ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.11 by root, Sat Mar 24 19:42:14 2007 UTC vs.
Revision 1.95 by root, Sun Feb 22 06:55:28 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>
		9	#include <limits.h>
		10	#include <float.h>
8		11
		12	#if defined(__BORLANDC__) || defined(_MSC_VER)
		13	# define snprintf _snprintf // C compilers have this in stdio.h
		14	#endif
		15
		16	// some old perls do not have this, try to make it work, no
		17	// guarentees, though. if it breaks, you get to keep the pieces.
		18	#ifndef UTF8_MAXBYTES
		19	# define UTF8_MAXBYTES 13
		20	#endif
		21
		22	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2)
		23
9	#define F_ASCII 0x00000001	24	#define F_ASCII 0x00000001UL
		25	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	26	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	27	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	28	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	29	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	30	#define F_SPACE_AFTER 0x00000040UL
15	#define F_ALLOW_NONREF 0x00000080	31	#define F_ALLOW_NONREF 0x00000100UL
16	#define F_SHRINK 0x00000100	32	#define F_SHRINK 0x00000200UL
		33	#define F_ALLOW_BLESSED 0x00000400UL
		34	#define F_CONV_BLESSED 0x00000800UL
		35	#define F_RELAXED 0x00001000UL
		36	#define F_ALLOW_UNKNOWN 0x00002000UL
		37	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
17		38
18	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	39	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
19	#define F_DEFAULT 0
20		40
21	#define INIT_SIZE 32 // initial scalar size to be allocated	41	#define INIT_SIZE 32 // initial scalar size to be allocated
		42	#define INDENT_STEP 3 // spaces per indentation level
		43
		44	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
22		45
23	#define SB do {	46	#define SB do {
24	#define SE } while (0)	47	#define SE } while (0)
25		48
26	static HV *json_stash;	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
		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	}
		107
		108	/////////////////////////////////////////////////////////////////////////////
		109	// utility functions
		110
		111	INLINE SV *
		112	get_bool (const char *name)
		113	{
		114	SV *sv = get_sv (name, 1);
		115
		116	SvREADONLY_on (sv);
		117	SvREADONLY_on (SvRV (sv));
		118
		119	return sv;
		120	}
		121
		122	INLINE void
		123	shrink (SV *sv)
		124	{
		125	sv_utf8_downgrade (sv, 1);
		126
		127	if (SvLEN (sv) > SvCUR (sv) + 1)
		128	{
		129	#ifdef SvPV_shrink_to_cur
		130	SvPV_shrink_to_cur (sv);
		131	#elif defined (SvPV_renew)
		132	SvPV_renew (sv, SvCUR (sv) + 1);
		133	#endif
		134	}
		135	}
		136
		137	// decode an utf-8 character and return it, or (UV)-1 in
		138	// case of an error.
		139	// we special-case "safe" characters from U+80 .. U+7FF,
		140	// but use the very good perl function to parse anything else.
		141	// note that we never call this function for a ascii codepoints
		142	INLINE UV
		143	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		144	{
		145	if (expect_true (len >= 2
		146	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
		147	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
		148	{
		149	*clen = 2;
		150	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		151	}
		152	else
		153	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		154	}
		155
		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	/////////////////////////////////////////////////////////////////////////////
		181	// encoder
27		182
28	// structure used for encoding JSON	183	// structure used for encoding JSON
29	typedef struct	184	typedef struct
30	{	185	{
31	char *cur;	186	char *cur; // SvPVX (sv) + current output position
32	STRLEN len; // SvLEN (sv)
33	char *end; // SvEND (sv)	187	char *end; // SvEND (sv)
34	SV *sv;	188	SV *sv; // result scalar
35	UV flags;	189	JSON json;
36	int max_recurse;	190	U32 indent; // indentation level
37	int indent;	191	UV limit; // escape character values >= this value when encoding
38	} enc_t;	192	} enc_t;
39		193
40	// structure used for decoding JSON	194	INLINE void
41	typedef struct
42	{
43	char *cur;
44	char *end;
45	const char *err;
46	UV flags;
47	} dec_t;
48
49	static UV *
50	SvJSON (SV *sv)
51	{
52	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
53	croak ("object is not of type JSON::XS");
54
55	return &SvUVX (SvRV (sv));
56	}
57
58	static void
59	shrink (SV *sv)
60	{
61	sv_utf8_downgrade (sv, 1);
62	#ifdef SvPV_shrink_to_cur
63	SvPV_shrink_to_cur (sv);
64	#endif
65	}
66
67	/////////////////////////////////////////////////////////////////////////////
68
69	static void
70	need (enc_t *enc, STRLEN len)	195	need (enc_t *enc, STRLEN len)
71	{	196	{
72	if (enc->cur + len >= enc->end)	197	if (expect_false (enc->cur + len >= enc->end))
73	{	198	{
74	STRLEN cur = enc->cur - SvPVX (enc->sv);	199	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
75	SvGROW (enc->sv, cur + len + 1);	200	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
76	enc->cur = SvPVX (enc->sv) + cur;	201	enc->cur = SvPVX (enc->sv) + cur;
77	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	202	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
78	}	203	}
79	}	204	}
80		205
81	static void	206	INLINE void
82	encode_ch (enc_t *enc, char ch)	207	encode_ch (enc_t *enc, char ch)
83	{	208	{
84	need (enc, 1);	209	need (enc, 1);
85	*enc->cur++ = ch;	210	*enc->cur++ = ch;
86	}	211	}
…		…
94		219
95	while (str < end)	220	while (str < end)
96	{	221	{
97	unsigned char ch = *(unsigned char *)str;	222	unsigned char ch = *(unsigned char *)str;
98		223
99	if (ch >= 0x20 && ch < 0x80) // most common case	224	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
100	{	225	{
101	if (ch == '"') // but with slow exceptions	226	if (expect_false (ch == '"')) // but with slow exceptions
102	{	227	{
103	need (enc, len += 1);	228	need (enc, len += 1);
104	*enc->cur++ = '\\';	229	*enc->cur++ = '\\';
105	*enc->cur++ = '"';	230	*enc->cur++ = '"';
106	}	231	}
107	else if (ch == '\\')	232	else if (expect_false (ch == '\\'))
108	{	233	{
109	need (enc, len += 1);	234	need (enc, len += 1);
110	*enc->cur++ = '\\';	235	*enc->cur++ = '\\';
111	*enc->cur++ = '\\';	236	*enc->cur++ = '\\';
112	}	237	}
…		…
130	STRLEN clen;	255	STRLEN clen;
131	UV uch;	256	UV uch;
132		257
133	if (is_utf8)	258	if (is_utf8)
134	{	259	{
135	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	260	uch = decode_utf8 (str, end - str, &clen);
136	if (clen == (STRLEN)-1)	261	if (clen == (STRLEN)-1)
137	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);
138	}	263	}
139	else	264	else
140	{	265	{
141	uch = ch;	266	uch = ch;
142	clen = 1;	267	clen = 1;
143	}	268	}
144		269
145	if (uch > 0x10FFFFUL)	270	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
146	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
147
148	if (uch < 0x80 || enc->flags & F_ASCII)
149	{	271	{
150	if (uch > 0xFFFFUL)	272	if (uch >= 0x10000UL)
151	{	273	{
		274	if (uch >= 0x110000UL)
		275	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		276
152	need (enc, len += 11);	277	need (enc, len += 11);
153	sprintf (enc->cur, "\\u%04x\\u%04x",	278	sprintf (enc->cur, "\\u%04x\\u%04x",
154	(int)((uch - 0x10000) / 0x400 + 0xD800),	279	(int)((uch - 0x10000) / 0x400 + 0xD800),
155	(int)((uch - 0x10000) % 0x400 + 0xDC00));	280	(int)((uch - 0x10000) % 0x400 + 0xDC00));
156	enc->cur += 12;	281	enc->cur += 12;
157	}	282	}
158	else	283	else
159	{	284	{
160	static char hexdigit [16] = "0123456789abcdef";
161	need (enc, len += 5);	285	need (enc, len += 5);
162	*enc->cur++ = '\\';	286	*enc->cur++ = '\\';
163	*enc->cur++ = 'u';	287	*enc->cur++ = 'u';
164	*enc->cur++ = hexdigit [ uch >> 12 ];	288	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
165	*enc->cur++ = hexdigit [(uch >> 8) & 15];	289	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
166	*enc->cur++ = hexdigit [(uch >> 4) & 15];	290	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
167	*enc->cur++ = hexdigit [(uch >> 0) & 15];	291	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
168	}	292	}
169		293
		294	str += clen;
		295	}
		296	else if (enc->json.flags & F_LATIN1)
		297	{
		298	*enc->cur++ = uch;
170	str += clen;	299	str += clen;
171	}	300	}
172	else if (is_utf8)	301	else if (is_utf8)
173	{	302	{
174	need (enc, len += clen);	303	need (enc, len += clen);
…		…
178	}	307	}
179	while (--clen);	308	while (--clen);
180	}	309	}
181	else	310	else
182	{	311	{
183	need (enc, len += 10); // never more than 11 bytes needed	312	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
184	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	313	enc->cur = encode_utf8 (enc->cur, uch);
185	++str;	314	++str;
186	}	315	}
187	}	316	}
188	}	317	}
189	}	318	}
190		319
191	--len;	320	--len;
192	}	321	}
193	}	322	}
194		323
195	#define INDENT SB \	324	INLINE void
		325	encode_indent (enc_t *enc)
		326	{
196	if (enc->flags & F_INDENT) \	327	if (enc->json.flags & F_INDENT)
197	{ \	328	{
198	int i_; \	329	int spaces = enc->indent * INDENT_STEP;
199	need (enc, enc->indent); \	330
200	for (i_ = enc->indent * 3; i_--; )\	331	need (enc, spaces);
		332	memset (enc->cur, ' ', spaces);
		333	enc->cur += spaces;
		334	}
		335	}
		336
		337	INLINE void
		338	encode_space (enc_t *enc)
		339	{
		340	need (enc, 1);
		341	encode_ch (enc, ' ');
		342	}
		343
		344	INLINE void
		345	encode_nl (enc_t *enc)
		346	{
		347	if (enc->json.flags & F_INDENT)
		348	{
		349	need (enc, 1);
201	encode_ch (enc, ' '); \	350	encode_ch (enc, '\n');
202	} \	351	}
203	SE	352	}
204		353
205	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	354	INLINE void
206	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	355	encode_comma (enc_t *enc)
207	#define COMMA SB \	356	{
208	encode_ch (enc, ','); \	357	encode_ch (enc, ',');
		358
209	if (enc->flags & F_INDENT) \	359	if (enc->json.flags & F_INDENT)
210	NL; \	360	encode_nl (enc);
211	else if (enc->flags & F_SPACE_AFTER) \	361	else if (enc->json.flags & F_SPACE_AFTER)
212	SPACE; \	362	encode_space (enc);
213	SE	363	}
214		364
215	static void encode_sv (enc_t *enc, SV *sv);	365	static void encode_sv (enc_t *enc, SV *sv);
216		366
217	static void	367	static void
218	encode_av (enc_t *enc, AV *av)	368	encode_av (enc_t *enc, AV *av)
219	{	369	{
220	int i, len = av_len (av);	370	int i, len = av_len (av);
221		371
		372	if (enc->indent >= enc->json.max_depth)
		373	croak (ERR_NESTING_EXCEEDED);
		374
222	encode_ch (enc, '['); NL;	375	encode_ch (enc, '[');
223	++enc->indent;	376
		377	if (len >= 0)
		378	{
		379	encode_nl (enc); ++enc->indent;
224		380
225	for (i = 0; i <= len; ++i)	381	for (i = 0; i <= len; ++i)
226	{	382	{
227	INDENT;	383	SV **svp = av_fetch (av, i, 0);
228	encode_sv (enc, *av_fetch (av, i, 0));
229		384
		385	encode_indent (enc);
		386
		387	if (svp)
		388	encode_sv (enc, *svp);
		389	else
		390	encode_str (enc, "null", 4, 0);
		391
230	if (i < len)	392	if (i < len)
231	COMMA;	393	encode_comma (enc);
232	}	394	}
233		395
234	NL;	396	encode_nl (enc); --enc->indent; encode_indent (enc);
235		397	}
236	--enc->indent;	398
237	INDENT; encode_ch (enc, ']');	399	encode_ch (enc, ']');
238	}	400	}
239		401
240	static void	402	static void
241	encode_he (enc_t *enc, HE *he)	403	encode_hk (enc_t *enc, HE *he)
242	{	404	{
243	encode_ch (enc, '"');	405	encode_ch (enc, '"');
244		406
245	if (HeKLEN (he) == HEf_SVKEY)	407	if (HeKLEN (he) == HEf_SVKEY)
246	{	408	{
…		…
256	else	418	else
257	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	419	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
258		420
259	encode_ch (enc, '"');	421	encode_ch (enc, '"');
260		422
261	if (enc->flags & F_SPACE_BEFORE) SPACE;	423	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
262	encode_ch (enc, ':');	424	encode_ch (enc, ':');
263	if (enc->flags & F_SPACE_AFTER ) SPACE;	425	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
264	encode_sv (enc, HeVAL (he));
265	}	426	}
266		427
267	// compare hash entries, used when all keys are bytestrings	428	// compare hash entries, used when all keys are bytestrings
268	static int	429	static int
269	he_cmp_fast (const void *a_, const void *b_)	430	he_cmp_fast (const void *a_, const void *b_)
…		…
274	HE *b = *(HE **)b_;	435	HE *b = *(HE **)b_;
275		436
276	STRLEN la = HeKLEN (a);	437	STRLEN la = HeKLEN (a);
277	STRLEN lb = HeKLEN (b);	438	STRLEN lb = HeKLEN (b);
278		439
279	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	440	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
280	cmp = la - lb;	441	cmp = lb - la;
281		442
282	return cmp;	443	return cmp;
283	}	444	}
284		445
285	// compare hash entries, used when some keys are sv's or utf-x	446	// compare hash entries, used when some keys are sv's or utf-x
286	static int	447	static int
287	he_cmp_slow (const void *a, const void *b)	448	he_cmp_slow (const void *a, const void *b)
288	{	449	{
289	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	450	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
290	}	451	}
291		452
292	static void	453	static void
293	encode_hv (enc_t *enc, HV *hv)	454	encode_hv (enc_t *enc, HV *hv)
294	{	455	{
295	int count, i;	456	HE *he;
296		457
297	encode_ch (enc, '{'); NL; ++enc->indent;	458	if (enc->indent >= enc->json.max_depth)
		459	croak (ERR_NESTING_EXCEEDED);
298		460
299	if ((count = hv_iterinit (hv)))	461	encode_ch (enc, '{');
300	{	462
301	// for canonical output we have to sort by keys first	463	// for canonical output we have to sort by keys first
302	// actually, this is mostly due to the stupid so-called	464	// actually, this is mostly due to the stupid so-called
303	// security workaround added somewhere in 5.8.x.	465	// security workaround added somewhere in 5.8.x
304	// that randomises hash orderings	466	// that randomises hash orderings
305	if (enc->flags & F_CANONICAL)	467	if (enc->json.flags & F_CANONICAL)
		468	{
		469	int count = hv_iterinit (hv);
		470
		471	if (SvMAGICAL (hv))
306	{	472	{
307	HE *he, *hes [count];	473	// need to count by iterating. could improve by dynamically building the vector below
		474	// but I don't care for the speed of this special case.
		475	// note also that we will run into undefined behaviour when the two iterations
		476	// do not result in the same count, something I might care for in some later release.
		477
		478	count = 0;
		479	while (hv_iternext (hv))
		480	++count;
		481
		482	hv_iterinit (hv);
		483	}
		484
		485	if (count)
		486	{
308	int fast = 1;	487	int i, fast = 1;
		488	#if defined(__BORLANDC__) || defined(_MSC_VER)
		489	HE **hes = _alloca (count * sizeof (HE));
		490	#else
		491	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		492	#endif
309		493
310	i = 0;	494	i = 0;
311	while ((he = hv_iternext (hv)))	495	while ((he = hv_iternext (hv)))
312	{	496	{
313	hes [i++] = he;	497	hes [i++] = he;
…		…
335		519
336	FREETMPS;	520	FREETMPS;
337	LEAVE;	521	LEAVE;
338	}	522	}
339		523
340	for (i = 0; i < count; ++i)	524	encode_nl (enc); ++enc->indent;
		525
		526	while (count--)
341	{	527	{
342	INDENT;	528	encode_indent (enc);
		529	he = hes [count];
343	encode_he (enc, hes [i]);	530	encode_hk (enc, he);
		531	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
344		532
345	if (i < count - 1)	533	if (count)
346	COMMA;	534	encode_comma (enc);
347	}	535	}
348		536
		537	encode_nl (enc); --enc->indent; encode_indent (enc);
		538	}
		539	}
		540	else
		541	{
		542	if (hv_iterinit (hv) || SvMAGICAL (hv))
		543	if ((he = hv_iternext (hv)))
349	NL;	544	{
		545	encode_nl (enc); ++enc->indent;
		546
		547	for (;;)
		548	{
		549	encode_indent (enc);
		550	encode_hk (enc, he);
		551	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		552
		553	if (!(he = hv_iternext (hv)))
		554	break;
		555
		556	encode_comma (enc);
		557	}
		558
		559	encode_nl (enc); --enc->indent; encode_indent (enc);
		560	}
		561	}
		562
		563	encode_ch (enc, '}');
		564	}
		565
		566	// encode objects, arrays and special \0=false and \1=true values.
		567	static void
		568	encode_rv (enc_t *enc, SV *sv)
		569	{
		570	svtype svt;
		571
		572	SvGETMAGIC (sv);
		573	svt = SvTYPE (sv);
		574
		575	if (expect_false (SvOBJECT (sv)))
		576	{
		577	HV *stash = !JSON_SLOW || json_boolean_stash
		578	? json_boolean_stash
		579	: gv_stashpv ("JSON::XS::Boolean", 1);
		580
		581	if (SvSTASH (sv) == stash)
		582	{
		583	if (SvIV (sv))
		584	encode_str (enc, "true", 4, 0);
		585	else
		586	encode_str (enc, "false", 5, 0);
350	}	587	}
351	else	588	else
352	{	589	{
353	SV *sv;	590	#if 0
354	HE *he = hv_iternext (hv);	591	if (0 && sv_derived_from (rv, "JSON::Literal"))
355
356	for (;;)
357	{	592	{
358	INDENT;	593	// not yet
359	encode_he (enc, he);
360
361	if (!(he = hv_iternext (hv)))
362	break;
363
364	COMMA;
365	}	594	}
366		595	#endif
		596	if (enc->json.flags & F_CONV_BLESSED)
367	NL;	597	{
		598	// we re-bless the reference to get overload and other niceties right
		599	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
		600
		601	if (to_json)
		602	{
		603	dSP;
		604
		605	ENTER; SAVETMPS; PUSHMARK (SP);
		606	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		607
		608	// calling with G_SCALAR ensures that we always get a 1 return value
		609	PUTBACK;
		610	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		611	SPAGAIN;
		612
		613	// catch this surprisingly common error
		614	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		615	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		616
		617	sv = POPs;
		618	PUTBACK;
		619
		620	encode_sv (enc, sv);
		621
		622	FREETMPS; LEAVE;
		623	}
		624	else if (enc->json.flags & F_ALLOW_BLESSED)
		625	encode_str (enc, "null", 4, 0);
		626	else
		627	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		628	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		629	}
		630	else if (enc->json.flags & F_ALLOW_BLESSED)
		631	encode_str (enc, "null", 4, 0);
		632	else
		633	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		634	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
368	}	635	}
369	}	636	}
		637	else if (svt == SVt_PVHV)
		638	encode_hv (enc, (HV *)sv);
		639	else if (svt == SVt_PVAV)
		640	encode_av (enc, (AV *)sv);
		641	else if (svt < SVt_PVAV)
		642	{
		643	STRLEN len = 0;
		644	char *pv = svt ? SvPV (sv, len) : 0;
370		645
371	--enc->indent; INDENT; encode_ch (enc, '}');	646	if (len == 1 && *pv == '1')
		647	encode_str (enc, "true", 4, 0);
		648	else if (len == 1 && *pv == '0')
		649	encode_str (enc, "false", 5, 0);
		650	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		651	encode_str (enc, "null", 4, 0);
		652	else
		653	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		654	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		655	}
		656	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		657	encode_str (enc, "null", 4, 0);
		658	else
		659	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		660	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
372	}	661	}
373		662
374	static void	663	static void
375	encode_sv (enc_t *enc, SV *sv)	664	encode_sv (enc_t *enc, SV *sv)
376	{	665	{
…		…
384	encode_str (enc, str, len, SvUTF8 (sv));	673	encode_str (enc, str, len, SvUTF8 (sv));
385	encode_ch (enc, '"');	674	encode_ch (enc, '"');
386	}	675	}
387	else if (SvNOKp (sv))	676	else if (SvNOKp (sv))
388	{	677	{
		678	// trust that perl will do the right thing w.r.t. JSON syntax.
389	need (enc, NV_DIG + 32);	679	need (enc, NV_DIG + 32);
390	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	680	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
391	enc->cur += strlen (enc->cur);	681	enc->cur += strlen (enc->cur);
392	}	682	}
393	else if (SvIOKp (sv))	683	else if (SvIOKp (sv))
394	{	684	{
		685	// we assume we can always read an IV as a UV and vice versa
		686	// we assume two's complement
		687	// we assume no aliasing issues in the union
		688	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		689	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		690	{
		691	// optimise the "small number case"
		692	// code will likely be branchless and use only a single multiplication
		693	// works for numbers up to 59074
		694	I32 i = SvIVX (sv);
		695	U32 u;
		696	char digit, nz = 0;
		697
395	need (enc, 64);	698	need (enc, 6);
		699
		700	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		701	u = i < 0 ? -i : i;
		702
		703	// convert to 4.28 fixed-point representation
		704	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		705
		706	// now output digit by digit, each time masking out the integer part
		707	// and multiplying by 5 while moving the decimal point one to the right,
		708	// resulting in a net multiplication by 10.
		709	// we always write the digit to memory but conditionally increment
		710	// the pointer, to enable the use of conditional move instructions.
		711	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		712	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		713	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		714	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		715	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		716	}
		717	else
		718	{
		719	// large integer, use the (rather slow) snprintf way.
		720	need (enc, IVUV_MAXCHARS);
396	enc->cur +=	721	enc->cur +=
397	SvIsUV(sv)	722	SvIsUV(sv)
398	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	723	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
399	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	724	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		725	}
400	}	726	}
401	else if (SvROK (sv))	727	else if (SvROK (sv))
402	{	728	encode_rv (enc, SvRV (sv));
403	SV *rv = SvRV (sv);	729	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
404
405	if (!--enc->max_recurse)
406	croak ("data structure too deep (hit recursion limit)");
407
408	switch (SvTYPE (rv))
409	{
410	case SVt_PVAV: encode_av (enc, (AV *)rv); break;
411	case SVt_PVHV: encode_hv (enc, (HV *)rv); break;
412
413	default:
414	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
415	SvPV_nolen (sv));
416	}
417	}
418	else if (!SvOK (sv))
419	encode_str (enc, "null", 4, 0);	730	encode_str (enc, "null", 4, 0);
420	else	731	else
421	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	732	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
422	SvPV_nolen (sv), SvFLAGS (sv));	733	SvPV_nolen (sv), SvFLAGS (sv));
423	}	734	}
424		735
425	static SV *	736	static SV *
426	encode_json (SV *scalar, UV flags)	737	encode_json (SV *scalar, JSON *json)
427	{	738	{
		739	enc_t enc;
		740
428	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	741	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
429	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	742	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
430		743
431	enc_t enc;	744	enc.json = *json;
432	enc.flags = flags;
433	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	745	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
434	enc.cur = SvPVX (enc.sv);	746	enc.cur = SvPVX (enc.sv);
435	enc.end = SvEND (enc.sv);	747	enc.end = SvEND (enc.sv);
436	enc.max_recurse = 0;
437	enc.indent = 0;	748	enc.indent = 0;
		749	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		750	: enc.json.flags & F_LATIN1 ? 0x000100UL
		751	: 0x110000UL;
438		752
439	SvPOK_only (enc.sv);	753	SvPOK_only (enc.sv);
440	encode_sv (&enc, scalar);	754	encode_sv (&enc, scalar);
441		755
		756	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		757	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		758
442	if (!(flags & (F_ASCII | F_UTF8)))	759	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
443	SvUTF8_on (enc.sv);	760	SvUTF8_on (enc.sv);
444		761
445	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
446
447	if (enc.flags & F_SHRINK)	762	if (enc.json.flags & F_SHRINK)
448	shrink (enc.sv);	763	shrink (enc.sv);
449		764
450	return enc.sv;	765	return enc.sv;
451	}	766	}
452		767
453	/////////////////////////////////////////////////////////////////////////////	768	/////////////////////////////////////////////////////////////////////////////
		769	// decoder
454		770
455	#define WS \	771	// structure used for decoding JSON
		772	typedef struct
		773	{
		774	char *cur; // current parser pointer
		775	char *end; // end of input string
		776	const char *err; // parse error, if != 0
		777	JSON json;
		778	U32 depth; // recursion depth
		779	U32 maxdepth; // recursion depth limit
		780	} dec_t;
		781
		782	INLINE void
		783	decode_comment (dec_t *dec)
		784	{
		785	// only '#'-style comments allowed a.t.m.
		786
		787	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		788	++dec->cur;
		789	}
		790
		791	INLINE void
		792	decode_ws (dec_t *dec)
		793	{
456	for (;;) \	794	for (;;)
457	{ \	795	{
458	char ch = *dec->cur; \	796	char ch = *dec->cur;
		797
459	if (ch > 0x20 \	798	if (ch > 0x20)
		799	{
		800	if (expect_false (ch == '#'))
		801	{
		802	if (dec->json.flags & F_RELAXED)
		803	decode_comment (dec);
		804	else
		805	break;
		806	}
		807	else
		808	break;
		809	}
460	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	810	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
461	break; \	811	break; // parse error, but let higher level handle it, gives better error messages
		812
462	++dec->cur; \	813	++dec->cur;
463	}	814	}
		815	}
464		816
465	#define ERR(reason) SB dec->err = reason; goto fail; SE	817	#define ERR(reason) SB dec->err = reason; goto fail; SE
		818
466	#define EXPECT_CH(ch) SB \	819	#define EXPECT_CH(ch) SB \
467	if (*dec->cur != ch) \	820	if (*dec->cur != ch) \
468	ERR (# ch " expected"); \	821	ERR (# ch " expected"); \
469	++dec->cur; \	822	++dec->cur; \
470	SE	823	SE
471		824
		825	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
		826	#define DEC_DEC_DEPTH --dec->depth
		827
472	static SV *decode_sv (dec_t *dec);	828	static SV *decode_sv (dec_t *dec);
473		829
474	static signed char decode_hexdigit[256];	830	static signed char decode_hexdigit[256];
475		831
476	static UV	832	static UV
477	decode_4hex (dec_t *dec)	833	decode_4hex (dec_t *dec)
478	{	834	{
479	signed char d1, d2, d3, d4;	835	signed char d1, d2, d3, d4;
		836	unsigned char *cur = (unsigned char *)dec->cur;
480		837
481	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	838	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
482	if (d1 < 0) ERR ("four hexadecimal digits expected");	839	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
483	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	840	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
484	if (d2 < 0) ERR ("four hexadecimal digits expected");	841	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
485	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
486	if (d3 < 0) ERR ("four hexadecimal digits expected");
487	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
488	if (d4 < 0) ERR ("four hexadecimal digits expected");
489		842
490	dec->cur += 4;	843	dec->cur += 4;
491		844
492	return ((UV)d1) << 12	845	return ((UV)d1) << 12
493	| ((UV)d2) << 8	846	| ((UV)d2) << 8
…		…
496		849
497	fail:	850	fail:
498	return (UV)-1;	851	return (UV)-1;
499	}	852	}
500		853
501	#define APPEND_GROW(n) SB \
502	if (cur + (n) >= end) \
503	{ \
504	STRLEN ofs = cur - SvPVX (sv); \
505	SvGROW (sv, ofs + (n) + 1); \
506	cur = SvPVX (sv) + ofs; \
507	end = SvEND (sv); \
508	} \
509	SE
510
511	#define APPEND_CH(ch) SB \
512	APPEND_GROW (1); \
513	*cur++ = (ch); \
514	SE
515
516	static SV *	854	static SV *
517	decode_str (dec_t *dec)	855	decode_str (dec_t *dec)
518	{	856	{
519	SV *sv = NEWSV (0,2);	857	SV *sv = 0;
520	int utf8 = 0;	858	int utf8 = 0;
521	char *cur = SvPVX (sv);	859	char *dec_cur = dec->cur;
522	char *end = SvEND (sv);
523		860
524	for (;;)	861	do
525	{	862	{
526	unsigned char ch = *(unsigned char *)dec->cur;	863	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		864	char *cur = buf;
527		865
528	if (ch == '"')	866	do
529	break;
530	else if (ch == '\\')
531	{	867	{
532	switch (*++dec->cur)	868	unsigned char ch = *(unsigned char *)dec_cur++;
		869
		870	if (expect_false (ch == '"'))
533	{	871	{
534	case '\\':	872	--dec_cur;
535	case '/':	873	break;
536	case '"': APPEND_CH (*dec->cur++); break;	874	}
537		875	else if (expect_false (ch == '\\'))
538	case 'b': APPEND_CH ('\010'); ++dec->cur; break;	876	{
539	case 't': APPEND_CH ('\011'); ++dec->cur; break;	877	switch (*dec_cur)
540	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
541	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
542	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
543
544	case 'u':
545	{	878	{
546	UV lo, hi;	879	case '\\':
547	++dec->cur;	880	case '/':
		881	case '"': *cur++ = *dec_cur++; break;
548		882
549	hi = decode_4hex (dec);	883	case 'b': ++dec_cur; *cur++ = '\010'; break;
550	if (hi == (UV)-1)	884	case 't': ++dec_cur; *cur++ = '\011'; break;
551	goto fail;	885	case 'n': ++dec_cur; *cur++ = '\012'; break;
		886	case 'f': ++dec_cur; *cur++ = '\014'; break;
		887	case 'r': ++dec_cur; *cur++ = '\015'; break;
552		888
553	// possibly a surrogate pair	889	case 'u':
554	if (hi >= 0xd800 && hi < 0xdc00)
555	{	890	{
556	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	891	UV lo, hi;
557	ERR ("missing low surrogate character in surrogate pair");	892	++dec_cur;
558		893
559	dec->cur += 2;	894	dec->cur = dec_cur;
560
561	lo = decode_4hex (dec);	895	hi = decode_4hex (dec);
		896	dec_cur = dec->cur;
562	if (lo == (UV)-1)	897	if (hi == (UV)-1)
563	goto fail;	898	goto fail;
564		899
		900	// possibly a surrogate pair
		901	if (hi >= 0xd800)
		902	if (hi < 0xdc00)
		903	{
		904	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		905	ERR ("missing low surrogate character in surrogate pair");
		906
		907	dec_cur += 2;
		908
		909	dec->cur = dec_cur;
		910	lo = decode_4hex (dec);
		911	dec_cur = dec->cur;
		912	if (lo == (UV)-1)
		913	goto fail;
		914
565	if (lo < 0xdc00 || lo >= 0xe000)	915	if (lo < 0xdc00 || lo >= 0xe000)
566	ERR ("surrogate pair expected");	916	ERR ("surrogate pair expected");
567		917
568	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	918	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		919	}
		920	else if (hi < 0xe000)
		921	ERR ("missing high surrogate character in surrogate pair");
		922
		923	if (hi >= 0x80)
		924	{
		925	utf8 = 1;
		926
		927	cur = encode_utf8 (cur, hi);
		928	}
		929	else
		930	*cur++ = hi;
569	}	931	}
570	else if (hi >= 0xdc00 && hi < 0xe000)
571	ERR ("missing high surrogate character in surrogate pair");
572
573	if (hi >= 0x80)
574	{	932	break;
575	utf8 = 1;
576		933
577	APPEND_GROW (4); // at most 4 bytes for 21 bits
578	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
579	}
580	else	934	default:
581	APPEND_CH (hi);	935	--dec_cur;
		936	ERR ("illegal backslash escape sequence in string");
582	}	937	}
583	break;	938	}
		939	else if (expect_true (ch >= 0x20 && ch < 0x80))
		940	*cur++ = ch;
		941	else if (ch >= 0x80)
		942	{
		943	STRLEN clen;
		944	UV uch;
584		945
585	default:
586	--dec->cur;	946	--dec_cur;
587	ERR ("illegal backslash escape sequence in string");	947
		948	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		949	if (clen == (STRLEN)-1)
		950	ERR ("malformed UTF-8 character in JSON string");
		951
		952	do
		953	*cur++ = *dec_cur++;
		954	while (--clen);
		955
		956	utf8 = 1;
		957	}
		958	else
		959	{
		960	--dec_cur;
		961
		962	if (!ch)
		963	ERR ("unexpected end of string while parsing JSON string");
		964	else
		965	ERR ("invalid character encountered while parsing JSON string");
588	}	966	}
589	}	967	}
590	else if (ch >= 0x20 && ch <= 0x7f)	968	while (cur < buf + SHORT_STRING_LEN);
591	APPEND_CH (*dec->cur++);	969
592	else if (ch >= 0x80)
593	{	970	{
594	STRLEN clen;	971	STRLEN len = cur - buf;
595	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
596	if (clen == (STRLEN)-1)
597	ERR ("malformed UTF-8 character in JSON string");
598		972
599	APPEND_GROW (clen);	973	if (sv)
600	do
601	{	974	{
602	*cur++ = *dec->cur++;	975	STRLEN cur = SvCUR (sv);
		976
		977	if (SvLEN (sv) <= cur + len)
		978	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		979
		980	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		981	SvCUR_set (sv, SvCUR (sv) + len);
603	}	982	}
604	while (--clen);
605
606	utf8 = 1;
607	}
608	else if (dec->cur == dec->end)
609	ERR ("unexpected end of string while parsing json string");
610	else	983	else
611	ERR ("invalid character encountered");	984	sv = newSVpvn (buf, len);
612	}	985	}
		986	}
		987	while (*dec_cur != '"');
613		988
614	++dec->cur;	989	++dec_cur;
615		990
616	SvCUR_set (sv, cur - SvPVX (sv));	991	if (sv)
617		992	{
618	SvPOK_only (sv);	993	SvPOK_only (sv);
619	*SvEND (sv) = 0;	994	*SvEND (sv) = 0;
620		995
621	if (utf8)	996	if (utf8)
622	SvUTF8_on (sv);	997	SvUTF8_on (sv);
		998	}
		999	else
		1000	sv = newSVpvn ("", 0);
623		1001
624	if (dec->flags & F_SHRINK)	1002	dec->cur = dec_cur;
625	shrink (sv);
626
627	return sv;	1003	return sv;
628		1004
629	fail:	1005	fail:
630	SvREFCNT_dec (sv);	1006	dec->cur = dec_cur;
631	return 0;	1007	return 0;
632	}	1008	}
633		1009
634	static SV *	1010	static SV *
635	decode_num (dec_t *dec)	1011	decode_num (dec_t *dec)
…		…
693	is_nv = 1;	1069	is_nv = 1;
694	}	1070	}
695		1071
696	if (!is_nv)	1072	if (!is_nv)
697	{	1073	{
698	UV uv;	1074	int len = dec->cur - start;
699	int numtype = grok_number (start, dec->cur - start, &uv);	1075
700	if (numtype & IS_NUMBER_IN_UV)	1076	// special case the rather common 1..5-digit-int case
701	if (numtype & IS_NUMBER_NEG)	1077	if (*start == '-')
		1078	switch (len)
702	{	1079	{
703	if (uv < (UV)IV_MIN)	1080	case 2: return newSViv (-( start [1] - '0' * 1));
704	return newSViv (-(IV)uv);	1081	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		1082	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1083	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1084	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
705	}	1085	}
		1086	else
		1087	switch (len)
		1088	{
		1089	case 1: return newSViv ( start [0] - '0' * 1);
		1090	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1091	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1092	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1093	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1094	}
		1095
		1096	{
		1097	UV uv;
		1098	int numtype = grok_number (start, len, &uv);
		1099	if (numtype & IS_NUMBER_IN_UV)
		1100	if (numtype & IS_NUMBER_NEG)
		1101	{
		1102	if (uv < (UV)IV_MIN)
		1103	return newSViv (-(IV)uv);
		1104	}
706	else	1105	else
707	return newSVuv (uv);	1106	return newSVuv (uv);
708	}	1107	}
709		1108
		1109	len -= *start == '-' ? 1 : 0;
		1110
		1111	// does not fit into IV or UV, try NV
		1112	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1113	#if defined (LDBL_DIG)
		1114	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1115	#endif
		1116	)
		1117	// fits into NV without loss of precision
		1118	return newSVnv (Atof (start));
		1119
		1120	// everything else fails, convert it to a string
		1121	return newSVpvn (start, dec->cur - start);
		1122	}
		1123
		1124	// loss of precision here
710	return newSVnv (Atof (start));	1125	return newSVnv (Atof (start));
711		1126
712	fail:	1127	fail:
713	return 0;	1128	return 0;
714	}	1129	}
…		…
716	static SV *	1131	static SV *
717	decode_av (dec_t *dec)	1132	decode_av (dec_t *dec)
718	{	1133	{
719	AV *av = newAV ();	1134	AV *av = newAV ();
720		1135
721	WS;	1136	DEC_INC_DEPTH;
		1137	decode_ws (dec);
		1138
722	if (*dec->cur == ']')	1139	if (*dec->cur == ']')
723	++dec->cur;	1140	++dec->cur;
724	else	1141	else
725	for (;;)	1142	for (;;)
726	{	1143	{
…		…
730	if (!value)	1147	if (!value)
731	goto fail;	1148	goto fail;
732		1149
733	av_push (av, value);	1150	av_push (av, value);
734		1151
735	WS;	1152	decode_ws (dec);
736		1153
737	if (*dec->cur == ']')	1154	if (*dec->cur == ']')
738	{	1155	{
739	++dec->cur;	1156	++dec->cur;
740	break;	1157	break;
…		…
742		1159
743	if (*dec->cur != ',')	1160	if (*dec->cur != ',')
744	ERR (", or ] expected while parsing array");	1161	ERR (", or ] expected while parsing array");
745		1162
746	++dec->cur;	1163	++dec->cur;
		1164
		1165	decode_ws (dec);
		1166
		1167	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1168	{
		1169	++dec->cur;
		1170	break;
		1171	}
747	}	1172	}
748		1173
		1174	DEC_DEC_DEPTH;
749	return newRV_noinc ((SV *)av);	1175	return newRV_noinc ((SV *)av);
750		1176
751	fail:	1177	fail:
752	SvREFCNT_dec (av);	1178	SvREFCNT_dec (av);
		1179	DEC_DEC_DEPTH;
753	return 0;	1180	return 0;
754	}	1181	}
755		1182
756	static SV *	1183	static SV *
757	decode_hv (dec_t *dec)	1184	decode_hv (dec_t *dec)
758	{	1185	{
		1186	SV *sv;
759	HV *hv = newHV ();	1187	HV *hv = newHV ();
760		1188
761	WS;	1189	DEC_INC_DEPTH;
		1190	decode_ws (dec);
		1191
762	if (*dec->cur == '}')	1192	if (*dec->cur == '}')
763	++dec->cur;	1193	++dec->cur;
764	else	1194	else
765	for (;;)	1195	for (;;)
766	{	1196	{
767	SV *key, *value;
768
769	WS; EXPECT_CH ('"');	1197	EXPECT_CH ('"');
770		1198
771	key = decode_str (dec);	1199	// heuristic: assume that
772	if (!key)	1200	// a) decode_str + hv_store_ent are abysmally slow.
773	goto fail;	1201	// b) most hash keys are short, simple ascii text.
		1202	// => try to "fast-match" such strings to avoid
		1203	// the overhead of decode_str + hv_store_ent.
		1204	{
		1205	SV *value;
		1206	char *p = dec->cur;
		1207	char *e = p + 24; // only try up to 24 bytes
774		1208
775	WS; EXPECT_CH (':');	1209	for (;;)
776
777	value = decode_sv (dec);
778	if (!value)
779	{	1210	{
		1211	// the >= 0x80 is false on most architectures
		1212	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1213	{
		1214	// slow path, back up and use decode_str
		1215	SV *key = decode_str (dec);
		1216	if (!key)
		1217	goto fail;
		1218
		1219	decode_ws (dec); EXPECT_CH (':');
		1220
		1221	decode_ws (dec);
		1222	value = decode_sv (dec);
		1223	if (!value)
		1224	{
		1225	SvREFCNT_dec (key);
		1226	goto fail;
		1227	}
		1228
		1229	hv_store_ent (hv, key, value, 0);
780	SvREFCNT_dec (key);	1230	SvREFCNT_dec (key);
		1231
		1232	break;
		1233	}
		1234	else if (*p == '"')
		1235	{
		1236	// fast path, got a simple key
		1237	char *key = dec->cur;
		1238	int len = p - key;
		1239	dec->cur = p + 1;
		1240
		1241	decode_ws (dec); EXPECT_CH (':');
		1242
		1243	decode_ws (dec);
		1244	value = decode_sv (dec);
		1245	if (!value)
781	goto fail;	1246	goto fail;
		1247
		1248	hv_store (hv, key, len, value, 0);
		1249
		1250	break;
		1251	}
		1252
		1253	++p;
782	}	1254	}
783
784	//TODO: optimise
785	hv_store_ent (hv, key, value, 0);
786
787	WS;	1255	}
		1256
		1257	decode_ws (dec);
788		1258
789	if (*dec->cur == '}')	1259	if (*dec->cur == '}')
790	{	1260	{
791	++dec->cur;	1261	++dec->cur;
792	break;	1262	break;
…		…
794		1264
795	if (*dec->cur != ',')	1265	if (*dec->cur != ',')
796	ERR (", or } expected while parsing object/hash");	1266	ERR (", or } expected while parsing object/hash");
797		1267
798	++dec->cur;	1268	++dec->cur;
		1269
		1270	decode_ws (dec);
		1271
		1272	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1273	{
		1274	++dec->cur;
		1275	break;
		1276	}
799	}	1277	}
800		1278
		1279	DEC_DEC_DEPTH;
801	return newRV_noinc ((SV *)hv);	1280	sv = newRV_noinc ((SV *)hv);
		1281
		1282	// check filter callbacks
		1283	if (dec->json.flags & F_HOOK)
		1284	{
		1285	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1286	{
		1287	HE *cb, *he;
		1288
		1289	hv_iterinit (hv);
		1290	he = hv_iternext (hv);
		1291	hv_iterinit (hv);
		1292
		1293	// the next line creates a mortal sv each time its called.
		1294	// might want to optimise this for common cases.
		1295	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1296
		1297	if (cb)
		1298	{
		1299	dSP;
		1300	int count;
		1301
		1302	ENTER; SAVETMPS; PUSHMARK (SP);
		1303	XPUSHs (HeVAL (he));
		1304
		1305	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1306
		1307	if (count == 1)
		1308	{
		1309	sv = newSVsv (POPs);
		1310	FREETMPS; LEAVE;
		1311	return sv;
		1312	}
		1313
		1314	FREETMPS; LEAVE;
		1315	}
		1316	}
		1317
		1318	if (dec->json.cb_object)
		1319	{
		1320	dSP;
		1321	int count;
		1322
		1323	ENTER; SAVETMPS; PUSHMARK (SP);
		1324	XPUSHs (sv_2mortal (sv));
		1325
		1326	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1327
		1328	if (count == 1)
		1329	{
		1330	sv = newSVsv (POPs);
		1331	FREETMPS; LEAVE;
		1332	return sv;
		1333	}
		1334
		1335	SvREFCNT_inc (sv);
		1336	FREETMPS; LEAVE;
		1337	}
		1338	}
		1339
		1340	return sv;
802		1341
803	fail:	1342	fail:
804	SvREFCNT_dec (hv);	1343	SvREFCNT_dec (hv);
		1344	DEC_DEC_DEPTH;
805	return 0;	1345	return 0;
806	}	1346	}
807		1347
808	static SV *	1348	static SV *
809	decode_sv (dec_t *dec)	1349	decode_sv (dec_t *dec)
810	{	1350	{
811	WS;	1351	// the beauty of JSON: you need exactly one character lookahead
		1352	// to parse everything.
812	switch (*dec->cur)	1353	switch (*dec->cur)
813	{	1354	{
814	case '"': ++dec->cur; return decode_str (dec);	1355	case '"': ++dec->cur; return decode_str (dec);
815	case '[': ++dec->cur; return decode_av (dec);	1356	case '[': ++dec->cur; return decode_av (dec);
816	case '{': ++dec->cur; return decode_hv (dec);	1357	case '{': ++dec->cur; return decode_hv (dec);
817		1358
818	case '-':	1359	case '-':
819	case '0': case '1': case '2': case '3': case '4':	1360	case '0': case '1': case '2': case '3': case '4':
820	case '5': case '6': case '7': case '8': case '9':	1361	case '5': case '6': case '7': case '8': case '9':
821	return decode_num (dec);	1362	return decode_num (dec);
822		1363
823	case 't':	1364	case 't':
824	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1365	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
825	{	1366	{
826	dec->cur += 4;	1367	dec->cur += 4;
		1368	#if JSON_SLOW
		1369	json_true = get_bool ("JSON::XS::true");
		1370	#endif
827	return newSViv (1);	1371	return newSVsv (json_true);
828	}	1372	}
829	else	1373	else
830	ERR ("'true' expected");	1374	ERR ("'true' expected");
831		1375
832	break;	1376	break;
833		1377
834	case 'f':	1378	case 'f':
835	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1379	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
836	{	1380	{
837	dec->cur += 5;	1381	dec->cur += 5;
		1382	#if JSON_SLOW
		1383	json_false = get_bool ("JSON::XS::false");
		1384	#endif
838	return newSViv (0);	1385	return newSVsv (json_false);
839	}	1386	}
840	else	1387	else
841	ERR ("'false' expected");	1388	ERR ("'false' expected");
842		1389
843	break;	1390	break;
…		…
852	ERR ("'null' expected");	1399	ERR ("'null' expected");
853		1400
854	break;	1401	break;
855		1402
856	default:	1403	default:
857	ERR ("malformed json string, neither array, object, number, string or atom");	1404	ERR ("malformed JSON string, neither array, object, number, string or atom");
858	break;	1405	break;
859	}	1406	}
860		1407
861	fail:	1408	fail:
862	return 0;	1409	return 0;
863	}	1410	}
864		1411
865	static SV *	1412	static SV *
866	decode_json (SV *string, UV flags)	1413	decode_json (SV *string, JSON *json, STRLEN *offset_return)
867	{	1414	{
		1415	dec_t dec;
		1416	STRLEN offset;
868	SV *sv;	1417	SV *sv;
869		1418
		1419	/* work around bugs in 5.10 where manipulating magic values
		1420	* will perl ignore the magic in subsequent accesses
		1421	*/
		1422	/*SvGETMAGIC (string);*/
		1423	if (SvMAGICAL (string))
		1424	string = sv_2mortal (newSVsv (string));
		1425
		1426	SvUPGRADE (string, SVt_PV);
		1427
		1428	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1429	* assertion with -DDEBUGGING, although SvCUR is documented to
		1430	* return the xpv_cur field which certainly exists after upgrading.
		1431	* according to nicholas clark, calling SvPOK fixes this.
		1432	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1433	* and hope for the best.
		1434	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1435	* assertion business is seriously broken, try yet another workaround
		1436	* for the broken -DDEBUGGING.
		1437	*/
		1438	#ifdef DEBUGGING
		1439	offset = SvOK (string) ? sv_len (string) : 0;
		1440	#else
		1441	offset = SvCUR (string);
		1442	#endif
		1443
		1444	if (offset > json->max_size && json->max_size)
		1445	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1446	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1447
870	if (flags & F_UTF8)	1448	if (json->flags & F_UTF8)
871	sv_utf8_downgrade (string, 0);	1449	sv_utf8_downgrade (string, 0);
872	else	1450	else
873	sv_utf8_upgrade (string);	1451	sv_utf8_upgrade (string);
874		1452
875	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1453	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
876		1454
877	dec_t dec;	1455	dec.json = *json;
878	dec.flags = flags;
879	dec.cur = SvPVX (string);	1456	dec.cur = SvPVX (string);
880	dec.end = SvEND (string);	1457	dec.end = SvEND (string);
881	dec.err = 0;	1458	dec.err = 0;
		1459	dec.depth = 0;
882		1460
		1461	if (dec.json.cb_object || dec.json.cb_sk_object)
		1462	dec.json.flags |= F_HOOK;
		1463
		1464	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1465
		1466	decode_ws (&dec);
883	sv = decode_sv (&dec);	1467	sv = decode_sv (&dec);
884		1468
		1469	if (!(offset_return || !sv))
		1470	{
		1471	// check for trailing garbage
		1472	decode_ws (&dec);
		1473
		1474	if (*dec.cur)
		1475	{
		1476	dec.err = "garbage after JSON object";
		1477	SvREFCNT_dec (sv);
		1478	sv = 0;
		1479	}
		1480	}
		1481
		1482	if (offset_return || !sv)
		1483	{
		1484	offset = dec.json.flags & F_UTF8
		1485	? dec.cur - SvPVX (string)
		1486	: utf8_distance (dec.cur, SvPVX (string));
		1487
		1488	if (offset_return)
		1489	*offset_return = offset;
		1490	}
		1491
885	if (!sv)	1492	if (!sv)
886	{	1493	{
887	IV offset = dec.flags & F_UTF8
888	? dec.cur - SvPVX (string)
889	: utf8_distance (dec.cur, SvPVX (string));
890	SV *uni = sv_newmortal ();	1494	SV *uni = sv_newmortal ();
891		1495
892	// horrible hack to silence warning inside pv_uni_display	1496	// horrible hack to silence warning inside pv_uni_display
893	COP cop = *PL_curcop;	1497	COP cop = *PL_curcop;
894	cop.cop_warnings = pWARN_NONE;	1498	cop.cop_warnings = pWARN_NONE;
…		…
896	SAVEVPTR (PL_curcop);	1500	SAVEVPTR (PL_curcop);
897	PL_curcop = &cop;	1501	PL_curcop = &cop;
898	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1502	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
899	LEAVE;	1503	LEAVE;
900		1504
901	croak ("%s, at character offset %d (%s)",	1505	croak ("%s, at character offset %d [\"%s\"]",
902	dec.err,	1506	dec.err,
903	(int)offset,	1507	(int)offset,
904	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1508	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
905	}	1509	}
906		1510
907	sv = sv_2mortal (sv);	1511	sv = sv_2mortal (sv);
908		1512
909	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1513	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
910	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1514	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
911		1515
912	return sv;	1516	return sv;
913	}	1517	}
914		1518
		1519	/////////////////////////////////////////////////////////////////////////////
		1520	// incremental parser
		1521
		1522	static void
		1523	incr_parse (JSON *self)
		1524	{
		1525	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1526
		1527	for (;;)
		1528	{
		1529	//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
		1530	switch (self->incr_mode)
		1531	{
		1532	// only used for intiial whitespace skipping
		1533	case INCR_M_WS:
		1534	for (;;)
		1535	{
		1536	if (*p > 0x20)
		1537	{
		1538	self->incr_mode = INCR_M_JSON;
		1539	goto incr_m_json;
		1540	}
		1541	else if (!*p)
		1542	goto interrupt;
		1543
		1544	++p;
		1545	}
		1546
		1547	// skip a single char inside a string (for \\-processing)
		1548	case INCR_M_BS:
		1549	if (!*p)
		1550	goto interrupt;
		1551
		1552	++p;
		1553	self->incr_mode = INCR_M_STR;
		1554	goto incr_m_str;
		1555
		1556	// inside a string
		1557	case INCR_M_STR:
		1558	incr_m_str:
		1559	for (;;)
		1560	{
		1561	if (*p == '"')
		1562	{
		1563	++p;
		1564	self->incr_mode = INCR_M_JSON;
		1565
		1566	if (!self->incr_nest)
		1567	goto interrupt;
		1568
		1569	goto incr_m_json;
		1570	}
		1571	else if (*p == '\\')
		1572	{
		1573	++p; // "virtually" consumes character after \
		1574
		1575	if (!*p) // if at end of string we have to switch modes
		1576	{
		1577	self->incr_mode = INCR_M_BS;
		1578	goto interrupt;
		1579	}
		1580	}
		1581	else if (!*p)
		1582	goto interrupt;
		1583
		1584	++p;
		1585	}
		1586
		1587	// after initial ws, outside string
		1588	case INCR_M_JSON:
		1589	incr_m_json:
		1590	for (;;)
		1591	{
		1592	switch (*p++)
		1593	{
		1594	case 0:
		1595	--p;
		1596	goto interrupt;
		1597
		1598	case 0x09:
		1599	case 0x0a:
		1600	case 0x0d:
		1601	case 0x20:
		1602	if (!self->incr_nest)
		1603	{
		1604	--p; // do not eat the whitespace, let the next round do it
		1605	goto interrupt;
		1606	}
		1607	break;
		1608
		1609	case '"':
		1610	self->incr_mode = INCR_M_STR;
		1611	goto incr_m_str;
		1612
		1613	case '[':
		1614	case '{':
		1615	if (++self->incr_nest > self->max_depth)
		1616	croak (ERR_NESTING_EXCEEDED);
		1617	break;
		1618
		1619	case ']':
		1620	case '}':
		1621	if (--self->incr_nest <= 0)
		1622	goto interrupt;
		1623	}
		1624	}
		1625	}
		1626
		1627	modechange:
		1628	;
		1629	}
		1630
		1631	interrupt:
		1632	self->incr_pos = p - SvPVX (self->incr_text);
		1633	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
		1634	}
		1635
		1636	/////////////////////////////////////////////////////////////////////////////
		1637	// XS interface functions
		1638
915	MODULE = JSON::XS PACKAGE = JSON::XS	1639	MODULE = JSON::XS PACKAGE = JSON::XS
916		1640
917	BOOT:	1641	BOOT:
918	{	1642	{
919	int i;	1643	int i;
920		1644
921	memset (decode_hexdigit, 0xff, 256);
922	for (i = 10; i--; )	1645	for (i = 0; i < 256; ++i)
923	decode_hexdigit ['0' + i] = i;	1646	decode_hexdigit [i] =
		1647	i >= '0' && i <= '9' ? i - '0'
		1648	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1649	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1650	: -1;
924		1651
925	for (i = 7; i--; )
926	{
927	decode_hexdigit ['a' + i] = 10 + i;
928	decode_hexdigit ['A' + i] = 10 + i;
929	}
930
931	json_stash = gv_stashpv ("JSON::XS", 1);	1652	json_stash = gv_stashpv ("JSON::XS" , 1);
		1653	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1654
		1655	json_true = get_bool ("JSON::XS::true");
		1656	json_false = get_bool ("JSON::XS::false");
932	}	1657	}
933		1658
934	PROTOTYPES: DISABLE	1659	PROTOTYPES: DISABLE
935		1660
936	SV *new (char *dummy)	1661	void CLONE (...)
937	CODE:	1662	CODE:
938	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1663	json_stash = 0;
		1664	json_boolean_stash = 0;
		1665
		1666	void new (char *klass)
		1667	PPCODE:
		1668	{
		1669	SV *pv = NEWSV (0, sizeof (JSON));
		1670	SvPOK_only (pv);
		1671	json_init ((JSON *)SvPVX (pv));
		1672	XPUSHs (sv_2mortal (sv_bless (
		1673	newRV_noinc (pv),
		1674	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1675	)));
		1676	}
		1677
		1678	void ascii (JSON *self, int enable = 1)
		1679	ALIAS:
		1680	ascii = F_ASCII
		1681	latin1 = F_LATIN1
		1682	utf8 = F_UTF8
		1683	indent = F_INDENT
		1684	canonical = F_CANONICAL
		1685	space_before = F_SPACE_BEFORE
		1686	space_after = F_SPACE_AFTER
		1687	pretty = F_PRETTY
		1688	allow_nonref = F_ALLOW_NONREF
		1689	shrink = F_SHRINK
		1690	allow_blessed = F_ALLOW_BLESSED
		1691	convert_blessed = F_CONV_BLESSED
		1692	relaxed = F_RELAXED
		1693	allow_unknown = F_ALLOW_UNKNOWN
		1694	PPCODE:
		1695	{
		1696	if (enable)
		1697	self->flags |= ix;
		1698	else
		1699	self->flags &= ~ix;
		1700
		1701	XPUSHs (ST (0));
		1702	}
		1703
		1704	void get_ascii (JSON *self)
		1705	ALIAS:
		1706	get_ascii = F_ASCII
		1707	get_latin1 = F_LATIN1
		1708	get_utf8 = F_UTF8
		1709	get_indent = F_INDENT
		1710	get_canonical = F_CANONICAL
		1711	get_space_before = F_SPACE_BEFORE
		1712	get_space_after = F_SPACE_AFTER
		1713	get_allow_nonref = F_ALLOW_NONREF
		1714	get_shrink = F_SHRINK
		1715	get_allow_blessed = F_ALLOW_BLESSED
		1716	get_convert_blessed = F_CONV_BLESSED
		1717	get_relaxed = F_RELAXED
		1718	get_allow_unknown = F_ALLOW_UNKNOWN
		1719	PPCODE:
		1720	XPUSHs (boolSV (self->flags & ix));
		1721
		1722	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		1723	PPCODE:
		1724	self->max_depth = max_depth;
		1725	XPUSHs (ST (0));
		1726
		1727	U32 get_max_depth (JSON *self)
		1728	CODE:
		1729	RETVAL = self->max_depth;
939	OUTPUT:	1730	OUTPUT:
940	RETVAL	1731	RETVAL
941		1732
942	SV *ascii (SV *self, int enable = 1)	1733	void max_size (JSON *self, U32 max_size = 0)
943	ALIAS:	1734	PPCODE:
944	ascii = F_ASCII	1735	self->max_size = max_size;
945	utf8 = F_UTF8	1736	XPUSHs (ST (0));
946	indent = F_INDENT	1737
947	canonical = F_CANONICAL	1738	int get_max_size (JSON *self)
948	space_before = F_SPACE_BEFORE
949	space_after = F_SPACE_AFTER
950	pretty = F_PRETTY
951	allow_nonref = F_ALLOW_NONREF
952	shrink = F_SHRINK
953	CODE:	1739	CODE:
954	{	1740	RETVAL = self->max_size;
955	UV *uv = SvJSON (self);
956	if (enable)
957	*uv |= ix;
958	else
959	*uv &= ~ix;
960
961	RETVAL = newSVsv (self);
962	}
963	OUTPUT:	1741	OUTPUT:
964	RETVAL	1742	RETVAL
965		1743
966	void encode (SV *self, SV *scalar)	1744	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
967	PPCODE:	1745	PPCODE:
968	XPUSHs (encode_json (scalar, *SvJSON (self)));	1746	{
		1747	SvREFCNT_dec (self->cb_object);
		1748	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
969		1749
970	void decode (SV *self, SV *jsonstr)	1750	XPUSHs (ST (0));
		1751	}
		1752
		1753	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
971	PPCODE:	1754	PPCODE:
		1755	{
		1756	if (!self->cb_sk_object)
		1757	self->cb_sk_object = newHV ();
		1758
		1759	if (SvOK (cb))
		1760	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1761	else
		1762	{
		1763	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1764
		1765	if (!HvKEYS (self->cb_sk_object))
		1766	{
		1767	SvREFCNT_dec (self->cb_sk_object);
		1768	self->cb_sk_object = 0;
		1769	}
		1770	}
		1771
		1772	XPUSHs (ST (0));
		1773	}
		1774
		1775	void encode (JSON *self, SV *scalar)
		1776	PPCODE:
		1777	XPUSHs (encode_json (scalar, self));
		1778
		1779	void decode (JSON *self, SV *jsonstr)
		1780	PPCODE:
972	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1781	XPUSHs (decode_json (jsonstr, self, 0));
		1782
		1783	void decode_prefix (JSON *self, SV *jsonstr)
		1784	PPCODE:
		1785	{
		1786	STRLEN offset;
		1787	EXTEND (SP, 2);
		1788	PUSHs (decode_json (jsonstr, self, &offset));
		1789	PUSHs (sv_2mortal (newSVuv (offset)));
		1790	}
		1791
		1792	void incr_parse (JSON *self, SV *jsonstr = 0)
		1793	PPCODE:
		1794	{
		1795	if (!self->incr_text)
		1796	self->incr_text = newSVpvn ("", 0);
		1797
		1798	// append data, if any
		1799	if (jsonstr)
		1800	{
		1801	if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text))
		1802	{
		1803	/* utf-8-ness differs, need to upgrade */
		1804	sv_utf8_upgrade (self->incr_text);
		1805
		1806	if (self->incr_pos)
		1807	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1808	- (U8 *)SvPVX (self->incr_text);
		1809	}
		1810
		1811	{
		1812	STRLEN len;
		1813	const char *str = SvPV (jsonstr, len);
		1814	STRLEN cur = SvCUR (self->incr_text);
		1815
		1816	if (SvLEN (self->incr_text) <= cur + len)
		1817	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1818
		1819	Move (str, SvEND (self->incr_text), len, char);
		1820	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1821	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1822	}
		1823	}
		1824
		1825	if (GIMME_V != G_VOID)
		1826	do
		1827	{
		1828	STRLEN offset;
		1829
		1830	if (!INCR_DONE (self))
		1831	{
		1832	incr_parse (self);
		1833
		1834	if (self->incr_pos > self->max_size && self->max_size)
		1835	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1836	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1837
		1838	if (!INCR_DONE (self))
		1839	break;
		1840	}
		1841
		1842	XPUSHs (decode_json (self->incr_text, self, &offset));
		1843
		1844	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
		1845	self->incr_pos -= offset;
		1846	self->incr_nest = 0;
		1847	self->incr_mode = 0;
		1848	}
		1849	while (GIMME_V == G_ARRAY);
		1850	}
		1851
		1852	SV *incr_text (JSON *self)
		1853	ATTRS: lvalue
		1854	CODE:
		1855	{
		1856	if (self->incr_pos)
		1857	croak ("incr_text can not be called when the incremental parser already started parsing");
		1858
		1859	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		1860	}
		1861	OUTPUT:
		1862	RETVAL
		1863
		1864	void incr_skip (JSON *self)
		1865	CODE:
		1866	{
		1867	if (self->incr_pos)
		1868	{
		1869	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1870	self->incr_pos = 0;
		1871	self->incr_nest = 0;
		1872	self->incr_mode = 0;
		1873	}
		1874	}
		1875
		1876	void incr_reset (JSON *self)
		1877	CODE:
		1878	{
		1879	SvREFCNT_dec (self->incr_text);
		1880	self->incr_text = 0;
		1881	self->incr_pos = 0;
		1882	self->incr_nest = 0;
		1883	self->incr_mode = 0;
		1884	}
		1885
		1886	void DESTROY (JSON *self)
		1887	CODE:
		1888	SvREFCNT_dec (self->cb_sk_object);
		1889	SvREFCNT_dec (self->cb_object);
		1890	SvREFCNT_dec (self->incr_text);
973		1891
974	PROTOTYPES: ENABLE	1892	PROTOTYPES: ENABLE
975		1893
976	void to_json (SV *scalar)	1894	void encode_json (SV *scalar)
		1895	ALIAS:
		1896	to_json_ = 0
		1897	encode_json = F_UTF8
977	PPCODE:	1898	PPCODE:
		1899	{
		1900	JSON json;
		1901	json_init (&json);
		1902	json.flags |= ix;
978	XPUSHs (encode_json (scalar, F_UTF8));	1903	XPUSHs (encode_json (scalar, &json));
		1904	}
979		1905
980	void from_json (SV *jsonstr)	1906	void decode_json (SV *jsonstr)
		1907	ALIAS:
		1908	from_json_ = 0
		1909	decode_json = F_UTF8
981	PPCODE:	1910	PPCODE:
		1911	{
		1912	JSON json;
		1913	json_init (&json);
		1914	json.flags |= ix;
982	XPUSHs (decode_json (jsonstr, F_UTF8));	1915	XPUSHs (decode_json (jsonstr, &json, 0));
		1916	}
983		1917
		1918

Diff Legend

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