ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.33 by root, Wed May 23 20:26:40 2007 UTC vs.
Revision 1.100 by root, Sat Sep 5 23:00:56 2009 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	// guarantees, 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
15	#define F_LATIN1 0x00000002UL	25	#define F_LATIN1 0x00000002UL
16	#define F_UTF8 0x00000004UL	26	#define F_UTF8 0x00000004UL
17	#define F_INDENT 0x00000008UL	27	#define F_INDENT 0x00000008UL
18	#define F_CANONICAL 0x00000010UL	28	#define F_CANONICAL 0x00000010UL
19	#define F_SPACE_BEFORE 0x00000020UL	29	#define F_SPACE_BEFORE 0x00000020UL
20	#define F_SPACE_AFTER 0x00000040UL	30	#define F_SPACE_AFTER 0x00000040UL
21	#define F_ALLOW_NONREF 0x00000100UL	31	#define F_ALLOW_NONREF 0x00000100UL
22	#define F_SHRINK 0x00000200UL	32	#define F_SHRINK 0x00000200UL
		33	#define F_ALLOW_BLESSED 0x00000400UL
		34	#define F_CONV_BLESSED 0x00000800UL
23	#define F_MAXDEPTH 0xf8000000UL	35	#define F_RELAXED 0x00001000UL
24	#define S_MAXDEPTH 27	36	#define F_ALLOW_UNKNOWN 0x00002000UL
25		37	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
26	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
27
28	// F_SELFCONVERT? <=> to_json/toJson
29	// F_BLESSED? <=> { $__class__$ => }
30		38
31	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	39	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
32	#define F_DEFAULT (9UL << S_MAXDEPTH)
33		40
34	#define INIT_SIZE 32 // initial scalar size to be allocated	41	#define INIT_SIZE 32 // initial scalar size to be allocated
35	#define INDENT_STEP 3 // spaces per indentation level	42	#define INDENT_STEP 3 // spaces per indentation level
36		43
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 <= 0 && (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	int 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;
		178	}
		179
		180	// convert offset pointer to character index, sv must be string
		181	static STRLEN
		182	ptr_to_index (SV *sv, char *offset)
		183	{
		184	return SvUTF8 (sv)
		185	? utf8_distance (offset, SvPVX (sv))
		186	: offset - SvPVX (sv);
90	}	187	}
91		188
92	/////////////////////////////////////////////////////////////////////////////	189	/////////////////////////////////////////////////////////////////////////////
93	// encoder	190	// encoder
94		191
…		…
96	typedef struct	193	typedef struct
97	{	194	{
98	char *cur; // SvPVX (sv) + current output position	195	char *cur; // SvPVX (sv) + current output position
99	char *end; // SvEND (sv)	196	char *end; // SvEND (sv)
100	SV *sv; // result scalar	197	SV *sv; // result scalar
101	U32 flags; // F_*	198	JSON json;
102	U32 indent; // indentation level	199	U32 indent; // indentation level
103	U32 maxdepth; // max. indentation/recursion level	200	UV limit; // escape character values >= this value when encoding
104	} enc_t;	201	} enc_t;
105		202
106	static void	203	INLINE void
107	need (enc_t *enc, STRLEN len)	204	need (enc_t *enc, STRLEN len)
108	{	205	{
109	if (enc->cur + len >= enc->end)	206	if (expect_false (enc->cur + len >= enc->end))
110	{	207	{
111	STRLEN cur = enc->cur - SvPVX (enc->sv);	208	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
112	SvGROW (enc->sv, cur + len + 1);	209	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
113	enc->cur = SvPVX (enc->sv) + cur;	210	enc->cur = SvPVX (enc->sv) + cur;
114	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	211	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
115	}	212	}
116	}	213	}
117		214
118	static void	215	INLINE void
119	encode_ch (enc_t *enc, char ch)	216	encode_ch (enc_t *enc, char ch)
120	{	217	{
121	need (enc, 1);	218	need (enc, 1);
122	*enc->cur++ = ch;	219	*enc->cur++ = ch;
123	}	220	}
…		…
131		228
132	while (str < end)	229	while (str < end)
133	{	230	{
134	unsigned char ch = *(unsigned char *)str;	231	unsigned char ch = *(unsigned char *)str;
135		232
136	if (ch >= 0x20 && ch < 0x80) // most common case	233	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
137	{	234	{
138	if (ch == '"') // but with slow exceptions	235	if (expect_false (ch == '"')) // but with slow exceptions
139	{	236	{
140	need (enc, len += 1);	237	need (enc, len += 1);
141	*enc->cur++ = '\\';	238	*enc->cur++ = '\\';
142	*enc->cur++ = '"';	239	*enc->cur++ = '"';
143	}	240	}
144	else if (ch == '\\')	241	else if (expect_false (ch == '\\'))
145	{	242	{
146	need (enc, len += 1);	243	need (enc, len += 1);
147	*enc->cur++ = '\\';	244	*enc->cur++ = '\\';
148	*enc->cur++ = '\\';	245	*enc->cur++ = '\\';
149	}	246	}
…		…
167	STRLEN clen;	264	STRLEN clen;
168	UV uch;	265	UV uch;
169		266
170	if (is_utf8)	267	if (is_utf8)
171	{	268	{
172	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
173	uch = decode_utf8 (str, end - str, &clen);	269	uch = decode_utf8 (str, end - str, &clen);
174	if (clen == (STRLEN)-1)	270	if (clen == (STRLEN)-1)
175	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	271	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
176	}	272	}
177	else	273	else
178	{	274	{
179	uch = ch;	275	uch = ch;
180	clen = 1;	276	clen = 1;
181	}	277	}
182		278
183	if (uch > 0x10FFFFUL)	279	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 || (enc->flags & F_LATIN1 && uch > 0xFF))
187	{	280	{
188	if (uch > 0xFFFFUL)	281	if (uch >= 0x10000UL)
189	{	282	{
		283	if (uch >= 0x110000UL)
		284	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		285
190	need (enc, len += 11);	286	need (enc, len += 11);
191	sprintf (enc->cur, "\\u%04x\\u%04x",	287	sprintf (enc->cur, "\\u%04x\\u%04x",
192	(int)((uch - 0x10000) / 0x400 + 0xD800),	288	(int)((uch - 0x10000) / 0x400 + 0xD800),
193	(int)((uch - 0x10000) % 0x400 + 0xDC00));	289	(int)((uch - 0x10000) % 0x400 + 0xDC00));
194	enc->cur += 12;	290	enc->cur += 12;
195	}	291	}
196	else	292	else
197	{	293	{
198	static char hexdigit [16] = "0123456789abcdef";
199	need (enc, len += 5);	294	need (enc, len += 5);
200	*enc->cur++ = '\\';	295	*enc->cur++ = '\\';
201	*enc->cur++ = 'u';	296	*enc->cur++ = 'u';
202	*enc->cur++ = hexdigit [ uch >> 12 ];	297	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
203	*enc->cur++ = hexdigit [(uch >> 8) & 15];	298	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
204	*enc->cur++ = hexdigit [(uch >> 4) & 15];	299	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
205	*enc->cur++ = hexdigit [(uch >> 0) & 15];	300	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
206	}	301	}
207		302
208	str += clen;	303	str += clen;
209	}	304	}
210	else if (enc->flags & F_LATIN1)	305	else if (enc->json.flags & F_LATIN1)
211	{	306	{
212	*enc->cur++ = uch;	307	*enc->cur++ = uch;
213	str += clen;	308	str += clen;
214	}	309	}
215	else if (is_utf8)	310	else if (is_utf8)
…		…
222	while (--clen);	317	while (--clen);
223	}	318	}
224	else	319	else
225	{	320	{
226	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	321	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
227	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	322	enc->cur = encode_utf8 (enc->cur, uch);
228	++str;	323	++str;
229	}	324	}
230	}	325	}
231	}	326	}
232	}	327	}
233		328
234	--len;	329	--len;
235	}	330	}
236	}	331	}
237		332
238	static void	333	INLINE void
239	encode_indent (enc_t *enc)	334	encode_indent (enc_t *enc)
240	{	335	{
241	if (enc->flags & F_INDENT)	336	if (enc->json.flags & F_INDENT)
242	{	337	{
243	int spaces = enc->indent * INDENT_STEP;	338	int spaces = enc->indent * INDENT_STEP;
244		339
245	need (enc, spaces);	340	need (enc, spaces);
246	memset (enc->cur, ' ', spaces);	341	memset (enc->cur, ' ', spaces);
247	enc->cur += spaces;	342	enc->cur += spaces;
248	}	343	}
249	}	344	}
250		345
251	static void	346	INLINE void
252	encode_space (enc_t *enc)	347	encode_space (enc_t *enc)
253	{	348	{
254	need (enc, 1);	349	need (enc, 1);
255	encode_ch (enc, ' ');	350	encode_ch (enc, ' ');
256	}	351	}
257		352
258	static void	353	INLINE void
259	encode_nl (enc_t *enc)	354	encode_nl (enc_t *enc)
260	{	355	{
261	if (enc->flags & F_INDENT)	356	if (enc->json.flags & F_INDENT)
262	{	357	{
263	need (enc, 1);	358	need (enc, 1);
264	encode_ch (enc, '\n');	359	encode_ch (enc, '\n');
265	}	360	}
266	}	361	}
267		362
268	static void	363	INLINE void
269	encode_comma (enc_t *enc)	364	encode_comma (enc_t *enc)
270	{	365	{
271	encode_ch (enc, ',');	366	encode_ch (enc, ',');
272		367
273	if (enc->flags & F_INDENT)	368	if (enc->json.flags & F_INDENT)
274	encode_nl (enc);	369	encode_nl (enc);
275	else if (enc->flags & F_SPACE_AFTER)	370	else if (enc->json.flags & F_SPACE_AFTER)
276	encode_space (enc);	371	encode_space (enc);
277	}	372	}
278		373
279	static void encode_sv (enc_t *enc, SV *sv);	374	static void encode_sv (enc_t *enc, SV *sv);
280		375
281	static void	376	static void
282	encode_av (enc_t *enc, AV *av)	377	encode_av (enc_t *enc, AV *av)
283	{	378	{
284	int i, len = av_len (av);	379	int i, len = av_len (av);
285		380
286	if (enc->indent >= enc->maxdepth)	381	if (enc->indent >= enc->json.max_depth)
287	croak ("data structure too deep (hit recursion limit)");	382	croak (ERR_NESTING_EXCEEDED);
288		383
289	encode_ch (enc, '['); encode_nl (enc);	384	encode_ch (enc, '[');
290	++enc->indent;	385
		386	if (len >= 0)
		387	{
		388	encode_nl (enc); ++enc->indent;
291		389
292	for (i = 0; i <= len; ++i)	390	for (i = 0; i <= len; ++i)
293	{	391	{
		392	SV **svp = av_fetch (av, i, 0);
		393
294	encode_indent (enc);	394	encode_indent (enc);
295	encode_sv (enc, *av_fetch (av, i, 0));
296		395
		396	if (svp)
		397	encode_sv (enc, *svp);
		398	else
		399	encode_str (enc, "null", 4, 0);
		400
297	if (i < len)	401	if (i < len)
298	encode_comma (enc);	402	encode_comma (enc);
299	}	403	}
300		404
		405	encode_nl (enc); --enc->indent; encode_indent (enc);
		406	}
		407
301	encode_nl (enc);	408	encode_ch (enc, ']');
302
303	--enc->indent;
304	encode_indent (enc); encode_ch (enc, ']');
305	}	409	}
306		410
307	static void	411	static void
308	encode_he (enc_t *enc, HE *he)	412	encode_hk (enc_t *enc, HE *he)
309	{	413	{
310	encode_ch (enc, '"');	414	encode_ch (enc, '"');
311		415
312	if (HeKLEN (he) == HEf_SVKEY)	416	if (HeKLEN (he) == HEf_SVKEY)
313	{	417	{
…		…
323	else	427	else
324	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	428	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
325		429
326	encode_ch (enc, '"');	430	encode_ch (enc, '"');
327		431
328	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	432	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
329	encode_ch (enc, ':');	433	encode_ch (enc, ':');
330	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	434	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
331	encode_sv (enc, HeVAL (he));
332	}	435	}
333		436
334	// compare hash entries, used when all keys are bytestrings	437	// compare hash entries, used when all keys are bytestrings
335	static int	438	static int
336	he_cmp_fast (const void *a_, const void *b_)	439	he_cmp_fast (const void *a_, const void *b_)
…		…
341	HE *b = *(HE **)b_;	444	HE *b = *(HE **)b_;
342		445
343	STRLEN la = HeKLEN (a);	446	STRLEN la = HeKLEN (a);
344	STRLEN lb = HeKLEN (b);	447	STRLEN lb = HeKLEN (b);
345		448
346	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	449	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
347	cmp = la - lb;	450	cmp = lb - la;
348		451
349	return cmp;	452	return cmp;
350	}	453	}
351		454
352	// compare hash entries, used when some keys are sv's or utf-x	455	// compare hash entries, used when some keys are sv's or utf-x
353	static int	456	static int
354	he_cmp_slow (const void *a, const void *b)	457	he_cmp_slow (const void *a, const void *b)
355	{	458	{
356	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	459	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
357	}	460	}
358		461
359	static void	462	static void
360	encode_hv (enc_t *enc, HV *hv)	463	encode_hv (enc_t *enc, HV *hv)
361	{	464	{
362	int count, i;	465	HE *he;
363		466
364	if (enc->indent >= enc->maxdepth)	467	if (enc->indent >= enc->json.max_depth)
365	croak ("data structure too deep (hit recursion limit)");	468	croak (ERR_NESTING_EXCEEDED);
366		469
367	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	470	encode_ch (enc, '{');
368		471
369	if ((count = hv_iterinit (hv)))
370	{
371	// for canonical output we have to sort by keys first	472	// for canonical output we have to sort by keys first
372	// actually, this is mostly due to the stupid so-called	473	// actually, this is mostly due to the stupid so-called
373	// security workaround added somewhere in 5.8.x.	474	// security workaround added somewhere in 5.8.x
374	// that randomises hash orderings	475	// that randomises hash orderings
375	if (enc->flags & F_CANONICAL)	476	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
		477	{
		478	int count = hv_iterinit (hv);
		479
		480	if (SvMAGICAL (hv))
376	{	481	{
		482	// need to count by iterating. could improve by dynamically building the vector below
		483	// but I don't care for the speed of this special case.
		484	// note also that we will run into undefined behaviour when the two iterations
		485	// do not result in the same count, something I might care for in some later release.
		486
		487	count = 0;
		488	while (hv_iternext (hv))
		489	++count;
		490
		491	hv_iterinit (hv);
		492	}
		493
		494	if (count)
		495	{
377	int fast = 1;	496	int i, fast = 1;
378	HE *he;	497	#if defined(__BORLANDC__) || defined(_MSC_VER)
379	#if WIN32
380	HE **hes = _alloca (count * sizeof (HE));	498	HE **hes = _alloca (count * sizeof (HE));
381	#else	499	#else
382	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode	500	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
383	#endif	501	#endif
384		502
…		…
410		528
411	FREETMPS;	529	FREETMPS;
412	LEAVE;	530	LEAVE;
413	}	531	}
414		532
415	for (i = 0; i < count; ++i)	533	encode_nl (enc); ++enc->indent;
		534
		535	while (count--)
416	{	536	{
417	encode_indent (enc);	537	encode_indent (enc);
		538	he = hes [count];
418	encode_he (enc, hes [i]);	539	encode_hk (enc, he);
		540	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
419		541
420	if (i < count - 1)	542	if (count)
421	encode_comma (enc);	543	encode_comma (enc);
422	}	544	}
423		545
		546	encode_nl (enc); --enc->indent; encode_indent (enc);
		547	}
		548	}
		549	else
		550	{
		551	if (hv_iterinit (hv) || SvMAGICAL (hv))
		552	if ((he = hv_iternext (hv)))
		553	{
		554	encode_nl (enc); ++enc->indent;
		555
		556	for (;;)
		557	{
424	encode_nl (enc);	558	encode_indent (enc);
		559	encode_hk (enc, he);
		560	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		561
		562	if (!(he = hv_iternext (hv)))
		563	break;
		564
		565	encode_comma (enc);
		566	}
		567
		568	encode_nl (enc); --enc->indent; encode_indent (enc);
		569	}
		570	}
		571
		572	encode_ch (enc, '}');
		573	}
		574
		575	// encode objects, arrays and special \0=false and \1=true values.
		576	static void
		577	encode_rv (enc_t *enc, SV *sv)
		578	{
		579	svtype svt;
		580
		581	SvGETMAGIC (sv);
		582	svt = SvTYPE (sv);
		583
		584	if (expect_false (SvOBJECT (sv)))
		585	{
		586	HV *stash = !JSON_SLOW || json_boolean_stash
		587	? json_boolean_stash
		588	: gv_stashpv ("JSON::XS::Boolean", 1);
		589
		590	if (SvSTASH (sv) == stash)
		591	{
		592	if (SvIV (sv))
		593	encode_str (enc, "true", 4, 0);
		594	else
		595	encode_str (enc, "false", 5, 0);
425	}	596	}
426	else	597	else
427	{	598	{
428	HE *he = hv_iternext (hv);	599	#if 0
429		600	if (0 && sv_derived_from (rv, "JSON::Literal"))
430	for (;;)
431	{	601	{
432	encode_indent (enc);	602	// not yet
433	encode_he (enc, he);
434
435	if (!(he = hv_iternext (hv)))
436	break;
437
438	encode_comma (enc);
439	}	603	}
		604	#endif
		605	if (enc->json.flags & F_CONV_BLESSED)
		606	{
		607	// we re-bless the reference to get overload and other niceties right
		608	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
440		609
		610	if (to_json)
		611	{
		612	dSP;
		613
		614	ENTER; SAVETMPS; PUSHMARK (SP);
		615	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		616
		617	// calling with G_SCALAR ensures that we always get a 1 return value
		618	PUTBACK;
		619	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		620	SPAGAIN;
		621
		622	// catch this surprisingly common error
		623	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		624	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		625
		626	sv = POPs;
		627	PUTBACK;
		628
441	encode_nl (enc);	629	encode_sv (enc, sv);
		630
		631	FREETMPS; LEAVE;
		632	}
		633	else if (enc->json.flags & F_ALLOW_BLESSED)
		634	encode_str (enc, "null", 4, 0);
		635	else
		636	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		637	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		638	}
		639	else if (enc->json.flags & F_ALLOW_BLESSED)
		640	encode_str (enc, "null", 4, 0);
		641	else
		642	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		643	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
442	}	644	}
443	}	645	}
444
445	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
446	}
447
448	// encode objects, arrays and special \0=false and \1=true values.
449	static void
450	encode_rv (enc_t *enc, SV *sv)
451	{
452	svtype svt;
453
454	SvGETMAGIC (sv);
455	svt = SvTYPE (sv);
456
457	if (svt == SVt_PVHV)	646	else if (svt == SVt_PVHV)
458	encode_hv (enc, (HV *)sv);	647	encode_hv (enc, (HV *)sv);
459	else if (svt == SVt_PVAV)	648	else if (svt == SVt_PVAV)
460	encode_av (enc, (AV *)sv);	649	encode_av (enc, (AV *)sv);
461	else if (svt < SVt_PVAV)	650	else if (svt < SVt_PVAV)
462	{	651	{
463	if (SvNIOK (sv) && SvIV (sv) == 0)	652	STRLEN len = 0;
		653	char *pv = svt ? SvPV (sv, len) : 0;
		654
		655	if (len == 1 && *pv == '1')
		656	encode_str (enc, "true", 4, 0);
		657	else if (len == 1 && *pv == '0')
464	encode_str (enc, "false", 5, 0);	658	encode_str (enc, "false", 5, 0);
465	else if (SvNIOK (sv) && SvIV (sv) == 1)	659	else if (enc->json.flags & F_ALLOW_UNKNOWN)
466	encode_str (enc, "true", 4, 0);	660	encode_str (enc, "null", 4, 0);
467	else	661	else
468	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	662	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
469	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	663	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
470	}	664	}
		665	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		666	encode_str (enc, "null", 4, 0);
471	else	667	else
472	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	668	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
473	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	669	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
474	}	670	}
475		671
…		…
486	encode_str (enc, str, len, SvUTF8 (sv));	682	encode_str (enc, str, len, SvUTF8 (sv));
487	encode_ch (enc, '"');	683	encode_ch (enc, '"');
488	}	684	}
489	else if (SvNOKp (sv))	685	else if (SvNOKp (sv))
490	{	686	{
		687	// trust that perl will do the right thing w.r.t. JSON syntax.
491	need (enc, NV_DIG + 32);	688	need (enc, NV_DIG + 32);
492	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	689	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
493	enc->cur += strlen (enc->cur);	690	enc->cur += strlen (enc->cur);
494	}	691	}
495	else if (SvIOKp (sv))	692	else if (SvIOKp (sv))
496	{	693	{
		694	// we assume we can always read an IV as a UV and vice versa
		695	// we assume two's complement
		696	// we assume no aliasing issues in the union
		697	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		698	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		699	{
		700	// optimise the "small number case"
		701	// code will likely be branchless and use only a single multiplication
		702	// works for numbers up to 59074
		703	I32 i = SvIVX (sv);
		704	U32 u;
		705	char digit, nz = 0;
		706
497	need (enc, 64);	707	need (enc, 6);
		708
		709	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		710	u = i < 0 ? -i : i;
		711
		712	// convert to 4.28 fixed-point representation
		713	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		714
		715	// now output digit by digit, each time masking out the integer part
		716	// and multiplying by 5 while moving the decimal point one to the right,
		717	// resulting in a net multiplication by 10.
		718	// we always write the digit to memory but conditionally increment
		719	// the pointer, to enable the use of conditional move instructions.
		720	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		721	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		722	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		723	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		724	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		725	}
		726	else
		727	{
		728	// large integer, use the (rather slow) snprintf way.
		729	need (enc, IVUV_MAXCHARS);
498	enc->cur +=	730	enc->cur +=
499	SvIsUV(sv)	731	SvIsUV(sv)
500	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	732	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
501	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	733	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		734	}
502	}	735	}
503	else if (SvROK (sv))	736	else if (SvROK (sv))
504	encode_rv (enc, SvRV (sv));	737	encode_rv (enc, SvRV (sv));
505	else if (!SvOK (sv))	738	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
506	encode_str (enc, "null", 4, 0);	739	encode_str (enc, "null", 4, 0);
507	else	740	else
508	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	741	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
509	SvPV_nolen (sv), SvFLAGS (sv));	742	SvPV_nolen (sv), SvFLAGS (sv));
510	}	743	}
511		744
512	static SV *	745	static SV *
513	encode_json (SV *scalar, U32 flags)	746	encode_json (SV *scalar, JSON *json)
514	{	747	{
515	enc_t enc;	748	enc_t enc;
516		749
517	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	750	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
518	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	751	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
519		752
520	enc.flags = flags;	753	enc.json = *json;
521	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	754	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
522	enc.cur = SvPVX (enc.sv);	755	enc.cur = SvPVX (enc.sv);
523	enc.end = SvEND (enc.sv);	756	enc.end = SvEND (enc.sv);
524	enc.indent = 0;	757	enc.indent = 0;
525	enc.maxdepth = DEC_DEPTH (flags);	758	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		759	: enc.json.flags & F_LATIN1 ? 0x000100UL
		760	: 0x110000UL;
526		761
527	SvPOK_only (enc.sv);	762	SvPOK_only (enc.sv);
528	encode_sv (&enc, scalar);	763	encode_sv (&enc, scalar);
		764	encode_nl (&enc);
529		765
530	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	766	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
531	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	767	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
532		768
533	if (!(flags & (F_ASCII | F_LATIN1 | F_UTF8)))	769	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
534	SvUTF8_on (enc.sv);	770	SvUTF8_on (enc.sv);
535		771
536	if (enc.flags & F_SHRINK)	772	if (enc.json.flags & F_SHRINK)
537	shrink (enc.sv);	773	shrink (enc.sv);
538		774
539	return enc.sv;	775	return enc.sv;
540	}	776	}
541		777
…		…
546	typedef struct	782	typedef struct
547	{	783	{
548	char *cur; // current parser pointer	784	char *cur; // current parser pointer
549	char *end; // end of input string	785	char *end; // end of input string
550	const char *err; // parse error, if != 0	786	const char *err; // parse error, if != 0
551	U32 flags; // F_*	787	JSON json;
552	U32 depth; // recursion depth	788	U32 depth; // recursion depth
553	U32 maxdepth; // recursion depth limit	789	U32 maxdepth; // recursion depth limit
554	} dec_t;	790	} dec_t;
555		791
556	static void	792	INLINE void
		793	decode_comment (dec_t *dec)
		794	{
		795	// only '#'-style comments allowed a.t.m.
		796
		797	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		798	++dec->cur;
		799	}
		800
		801	INLINE void
557	decode_ws (dec_t *dec)	802	decode_ws (dec_t *dec)
558	{	803	{
559	for (;;)	804	for (;;)
560	{	805	{
561	char ch = *dec->cur;	806	char ch = *dec->cur;
562		807
563	if (ch > 0x20	808	if (ch > 0x20)
		809	{
		810	if (expect_false (ch == '#'))
		811	{
		812	if (dec->json.flags & F_RELAXED)
		813	decode_comment (dec);
		814	else
		815	break;
		816	}
		817	else
		818	break;
		819	}
564	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	820	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
565	break;	821	break; // parse error, but let higher level handle it, gives better error messages
566		822
567	++dec->cur;	823	++dec->cur;
568	}	824	}
569	}	825	}
570		826
…		…
574	if (*dec->cur != ch) \	830	if (*dec->cur != ch) \
575	ERR (# ch " expected"); \	831	ERR (# ch " expected"); \
576	++dec->cur; \	832	++dec->cur; \
577	SE	833	SE
578		834
579	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	835	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
580	#define DEC_DEC_DEPTH --dec->depth	836	#define DEC_DEC_DEPTH --dec->depth
581		837
582	static SV *decode_sv (dec_t *dec);	838	static SV *decode_sv (dec_t *dec);
583		839
584	static signed char decode_hexdigit[256];	840	static signed char decode_hexdigit[256];
…		…
587	decode_4hex (dec_t *dec)	843	decode_4hex (dec_t *dec)
588	{	844	{
589	signed char d1, d2, d3, d4;	845	signed char d1, d2, d3, d4;
590	unsigned char *cur = (unsigned char *)dec->cur;	846	unsigned char *cur = (unsigned char *)dec->cur;
591		847
592	d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected");	848	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
593	d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected");	849	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
594	d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected");	850	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
595	d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected");	851	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
596		852
597	dec->cur += 4;	853	dec->cur += 4;
598		854
599	return ((UV)d1) << 12	855	return ((UV)d1) << 12
600	| ((UV)d2) << 8	856	| ((UV)d2) << 8
…		…
608	static SV *	864	static SV *
609	decode_str (dec_t *dec)	865	decode_str (dec_t *dec)
610	{	866	{
611	SV *sv = 0;	867	SV *sv = 0;
612	int utf8 = 0;	868	int utf8 = 0;
		869	char *dec_cur = dec->cur;
613		870
614	do	871	do
615	{	872	{
616	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];	873	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
617	char *cur = buf;	874	char *cur = buf;
618		875
619	do	876	do
620	{	877	{
621	unsigned char ch = *(unsigned char *)dec->cur++;	878	unsigned char ch = *(unsigned char *)dec_cur++;
622		879
623	if (ch == '"')	880	if (expect_false (ch == '"'))
624	{	881	{
625	--dec->cur;	882	--dec_cur;
626	break;	883	break;
627	}	884	}
628	else if (ch == '\\')	885	else if (expect_false (ch == '\\'))
629	{	886	{
630	switch (*dec->cur)	887	switch (*dec_cur)
631	{	888	{
632	case '\\':	889	case '\\':
633	case '/':	890	case '/':
634	case '"': *cur++ = *dec->cur++; break;	891	case '"': *cur++ = *dec_cur++; break;
635		892
636	case 'b': ++dec->cur; *cur++ = '\010'; break;	893	case 'b': ++dec_cur; *cur++ = '\010'; break;
637	case 't': ++dec->cur; *cur++ = '\011'; break;	894	case 't': ++dec_cur; *cur++ = '\011'; break;
638	case 'n': ++dec->cur; *cur++ = '\012'; break;	895	case 'n': ++dec_cur; *cur++ = '\012'; break;
639	case 'f': ++dec->cur; *cur++ = '\014'; break;	896	case 'f': ++dec_cur; *cur++ = '\014'; break;
640	case 'r': ++dec->cur; *cur++ = '\015'; break;	897	case 'r': ++dec_cur; *cur++ = '\015'; break;
641		898
642	case 'u':	899	case 'u':
643	{	900	{
644	UV lo, hi;	901	UV lo, hi;
645	++dec->cur;	902	++dec_cur;
646		903
		904	dec->cur = dec_cur;
647	hi = decode_4hex (dec);	905	hi = decode_4hex (dec);
		906	dec_cur = dec->cur;
648	if (hi == (UV)-1)	907	if (hi == (UV)-1)
649	goto fail;	908	goto fail;
650		909
651	// possibly a surrogate pair	910	// possibly a surrogate pair
652	if (hi >= 0xd800)	911	if (hi >= 0xd800)
653	if (hi < 0xdc00)	912	if (hi < 0xdc00)
654	{	913	{
655	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	914	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
656	ERR ("missing low surrogate character in surrogate pair");	915	ERR ("missing low surrogate character in surrogate pair");
657		916
658	dec->cur += 2;	917	dec_cur += 2;
659		918
		919	dec->cur = dec_cur;
660	lo = decode_4hex (dec);	920	lo = decode_4hex (dec);
		921	dec_cur = dec->cur;
661	if (lo == (UV)-1)	922	if (lo == (UV)-1)
662	goto fail;	923	goto fail;
663		924
664	if (lo < 0xdc00 || lo >= 0xe000)	925	if (lo < 0xdc00 || lo >= 0xe000)
665	ERR ("surrogate pair expected");	926	ERR ("surrogate pair expected");
…		…
671		932
672	if (hi >= 0x80)	933	if (hi >= 0x80)
673	{	934	{
674	utf8 = 1;	935	utf8 = 1;
675		936
676	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	937	cur = encode_utf8 (cur, hi);
677	}	938	}
678	else	939	else
679	*cur++ = hi;	940	*cur++ = hi;
680	}	941	}
681	break;	942	break;
682		943
683	default:	944	default:
684	--dec->cur;	945	--dec_cur;
685	ERR ("illegal backslash escape sequence in string");	946	ERR ("illegal backslash escape sequence in string");
686	}	947	}
687	}	948	}
688	else if (ch >= 0x20 && ch <= 0x7f)	949	else if (expect_true (ch >= 0x20 && ch < 0x80))
689	*cur++ = ch;	950	*cur++ = ch;
690	else if (ch >= 0x80)	951	else if (ch >= 0x80)
691	{	952	{
692	STRLEN clen;	953	STRLEN clen;
693	UV uch;	954	UV uch;
694		955
695	--dec->cur;	956	--dec_cur;
696		957
697	uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	958	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
698	if (clen == (STRLEN)-1)	959	if (clen == (STRLEN)-1)
699	ERR ("malformed UTF-8 character in JSON string");	960	ERR ("malformed UTF-8 character in JSON string");
700		961
701	do	962	do
702	*cur++ = *dec->cur++;	963	*cur++ = *dec_cur++;
703	while (--clen);	964	while (--clen);
704		965
705	utf8 = 1;	966	utf8 = 1;
706	}	967	}
707	else	968	else
708	{	969	{
709	--dec->cur;	970	--dec_cur;
710		971
711	if (!ch)	972	if (!ch)
712	ERR ("unexpected end of string while parsing JSON string");	973	ERR ("unexpected end of string while parsing JSON string");
713	else	974	else
714	ERR ("invalid character encountered while parsing JSON string");	975	ERR ("invalid character encountered while parsing JSON string");
…		…
719	{	980	{
720	STRLEN len = cur - buf;	981	STRLEN len = cur - buf;
721		982
722	if (sv)	983	if (sv)
723	{	984	{
724	SvGROW (sv, SvCUR (sv) + len + 1);	985	STRLEN cur = SvCUR (sv);
		986
		987	if (SvLEN (sv) <= cur + len)
		988	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		989
725	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	990	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
726	SvCUR_set (sv, SvCUR (sv) + len);	991	SvCUR_set (sv, SvCUR (sv) + len);
727	}	992	}
728	else	993	else
729	sv = newSVpvn (buf, len);	994	sv = newSVpvn (buf, len);
730	}	995	}
731	}	996	}
732	while (*dec->cur != '"');	997	while (*dec_cur != '"');
733		998
734	++dec->cur;	999	++dec_cur;
735		1000
736	if (sv)	1001	if (sv)
737	{	1002	{
738	SvPOK_only (sv);	1003	SvPOK_only (sv);
739	*SvEND (sv) = 0;	1004	*SvEND (sv) = 0;
…		…
742	SvUTF8_on (sv);	1007	SvUTF8_on (sv);
743	}	1008	}
744	else	1009	else
745	sv = newSVpvn ("", 0);	1010	sv = newSVpvn ("", 0);
746		1011
		1012	dec->cur = dec_cur;
747	return sv;	1013	return sv;
748		1014
749	fail:	1015	fail:
		1016	dec->cur = dec_cur;
750	return 0;	1017	return 0;
751	}	1018	}
752		1019
753	static SV *	1020	static SV *
754	decode_num (dec_t *dec)	1021	decode_num (dec_t *dec)
…		…
812	is_nv = 1;	1079	is_nv = 1;
813	}	1080	}
814		1081
815	if (!is_nv)	1082	if (!is_nv)
816	{	1083	{
817	UV uv;	1084	int len = dec->cur - start;
818	int numtype = grok_number (start, dec->cur - start, &uv);	1085
819	if (numtype & IS_NUMBER_IN_UV)	1086	// special case the rather common 1..5-digit-int case
820	if (numtype & IS_NUMBER_NEG)	1087	if (*start == '-')
		1088	switch (len)
821	{	1089	{
822	if (uv < (UV)IV_MIN)	1090	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
823	return newSViv (-(IV)uv);	1091	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
		1092	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1093	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1094	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
824	}	1095	}
		1096	else
		1097	switch (len)
		1098	{
		1099	case 1: return newSViv ( start [0] - '0' * 1);
		1100	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1101	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1102	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1103	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1104	}
		1105
		1106	{
		1107	UV uv;
		1108	int numtype = grok_number (start, len, &uv);
		1109	if (numtype & IS_NUMBER_IN_UV)
		1110	if (numtype & IS_NUMBER_NEG)
		1111	{
		1112	if (uv < (UV)IV_MIN)
		1113	return newSViv (-(IV)uv);
		1114	}
825	else	1115	else
826	return newSVuv (uv);	1116	return newSVuv (uv);
827	}	1117	}
828		1118
		1119	len -= *start == '-' ? 1 : 0;
		1120
		1121	// does not fit into IV or UV, try NV
		1122	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1123	#if defined (LDBL_DIG)
		1124	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1125	#endif
		1126	)
		1127	// fits into NV without loss of precision
		1128	return newSVnv (Atof (start));
		1129
		1130	// everything else fails, convert it to a string
		1131	return newSVpvn (start, dec->cur - start);
		1132	}
		1133
		1134	// loss of precision here
829	return newSVnv (Atof (start));	1135	return newSVnv (Atof (start));
830		1136
831	fail:	1137	fail:
832	return 0;	1138	return 0;
833	}	1139	}
…		…
863		1169
864	if (*dec->cur != ',')	1170	if (*dec->cur != ',')
865	ERR (", or ] expected while parsing array");	1171	ERR (", or ] expected while parsing array");
866		1172
867	++dec->cur;	1173	++dec->cur;
		1174
		1175	decode_ws (dec);
		1176
		1177	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1178	{
		1179	++dec->cur;
		1180	break;
		1181	}
868	}	1182	}
869		1183
870	DEC_DEC_DEPTH;	1184	DEC_DEC_DEPTH;
871	return newRV_noinc ((SV *)av);	1185	return newRV_noinc ((SV *)av);
872		1186
…		…
877	}	1191	}
878		1192
879	static SV *	1193	static SV *
880	decode_hv (dec_t *dec)	1194	decode_hv (dec_t *dec)
881	{	1195	{
		1196	SV *sv;
882	HV *hv = newHV ();	1197	HV *hv = newHV ();
883		1198
884	DEC_INC_DEPTH;	1199	DEC_INC_DEPTH;
885	decode_ws (dec);	1200	decode_ws (dec);
886		1201
887	if (*dec->cur == '}')	1202	if (*dec->cur == '}')
888	++dec->cur;	1203	++dec->cur;
889	else	1204	else
890	for (;;)	1205	for (;;)
891	{	1206	{
		1207	EXPECT_CH ('"');
		1208
		1209	// heuristic: assume that
		1210	// a) decode_str + hv_store_ent are abysmally slow.
		1211	// b) most hash keys are short, simple ascii text.
		1212	// => try to "fast-match" such strings to avoid
		1213	// the overhead of decode_str + hv_store_ent.
		1214	{
892	SV *key, *value;	1215	SV *value;
		1216	char *p = dec->cur;
		1217	char *e = p + 24; // only try up to 24 bytes
893		1218
894	decode_ws (dec); EXPECT_CH ('"');	1219	for (;;)
895
896	key = decode_str (dec);
897	if (!key)
898	goto fail;
899
900	decode_ws (dec); EXPECT_CH (':');
901
902	value = decode_sv (dec);
903	if (!value)
904	{	1220	{
		1221	// the >= 0x80 is false on most architectures
		1222	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1223	{
		1224	// slow path, back up and use decode_str
		1225	SV *key = decode_str (dec);
		1226	if (!key)
		1227	goto fail;
		1228
		1229	decode_ws (dec); EXPECT_CH (':');
		1230
		1231	decode_ws (dec);
		1232	value = decode_sv (dec);
		1233	if (!value)
		1234	{
		1235	SvREFCNT_dec (key);
		1236	goto fail;
		1237	}
		1238
		1239	hv_store_ent (hv, key, value, 0);
905	SvREFCNT_dec (key);	1240	SvREFCNT_dec (key);
		1241
		1242	break;
		1243	}
		1244	else if (*p == '"')
		1245	{
		1246	// fast path, got a simple key
		1247	char *key = dec->cur;
		1248	int len = p - key;
		1249	dec->cur = p + 1;
		1250
		1251	decode_ws (dec); EXPECT_CH (':');
		1252
		1253	decode_ws (dec);
		1254	value = decode_sv (dec);
		1255	if (!value)
906	goto fail;	1256	goto fail;
		1257
		1258	hv_store (hv, key, len, value, 0);
		1259
		1260	break;
		1261	}
		1262
		1263	++p;
907	}	1264	}
908		1265	}
909	hv_store_ent (hv, key, value, 0);
910	SvREFCNT_dec (key);
911		1266
912	decode_ws (dec);	1267	decode_ws (dec);
913		1268
914	if (*dec->cur == '}')	1269	if (*dec->cur == '}')
915	{	1270	{
…		…
919		1274
920	if (*dec->cur != ',')	1275	if (*dec->cur != ',')
921	ERR (", or } expected while parsing object/hash");	1276	ERR (", or } expected while parsing object/hash");
922		1277
923	++dec->cur;	1278	++dec->cur;
		1279
		1280	decode_ws (dec);
		1281
		1282	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1283	{
		1284	++dec->cur;
		1285	break;
		1286	}
924	}	1287	}
925		1288
926	DEC_DEC_DEPTH;	1289	DEC_DEC_DEPTH;
927	return newRV_noinc ((SV *)hv);	1290	sv = newRV_noinc ((SV *)hv);
		1291
		1292	// check filter callbacks
		1293	if (dec->json.flags & F_HOOK)
		1294	{
		1295	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1296	{
		1297	HE *cb, *he;
		1298
		1299	hv_iterinit (hv);
		1300	he = hv_iternext (hv);
		1301	hv_iterinit (hv);
		1302
		1303	// the next line creates a mortal sv each time its called.
		1304	// might want to optimise this for common cases.
		1305	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1306
		1307	if (cb)
		1308	{
		1309	dSP;
		1310	int count;
		1311
		1312	ENTER; SAVETMPS; PUSHMARK (SP);
		1313	XPUSHs (HeVAL (he));
		1314
		1315	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1316
		1317	if (count == 1)
		1318	{
		1319	sv = newSVsv (POPs);
		1320	FREETMPS; LEAVE;
		1321	return sv;
		1322	}
		1323
		1324	FREETMPS; LEAVE;
		1325	}
		1326	}
		1327
		1328	if (dec->json.cb_object)
		1329	{
		1330	dSP;
		1331	int count;
		1332
		1333	ENTER; SAVETMPS; PUSHMARK (SP);
		1334	XPUSHs (sv_2mortal (sv));
		1335
		1336	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1337
		1338	if (count == 1)
		1339	{
		1340	sv = newSVsv (POPs);
		1341	FREETMPS; LEAVE;
		1342	return sv;
		1343	}
		1344
		1345	SvREFCNT_inc (sv);
		1346	FREETMPS; LEAVE;
		1347	}
		1348	}
		1349
		1350	return sv;
928		1351
929	fail:	1352	fail:
930	SvREFCNT_dec (hv);	1353	SvREFCNT_dec (hv);
931	DEC_DEC_DEPTH;	1354	DEC_DEC_DEPTH;
932	return 0;	1355	return 0;
933	}	1356	}
934		1357
935	static SV *	1358	static SV *
936	decode_sv (dec_t *dec)	1359	decode_sv (dec_t *dec)
937	{	1360	{
938	decode_ws (dec);	1361	// the beauty of JSON: you need exactly one character lookahead
		1362	// to parse everything.
939	switch (*dec->cur)	1363	switch (*dec->cur)
940	{	1364	{
941	case '"': ++dec->cur; return decode_str (dec);	1365	case '"': ++dec->cur; return decode_str (dec);
942	case '[': ++dec->cur; return decode_av (dec);	1366	case '[': ++dec->cur; return decode_av (dec);
943	case '{': ++dec->cur; return decode_hv (dec);	1367	case '{': ++dec->cur; return decode_hv (dec);
944		1368
945	case '-':	1369	case '-':
946	case '0': case '1': case '2': case '3': case '4':	1370	case '0': case '1': case '2': case '3': case '4':
947	case '5': case '6': case '7': case '8': case '9':	1371	case '5': case '6': case '7': case '8': case '9':
948	return decode_num (dec);	1372	return decode_num (dec);
949		1373
950	case 't':	1374	case 't':
951	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1375	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
952	{	1376	{
953	dec->cur += 4;	1377	dec->cur += 4;
		1378	#if JSON_SLOW
		1379	json_true = get_bool ("JSON::XS::true");
		1380	#endif
954	return newSViv (1);	1381	return newSVsv (json_true);
955	}	1382	}
956	else	1383	else
957	ERR ("'true' expected");	1384	ERR ("'true' expected");
958		1385
959	break;	1386	break;
960		1387
961	case 'f':	1388	case 'f':
962	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1389	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
963	{	1390	{
964	dec->cur += 5;	1391	dec->cur += 5;
		1392	#if JSON_SLOW
		1393	json_false = get_bool ("JSON::XS::false");
		1394	#endif
965	return newSViv (0);	1395	return newSVsv (json_false);
966	}	1396	}
967	else	1397	else
968	ERR ("'false' expected");	1398	ERR ("'false' expected");
969		1399
970	break;	1400	break;
…		…
988	fail:	1418	fail:
989	return 0;	1419	return 0;
990	}	1420	}
991		1421
992	static SV *	1422	static SV *
993	decode_json (SV *string, U32 flags, UV *offset_return)	1423	decode_json (SV *string, JSON *json, char **offset_return)
994	{	1424	{
995	dec_t dec;	1425	dec_t dec;
996	UV offset;
997	SV *sv;	1426	SV *sv;
998		1427
		1428	/* work around bugs in 5.10 where manipulating magic values
		1429	* will perl ignore the magic in subsequent accesses
		1430	*/
999	SvGETMAGIC (string);	1431	/*SvGETMAGIC (string);*/
		1432	if (SvMAGICAL (string))
		1433	string = sv_2mortal (newSVsv (string));
		1434
1000	SvUPGRADE (string, SVt_PV);	1435	SvUPGRADE (string, SVt_PV);
1001		1436
		1437	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1438	* assertion with -DDEBUGGING, although SvCUR is documented to
		1439	* return the xpv_cur field which certainly exists after upgrading.
		1440	* according to nicholas clark, calling SvPOK fixes this.
		1441	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1442	* and hope for the best.
		1443	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1444	* assertion business is seriously broken, try yet another workaround
		1445	* for the broken -DDEBUGGING.
		1446	*/
		1447	{
		1448	#ifdef DEBUGGING
		1449	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1450	#else
		1451	STRLEN offset = SvCUR (string);
		1452	#endif
		1453
		1454	if (offset > json->max_size && json->max_size)
		1455	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1456	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1457	}
		1458
1002	if (flags & F_UTF8)	1459	if (json->flags & F_UTF8)
1003	sv_utf8_downgrade (string, 0);	1460	sv_utf8_downgrade (string, 0);
1004	else	1461	else
1005	sv_utf8_upgrade (string);	1462	sv_utf8_upgrade (string);
1006		1463
1007	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1464	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1008		1465
1009	dec.flags = flags;	1466	dec.json = *json;
1010	dec.cur = SvPVX (string);	1467	dec.cur = SvPVX (string);
1011	dec.end = SvEND (string);	1468	dec.end = SvEND (string);
1012	dec.err = 0;	1469	dec.err = 0;
1013	dec.depth = 0;	1470	dec.depth = 0;
1014	dec.maxdepth = DEC_DEPTH (dec.flags);	1471
		1472	if (dec.json.cb_object || dec.json.cb_sk_object)
		1473	dec.json.flags |= F_HOOK;
1015		1474
1016	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1475	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1476
		1477	decode_ws (&dec);
1017	sv = decode_sv (&dec);	1478	sv = decode_sv (&dec);
		1479
		1480	if (offset_return)
		1481	*offset_return = dec.cur;
1018		1482
1019	if (!(offset_return || !sv))	1483	if (!(offset_return || !sv))
1020	{	1484	{
1021	// check for trailing garbage	1485	// check for trailing garbage
1022	decode_ws (&dec);	1486	decode_ws (&dec);
…		…
1025	{	1489	{
1026	dec.err = "garbage after JSON object";	1490	dec.err = "garbage after JSON object";
1027	SvREFCNT_dec (sv);	1491	SvREFCNT_dec (sv);
1028	sv = 0;	1492	sv = 0;
1029	}	1493	}
1030	}
1031
1032	if (offset_return || !sv)
1033	{
1034	offset = dec.flags & F_UTF8
1035	? dec.cur - SvPVX (string)
1036	: utf8_distance (dec.cur, SvPVX (string));
1037
1038	if (offset_return)
1039	*offset_return = offset;
1040	}	1494	}
1041		1495
1042	if (!sv)	1496	if (!sv)
1043	{	1497	{
1044	SV *uni = sv_newmortal ();	1498	SV *uni = sv_newmortal ();
…		…
1050	SAVEVPTR (PL_curcop);	1504	SAVEVPTR (PL_curcop);
1051	PL_curcop = &cop;	1505	PL_curcop = &cop;
1052	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1506	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1053	LEAVE;	1507	LEAVE;
1054		1508
1055	croak ("%s, at character offset %d [\"%s\"]",	1509	croak ("%s, at character offset %d (before \"%s\")",
1056	dec.err,	1510	dec.err,
1057	(int)offset,	1511	ptr_to_index (string, dec.cur),
1058	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1512	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1059	}	1513	}
1060		1514
1061	sv = sv_2mortal (sv);	1515	sv = sv_2mortal (sv);
1062		1516
1063	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1517	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1064	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1518	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1065		1519
1066	return sv;	1520	return sv;
		1521	}
		1522
		1523	/////////////////////////////////////////////////////////////////////////////
		1524	// incremental parser
		1525
		1526	static void
		1527	incr_parse (JSON *self)
		1528	{
		1529	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1530
		1531	for (;;)
		1532	{
		1533	//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
		1534	switch (self->incr_mode)
		1535	{
		1536	// only used for intiial whitespace skipping
		1537	case INCR_M_WS:
		1538	for (;;)
		1539	{
		1540	if (*p > 0x20)
		1541	{
		1542	self->incr_mode = INCR_M_JSON;
		1543	goto incr_m_json;
		1544	}
		1545	else if (!*p)
		1546	goto interrupt;
		1547
		1548	++p;
		1549	}
		1550
		1551	// skip a single char inside a string (for \\-processing)
		1552	case INCR_M_BS:
		1553	if (!*p)
		1554	goto interrupt;
		1555
		1556	++p;
		1557	self->incr_mode = INCR_M_STR;
		1558	goto incr_m_str;
		1559
		1560	// inside a string
		1561	case INCR_M_STR:
		1562	incr_m_str:
		1563	for (;;)
		1564	{
		1565	if (*p == '"')
		1566	{
		1567	++p;
		1568	self->incr_mode = INCR_M_JSON;
		1569
		1570	if (!self->incr_nest)
		1571	goto interrupt;
		1572
		1573	goto incr_m_json;
		1574	}
		1575	else if (*p == '\\')
		1576	{
		1577	++p; // "virtually" consumes character after \
		1578
		1579	if (!*p) // if at end of string we have to switch modes
		1580	{
		1581	self->incr_mode = INCR_M_BS;
		1582	goto interrupt;
		1583	}
		1584	}
		1585	else if (!*p)
		1586	goto interrupt;
		1587
		1588	++p;
		1589	}
		1590
		1591	// after initial ws, outside string
		1592	case INCR_M_JSON:
		1593	incr_m_json:
		1594	for (;;)
		1595	{
		1596	switch (*p++)
		1597	{
		1598	case 0:
		1599	--p;
		1600	goto interrupt;
		1601
		1602	case 0x09:
		1603	case 0x0a:
		1604	case 0x0d:
		1605	case 0x20:
		1606	if (!self->incr_nest)
		1607	{
		1608	--p; // do not eat the whitespace, let the next round do it
		1609	goto interrupt;
		1610	}
		1611	break;
		1612
		1613	case '"':
		1614	self->incr_mode = INCR_M_STR;
		1615	goto incr_m_str;
		1616
		1617	case '[':
		1618	case '{':
		1619	if (++self->incr_nest > self->max_depth)
		1620	croak (ERR_NESTING_EXCEEDED);
		1621	break;
		1622
		1623	case ']':
		1624	case '}':
		1625	if (--self->incr_nest <= 0)
		1626	goto interrupt;
		1627	}
		1628	}
		1629	}
		1630
		1631	modechange:
		1632	;
		1633	}
		1634
		1635	interrupt:
		1636	self->incr_pos = p - SvPVX (self->incr_text);
		1637	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1067	}	1638	}
1068		1639
1069	/////////////////////////////////////////////////////////////////////////////	1640	/////////////////////////////////////////////////////////////////////////////
1070	// XS interface functions	1641	// XS interface functions
1071		1642
1072	MODULE = JSON::XS PACKAGE = JSON::XS	1643	MODULE = JSON::XS PACKAGE = JSON::XS
1073		1644
1074	BOOT:	1645	BOOT:
1075	{	1646	{
1076	int i;	1647	int i;
1077
1078	memset (decode_hexdigit, 0xff, 256);
1079		1648
1080	for (i = 0; i < 256; ++i)	1649	for (i = 0; i < 256; ++i)
1081	decode_hexdigit [i] =	1650	decode_hexdigit [i] =
1082	i >= '0' && i <= '9' ? i - '0'	1651	i >= '0' && i <= '9' ? i - '0'
1083	: i >= 'a' && i <= 'f' ? i - 'a' + 10	1652	: i >= 'a' && i <= 'f' ? i - 'a' + 10
1084	: i >= 'A' && i <= 'F' ? i - 'A' + 10	1653	: i >= 'A' && i <= 'F' ? i - 'A' + 10
1085	: -1;	1654	: -1;
1086		1655
1087	json_stash = gv_stashpv ("JSON::XS", 1);	1656	json_stash = gv_stashpv ("JSON::XS" , 1);
		1657	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1658
		1659	json_true = get_bool ("JSON::XS::true");
		1660	json_false = get_bool ("JSON::XS::false");
		1661
		1662	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1088	}	1663	}
1089		1664
1090	PROTOTYPES: DISABLE	1665	PROTOTYPES: DISABLE
1091		1666
1092	SV *new (char *dummy)	1667	void CLONE (...)
1093	CODE:	1668	CODE:
1094	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1669	json_stash = 0;
		1670	json_boolean_stash = 0;
		1671
		1672	void new (char *klass)
		1673	PPCODE:
		1674	{
		1675	SV *pv = NEWSV (0, sizeof (JSON));
		1676	SvPOK_only (pv);
		1677	json_init ((JSON *)SvPVX (pv));
		1678	XPUSHs (sv_2mortal (sv_bless (
		1679	newRV_noinc (pv),
		1680	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1681	)));
		1682	}
		1683
		1684	void ascii (JSON *self, int enable = 1)
		1685	ALIAS:
		1686	ascii = F_ASCII
		1687	latin1 = F_LATIN1
		1688	utf8 = F_UTF8
		1689	indent = F_INDENT
		1690	canonical = F_CANONICAL
		1691	space_before = F_SPACE_BEFORE
		1692	space_after = F_SPACE_AFTER
		1693	pretty = F_PRETTY
		1694	allow_nonref = F_ALLOW_NONREF
		1695	shrink = F_SHRINK
		1696	allow_blessed = F_ALLOW_BLESSED
		1697	convert_blessed = F_CONV_BLESSED
		1698	relaxed = F_RELAXED
		1699	allow_unknown = F_ALLOW_UNKNOWN
		1700	PPCODE:
		1701	{
		1702	if (enable)
		1703	self->flags |= ix;
		1704	else
		1705	self->flags &= ~ix;
		1706
		1707	XPUSHs (ST (0));
		1708	}
		1709
		1710	void get_ascii (JSON *self)
		1711	ALIAS:
		1712	get_ascii = F_ASCII
		1713	get_latin1 = F_LATIN1
		1714	get_utf8 = F_UTF8
		1715	get_indent = F_INDENT
		1716	get_canonical = F_CANONICAL
		1717	get_space_before = F_SPACE_BEFORE
		1718	get_space_after = F_SPACE_AFTER
		1719	get_allow_nonref = F_ALLOW_NONREF
		1720	get_shrink = F_SHRINK
		1721	get_allow_blessed = F_ALLOW_BLESSED
		1722	get_convert_blessed = F_CONV_BLESSED
		1723	get_relaxed = F_RELAXED
		1724	get_allow_unknown = F_ALLOW_UNKNOWN
		1725	PPCODE:
		1726	XPUSHs (boolSV (self->flags & ix));
		1727
		1728	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		1729	PPCODE:
		1730	self->max_depth = max_depth;
		1731	XPUSHs (ST (0));
		1732
		1733	U32 get_max_depth (JSON *self)
		1734	CODE:
		1735	RETVAL = self->max_depth;
1095	OUTPUT:	1736	OUTPUT:
1096	RETVAL	1737	RETVAL
1097		1738
1098	SV *ascii (SV *self, int enable = 1)	1739	void max_size (JSON *self, U32 max_size = 0)
1099	ALIAS:	1740	PPCODE:
1100	ascii = F_ASCII	1741	self->max_size = max_size;
1101	latin1 = F_LATIN1	1742	XPUSHs (ST (0));
1102	utf8 = F_UTF8	1743
1103	indent = F_INDENT	1744	int get_max_size (JSON *self)
1104	canonical = F_CANONICAL
1105	space_before = F_SPACE_BEFORE
1106	space_after = F_SPACE_AFTER
1107	pretty = F_PRETTY
1108	allow_nonref = F_ALLOW_NONREF
1109	shrink = F_SHRINK
1110	CODE:	1745	CODE:
1111	{	1746	RETVAL = self->max_size;
1112	UV *uv = SvJSON (self);
1113	if (enable)
1114	*uv |= ix;
1115	else
1116	*uv &= ~ix;
1117
1118	RETVAL = newSVsv (self);
1119	}
1120	OUTPUT:	1747	OUTPUT:
1121	RETVAL	1748	RETVAL
1122		1749
1123	SV *max_depth (SV *self, UV max_depth = 0x80000000UL)	1750	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
		1751	PPCODE:
		1752	{
		1753	SvREFCNT_dec (self->cb_object);
		1754	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1755
		1756	XPUSHs (ST (0));
		1757	}
		1758
		1759	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
		1760	PPCODE:
		1761	{
		1762	if (!self->cb_sk_object)
		1763	self->cb_sk_object = newHV ();
		1764
		1765	if (SvOK (cb))
		1766	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1767	else
		1768	{
		1769	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1770
		1771	if (!HvKEYS (self->cb_sk_object))
		1772	{
		1773	SvREFCNT_dec (self->cb_sk_object);
		1774	self->cb_sk_object = 0;
		1775	}
		1776	}
		1777
		1778	XPUSHs (ST (0));
		1779	}
		1780
		1781	void encode (JSON *self, SV *scalar)
		1782	PPCODE:
		1783	XPUSHs (encode_json (scalar, self));
		1784
		1785	void decode (JSON *self, SV *jsonstr)
		1786	PPCODE:
		1787	XPUSHs (decode_json (jsonstr, self, 0));
		1788
		1789	void decode_prefix (JSON *self, SV *jsonstr)
		1790	PPCODE:
		1791	{
		1792	char *offset;
		1793	EXTEND (SP, 2);
		1794	PUSHs (decode_json (jsonstr, self, &offset));
		1795	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		1796	}
		1797
		1798	void incr_parse (JSON *self, SV *jsonstr = 0)
		1799	PPCODE:
		1800	{
		1801	if (!self->incr_text)
		1802	self->incr_text = newSVpvn ("", 0);
		1803
		1804	// append data, if any
		1805	if (jsonstr)
		1806	{
		1807	if (SvUTF8 (jsonstr))
		1808	{
		1809	if (!SvUTF8 (self->incr_text))
		1810	{
		1811	/* utf-8-ness differs, need to upgrade */
		1812	sv_utf8_upgrade (self->incr_text);
		1813
		1814	if (self->incr_pos)
		1815	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1816	- (U8 *)SvPVX (self->incr_text);
		1817	}
		1818	}
		1819	else if (SvUTF8 (self->incr_text))
		1820	sv_utf8_upgrade (jsonstr);
		1821
		1822	{
		1823	STRLEN len;
		1824	const char *str = SvPV (jsonstr, len);
		1825	STRLEN cur = SvCUR (self->incr_text);
		1826
		1827	if (SvLEN (self->incr_text) <= cur + len)
		1828	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1829
		1830	Move (str, SvEND (self->incr_text), len, char);
		1831	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1832	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1833	}
		1834	}
		1835
		1836	if (GIMME_V != G_VOID)
		1837	do
		1838	{
		1839	char *offset;
		1840
		1841	if (!INCR_DONE (self))
		1842	{
		1843	incr_parse (self);
		1844
		1845	if (self->incr_pos > self->max_size && self->max_size)
		1846	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1847	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1848
		1849	if (!INCR_DONE (self))
		1850	break;
		1851	}
		1852
		1853	XPUSHs (decode_json (self->incr_text, self, &offset));
		1854
		1855	self->incr_pos -= offset - SvPVX (self->incr_text);
		1856	self->incr_nest = 0;
		1857	self->incr_mode = 0;
		1858
		1859	sv_chop (self->incr_text, offset);
		1860	}
		1861	while (GIMME_V == G_ARRAY);
		1862	}
		1863
		1864	SV *incr_text (JSON *self)
		1865	ATTRS: lvalue
1124	CODE:	1866	CODE:
1125	{	1867	{
1126	UV *uv = SvJSON (self);	1868	if (self->incr_pos)
1127	UV log2 = 0;	1869	croak ("incr_text can not be called when the incremental parser already started parsing");
1128		1870
1129	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	1871	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
1130
1131	while ((1UL << log2) < max_depth)
1132	++log2;
1133
1134	*uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1135
1136	RETVAL = newSVsv (self);
1137	}	1872	}
1138	OUTPUT:	1873	OUTPUT:
1139	RETVAL	1874	RETVAL
1140		1875
		1876	void incr_skip (JSON *self)
		1877	CODE:
		1878	{
		1879	if (self->incr_pos)
		1880	{
		1881	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1882	self->incr_pos = 0;
		1883	self->incr_nest = 0;
		1884	self->incr_mode = 0;
		1885	}
		1886	}
		1887
		1888	void incr_reset (JSON *self)
		1889	CODE:
		1890	{
		1891	SvREFCNT_dec (self->incr_text);
		1892	self->incr_text = 0;
		1893	self->incr_pos = 0;
		1894	self->incr_nest = 0;
		1895	self->incr_mode = 0;
		1896	}
		1897
		1898	void DESTROY (JSON *self)
		1899	CODE:
		1900	SvREFCNT_dec (self->cb_sk_object);
		1901	SvREFCNT_dec (self->cb_object);
		1902	SvREFCNT_dec (self->incr_text);
		1903
		1904	PROTOTYPES: ENABLE
		1905
1141	void encode (SV *self, SV *scalar)	1906	void encode_json (SV *scalar)
		1907	ALIAS:
		1908	to_json_ = 0
		1909	encode_json = F_UTF8
1142	PPCODE:	1910	PPCODE:
		1911	{
		1912	JSON json;
		1913	json_init (&json);
		1914	json.flags |= ix;
1143	XPUSHs (encode_json (scalar, *SvJSON (self)));	1915	XPUSHs (encode_json (scalar, &json));
		1916	}
1144		1917
1145	void decode (SV *self, SV *jsonstr)	1918	void decode_json (SV *jsonstr)
		1919	ALIAS:
		1920	from_json_ = 0
		1921	decode_json = F_UTF8
1146	PPCODE:	1922	PPCODE:
		1923	{
		1924	JSON json;
		1925	json_init (&json);
		1926	json.flags |= ix;
1147	XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));	1927	XPUSHs (decode_json (jsonstr, &json, 0));
1148
1149	void decode_prefix (SV *self, SV *jsonstr)
1150	PPCODE:
1151	{
1152	UV offset;
1153	EXTEND (SP, 2);
1154	PUSHs (decode_json (jsonstr, *SvJSON (self), &offset));
1155	PUSHs (sv_2mortal (newSVuv (offset)));
1156	}	1928	}
1157		1929
1158	PROTOTYPES: ENABLE
1159		1930
1160	void to_json (SV *scalar)
1161	ALIAS:
1162	objToJson = 0
1163	PPCODE:
1164	XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8));
1165
1166	void from_json (SV *jsonstr)
1167	ALIAS:
1168	jsonToObj = 0
1169	PPCODE:
1170	XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8, 0));
1171

Diff Legend

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