ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.7 by root, Fri Mar 23 15:57:18 2007 UTC vs.
Revision 1.83 by root, Wed Mar 26 22:54:38 2008 UTC

1	#include "EXTERN.h"	1	#include "EXTERN.h"
2	#include "perl.h"	2	#include "perl.h"
3	#include "XSUB.h"	3	#include "XSUB.h"
4		4
5	#include "assert.h"	5	#include <assert.h>
6	#include "string.h"	6	#include <string.h>
7	#include "stdlib.h"	7	#include <stdlib.h>
		8	#include <stdio.h>
		9	#include <limits.h>
		10	#include <float.h>
8		11
		12	#if defined(__BORLANDC__) || defined(_MSC_VER)
		13	# define snprintf _snprintf // C compilers have this in stdio.h
		14	#endif
		15
		16	// some old perls do not have this, try to make it work, no
		17	// guarentees, though. if it breaks, you get to keep the pieces.
		18	#ifndef UTF8_MAXBYTES
		19	# define UTF8_MAXBYTES 13
		20	#endif
		21
		22	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2)
		23
9	#define F_ASCII 0x00000001	24	#define F_ASCII 0x00000001UL
		25	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	26	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	27	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	28	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	29	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	30	#define F_SPACE_AFTER 0x00000040UL
15	#define F_JSON_RPC 0x00000040
16	#define F_ALLOW_NONREF 0x00000080	31	#define F_ALLOW_NONREF 0x00000100UL
17	#define F_SHRINK 0x00000100	32	#define F_SHRINK 0x00000200UL
		33	#define F_ALLOW_BLESSED 0x00000400UL
		34	#define F_CONV_BLESSED 0x00000800UL
		35	#define F_RELAXED 0x00001000UL
		36
		37	#define F_MAXDEPTH 0xf8000000UL
		38	#define S_MAXDEPTH 27
		39	#define F_MAXSIZE 0x01f00000UL
		40	#define S_MAXSIZE 20
		41	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
		42
		43	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
		44	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
18		45
19	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	46	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
20	#define F_DEFAULT 0	47	#define F_DEFAULT (9UL << S_MAXDEPTH)
21		48
22	#define INIT_SIZE 32 // initial scalar size to be allocated	49	#define INIT_SIZE 32 // initial scalar size to be allocated
		50	#define INDENT_STEP 3 // spaces per indentation level
		51
		52	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
23		53
24	#define SB do {	54	#define SB do {
25	#define SE } while (0)	55	#define SE } while (0)
26		56
27	static HV *json_stash;	57	#if __GNUC__ >= 3
		58	# define expect(expr,value) __builtin_expect ((expr), (value))
		59	# define INLINE static inline
		60	#else
		61	# define expect(expr,value) (expr)
		62	# define INLINE static
		63	#endif
		64
		65	#define expect_false(expr) expect ((expr) != 0, 0)
		66	#define expect_true(expr) expect ((expr) != 0, 1)
		67
		68	#define IN_RANGE_INC(type,val,beg,end) \
		69	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		70	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		71
		72	#ifdef USE_ITHREADS
		73	# define JSON_SLOW 1
		74	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		75	#else
		76	# define JSON_SLOW 0
		77	# define JSON_STASH json_stash
		78	#endif
		79
		80	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		81	static SV *json_true, *json_false;
		82
		83	enum {
		84	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		85	INCR_M_STR, // inside string
		86	INCR_M_BS, // inside backslash
		87	INCR_M_JSON // outside anything, count nesting
		88	};
		89
		90	#define INCR_DONE(json) (!(json)->incr_nest && (json)->incr_mode == INCR_M_JSON)
		91
		92	typedef struct {
		93	U32 flags;
		94	SV *cb_object;
		95	HV *cb_sk_object;
		96
		97	// for the incremental parser
		98	SV *incr_text; // the source text so far
		99	STRLEN incr_pos; // the current offset into the text
		100	int incr_nest; // {[]}-nesting level
		101	int incr_mode;
		102	} JSON;
		103
		104	/////////////////////////////////////////////////////////////////////////////
		105	// utility functions
		106
		107	INLINE SV *
		108	get_bool (const char *name)
		109	{
		110	SV *sv = get_sv (name, 1);
		111
		112	SvREADONLY_on (sv);
		113	SvREADONLY_on (SvRV (sv));
		114
		115	return sv;
		116	}
		117
		118	INLINE void
		119	shrink (SV *sv)
		120	{
		121	sv_utf8_downgrade (sv, 1);
		122	if (SvLEN (sv) > SvCUR (sv) + 1)
		123	{
		124	#ifdef SvPV_shrink_to_cur
		125	SvPV_shrink_to_cur (sv);
		126	#elif defined (SvPV_renew)
		127	SvPV_renew (sv, SvCUR (sv) + 1);
		128	#endif
		129	}
		130	}
		131
		132	// decode an utf-8 character and return it, or (UV)-1 in
		133	// case of an error.
		134	// we special-case "safe" characters from U+80 .. U+7FF,
		135	// but use the very good perl function to parse anything else.
		136	// note that we never call this function for a ascii codepoints
		137	INLINE UV
		138	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		139	{
		140	if (expect_true (len >= 2
		141	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
		142	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
		143	{
		144	*clen = 2;
		145	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		146	}
		147	else
		148	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		149	}
		150
		151	// likewise for encoding, also never called for ascii codepoints
		152	// this function takes advantage of this fact, although current gccs
		153	// seem to optimise the check for >= 0x80 away anyways
		154	INLINE unsigned char *
		155	encode_utf8 (unsigned char *s, UV ch)
		156	{
		157	if (expect_false (ch < 0x000080))
		158	*s++ = ch;
		159	else if (expect_true (ch < 0x000800))
		160	*s++ = 0xc0 | ( ch >> 6),
		161	*s++ = 0x80 | ( ch & 0x3f);
		162	else if ( ch < 0x010000)
		163	*s++ = 0xe0 | ( ch >> 12),
		164	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		165	*s++ = 0x80 | ( ch & 0x3f);
		166	else if ( ch < 0x110000)
		167	*s++ = 0xf0 | ( ch >> 18),
		168	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		169	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		170	*s++ = 0x80 | ( ch & 0x3f);
		171
		172	return s;
		173	}
		174
		175	/////////////////////////////////////////////////////////////////////////////
		176	// encoder
28		177
29	// structure used for encoding JSON	178	// structure used for encoding JSON
30	typedef struct	179	typedef struct
31	{	180	{
32	char *cur;	181	char *cur; // SvPVX (sv) + current output position
33	STRLEN len; // SvLEN (sv)
34	char *end; // SvEND (sv)	182	char *end; // SvEND (sv)
35	SV *sv;	183	SV *sv; // result scalar
36	UV flags;	184	JSON json;
37	int max_recurse;	185	U32 indent; // indentation level
38	int indent;	186	U32 maxdepth; // max. indentation/recursion level
		187	UV limit; // escape character values >= this value when encoding
39	} enc_t;	188	} enc_t;
40		189
41	// structure used for decoding JSON	190	INLINE void
42	typedef struct
43	{
44	char *cur;
45	char *end;
46	const char *err;
47	UV flags;
48	} dec_t;
49
50	static UV *
51	SvJSON (SV *sv)
52	{
53	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
54	croak ("object is not of type JSON::XS");
55
56	return &SvUVX (SvRV (sv));
57	}
58
59	static void
60	shrink (SV *sv)
61	{
62	sv_utf8_downgrade (sv, 1);
63	#ifdef SvPV_shrink_to_cur
64	SvPV_shrink_to_cur (sv);
65	#endif
66	}
67
68	/////////////////////////////////////////////////////////////////////////////
69
70	static void
71	need (enc_t *enc, STRLEN len)	191	need (enc_t *enc, STRLEN len)
72	{	192	{
73	if (enc->cur + len >= enc->end)	193	if (expect_false (enc->cur + len >= enc->end))
74	{	194	{
75	STRLEN cur = enc->cur - SvPVX (enc->sv);	195	STRLEN cur = enc->cur - SvPVX (enc->sv);
76	SvGROW (enc->sv, cur + len + 1);	196	SvGROW (enc->sv, cur + len + 1);
77	enc->cur = SvPVX (enc->sv) + cur;	197	enc->cur = SvPVX (enc->sv) + cur;
78	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	198	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
79	}	199	}
80	}	200	}
81		201
82	static void	202	INLINE void
83	encode_ch (enc_t *enc, char ch)	203	encode_ch (enc_t *enc, char ch)
84	{	204	{
85	need (enc, 1);	205	need (enc, 1);
86	*enc->cur++ = ch;	206	*enc->cur++ = ch;
87	}	207	}
…		…
95		215
96	while (str < end)	216	while (str < end)
97	{	217	{
98	unsigned char ch = *(unsigned char *)str;	218	unsigned char ch = *(unsigned char *)str;
99		219
100	if (ch >= 0x20 && ch < 0x80) // most common case	220	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
101	{	221	{
102	if (ch == '"') // but with slow exceptions	222	if (expect_false (ch == '"')) // but with slow exceptions
103	{	223	{
104	need (enc, len += 1);	224	need (enc, len += 1);
105	*enc->cur++ = '\\';	225	*enc->cur++ = '\\';
106	*enc->cur++ = '"';	226	*enc->cur++ = '"';
107	}	227	}
108	else if (ch == '\\')	228	else if (expect_false (ch == '\\'))
109	{	229	{
110	need (enc, len += 1);	230	need (enc, len += 1);
111	*enc->cur++ = '\\';	231	*enc->cur++ = '\\';
112	*enc->cur++ = '\\';	232	*enc->cur++ = '\\';
113	}	233	}
…		…
131	STRLEN clen;	251	STRLEN clen;
132	UV uch;	252	UV uch;
133		253
134	if (is_utf8)	254	if (is_utf8)
135	{	255	{
136	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	256	uch = decode_utf8 (str, end - str, &clen);
137	if (clen == (STRLEN)-1)	257	if (clen == (STRLEN)-1)
138	croak ("malformed UTF-8 character in string, cannot convert to JSON");	258	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
139	}	259	}
140	else	260	else
141	{	261	{
142	uch = ch;	262	uch = ch;
143	clen = 1;	263	clen = 1;
144	}	264	}
145		265
146	if (uch < 0x80 || enc->flags & F_ASCII)	266	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
147	{	267	{
148	if (uch > 0xFFFFUL)	268	if (uch >= 0x10000UL)
149	{	269	{
		270	if (uch >= 0x110000UL)
		271	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		272
150	need (enc, len += 11);	273	need (enc, len += 11);
151	sprintf (enc->cur, "\\u%04x\\u%04x",	274	sprintf (enc->cur, "\\u%04x\\u%04x",
152	(uch - 0x10000) / 0x400 + 0xD800,	275	(int)((uch - 0x10000) / 0x400 + 0xD800),
153	(uch - 0x10000) % 0x400 + 0xDC00);	276	(int)((uch - 0x10000) % 0x400 + 0xDC00));
154	enc->cur += 12;	277	enc->cur += 12;
155	}	278	}
156	else	279	else
157	{	280	{
158	static char hexdigit [16] = "0123456789abcdef";	281	static char hexdigit [16] = "0123456789abcdef";
…		…
165	*enc->cur++ = hexdigit [(uch >> 0) & 15];	288	*enc->cur++ = hexdigit [(uch >> 0) & 15];
166	}	289	}
167		290
168	str += clen;	291	str += clen;
169	}	292	}
		293	else if (enc->json.flags & F_LATIN1)
		294	{
		295	*enc->cur++ = uch;
		296	str += clen;
		297	}
170	else if (is_utf8)	298	else if (is_utf8)
171	{	299	{
172	need (enc, len += clen);	300	need (enc, len += clen);
173	do	301	do
174	{	302	{
…		…
176	}	304	}
177	while (--clen);	305	while (--clen);
178	}	306	}
179	else	307	else
180	{	308	{
181	need (enc, len += 10); // never more than 11 bytes needed	309	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
182	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	310	enc->cur = encode_utf8 (enc->cur, uch);
183	++str;	311	++str;
184	}	312	}
185	}	313	}
186	}	314	}
187	}	315	}
188		316
189	--len;	317	--len;
190	}	318	}
191	}	319	}
192		320
193	#define INDENT SB \	321	INLINE void
		322	encode_indent (enc_t *enc)
		323	{
194	if (enc->flags & F_INDENT) \	324	if (enc->json.flags & F_INDENT)
195	{ \	325	{
196	int i_; \	326	int spaces = enc->indent * INDENT_STEP;
197	need (enc, enc->indent); \	327
198	for (i_ = enc->indent * 3; i_--; )\	328	need (enc, spaces);
		329	memset (enc->cur, ' ', spaces);
		330	enc->cur += spaces;
		331	}
		332	}
		333
		334	INLINE void
		335	encode_space (enc_t *enc)
		336	{
		337	need (enc, 1);
		338	encode_ch (enc, ' ');
		339	}
		340
		341	INLINE void
		342	encode_nl (enc_t *enc)
		343	{
		344	if (enc->json.flags & F_INDENT)
		345	{
		346	need (enc, 1);
199	encode_ch (enc, ' '); \	347	encode_ch (enc, '\n');
200	} \	348	}
201	SE	349	}
202		350
203	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	351	INLINE void
204	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	352	encode_comma (enc_t *enc)
205	#define COMMA SB \	353	{
206	encode_ch (enc, ','); \	354	encode_ch (enc, ',');
		355
207	if (enc->flags & F_INDENT) \	356	if (enc->json.flags & F_INDENT)
208	NL; \	357	encode_nl (enc);
209	else if (enc->flags & F_SPACE_AFTER) \	358	else if (enc->json.flags & F_SPACE_AFTER)
210	SPACE; \	359	encode_space (enc);
211	SE	360	}
212		361
213	static void encode_sv (enc_t *enc, SV *sv);	362	static void encode_sv (enc_t *enc, SV *sv);
214		363
215	static void	364	static void
216	encode_av (enc_t *enc, AV *av)	365	encode_av (enc_t *enc, AV *av)
217	{	366	{
218	int i, len = av_len (av);	367	int i, len = av_len (av);
219		368
		369	if (enc->indent >= enc->maxdepth)
		370	croak ("data structure too deep (hit recursion limit)");
		371
220	encode_ch (enc, '['); NL;	372	encode_ch (enc, '[');
221	++enc->indent;	373
		374	if (len >= 0)
		375	{
		376	encode_nl (enc); ++enc->indent;
222		377
223	for (i = 0; i <= len; ++i)	378	for (i = 0; i <= len; ++i)
224	{	379	{
225	INDENT;	380	SV **svp = av_fetch (av, i, 0);
226	encode_sv (enc, *av_fetch (av, i, 0));
227		381
		382	encode_indent (enc);
		383
		384	if (svp)
		385	encode_sv (enc, *svp);
		386	else
		387	encode_str (enc, "null", 4, 0);
		388
228	if (i < len)	389	if (i < len)
229	COMMA;	390	encode_comma (enc);
230	}	391	}
231		392
232	NL;	393	encode_nl (enc); --enc->indent; encode_indent (enc);
233		394	}
234	--enc->indent;	395
235	INDENT; encode_ch (enc, ']');	396	encode_ch (enc, ']');
236	}	397	}
237		398
238	static void	399	static void
239	encode_he (enc_t *enc, HE *he)	400	encode_hk (enc_t *enc, HE *he)
240	{	401	{
241	encode_ch (enc, '"');	402	encode_ch (enc, '"');
242		403
243	if (HeKLEN (he) == HEf_SVKEY)	404	if (HeKLEN (he) == HEf_SVKEY)
244	{	405	{
…		…
254	else	415	else
255	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	416	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
256		417
257	encode_ch (enc, '"');	418	encode_ch (enc, '"');
258		419
259	if (enc->flags & F_SPACE_BEFORE) SPACE;	420	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
260	encode_ch (enc, ':');	421	encode_ch (enc, ':');
261	if (enc->flags & F_SPACE_AFTER ) SPACE;	422	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
262	encode_sv (enc, HeVAL (he));
263	}	423	}
264		424
265	// compare hash entries, used when all keys are bytestrings	425	// compare hash entries, used when all keys are bytestrings
266	static int	426	static int
267	he_cmp_fast (const void *a_, const void *b_)	427	he_cmp_fast (const void *a_, const void *b_)
…		…
272	HE *b = *(HE **)b_;	432	HE *b = *(HE **)b_;
273		433
274	STRLEN la = HeKLEN (a);	434	STRLEN la = HeKLEN (a);
275	STRLEN lb = HeKLEN (b);	435	STRLEN lb = HeKLEN (b);
276		436
277	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	437	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
278	cmp = la < lb ? -1 : la == lb ? 0 : 1;	438	cmp = lb - la;
279		439
280	return cmp;	440	return cmp;
281	}	441	}
282		442
283	// compare hash entries, used when some keys are sv's or utf-x	443	// compare hash entries, used when some keys are sv's or utf-x
284	static int	444	static int
285	he_cmp_slow (const void *a, const void *b)	445	he_cmp_slow (const void *a, const void *b)
286	{	446	{
287	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	447	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
288	}	448	}
289		449
290	static void	450	static void
291	encode_hv (enc_t *enc, HV *hv)	451	encode_hv (enc_t *enc, HV *hv)
292	{	452	{
293	int count, i;	453	HE *he;
294		454
295	encode_ch (enc, '{'); NL; ++enc->indent;	455	if (enc->indent >= enc->maxdepth)
		456	croak ("data structure too deep (hit recursion limit)");
296		457
297	if ((count = hv_iterinit (hv)))	458	encode_ch (enc, '{');
298	{	459
299	// for canonical output we have to sort by keys first	460	// for canonical output we have to sort by keys first
300	// actually, this is mostly due to the stupid so-called	461	// actually, this is mostly due to the stupid so-called
301	// security workaround added somewhere in 5.8.x.	462	// security workaround added somewhere in 5.8.x.
302	// that randomises hash orderings	463	// that randomises hash orderings
303	if (enc->flags & F_CANONICAL)	464	if (enc->json.flags & F_CANONICAL)
		465	{
		466	int count = hv_iterinit (hv);
		467
		468	if (SvMAGICAL (hv))
304	{	469	{
305	HE *he, *hes [count];	470	// need to count by iterating. could improve by dynamically building the vector below
		471	// but I don't care for the speed of this special case.
		472	// note also that we will run into undefined behaviour when the two iterations
		473	// do not result in the same count, something I might care for in some later release.
		474
		475	count = 0;
		476	while (hv_iternext (hv))
		477	++count;
		478
		479	hv_iterinit (hv);
		480	}
		481
		482	if (count)
		483	{
306	int fast = 1;	484	int i, fast = 1;
		485	#if defined(__BORLANDC__) || defined(_MSC_VER)
		486	HE **hes = _alloca (count * sizeof (HE));
		487	#else
		488	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		489	#endif
307		490
308	i = 0;	491	i = 0;
309	while ((he = hv_iternext (hv)))	492	while ((he = hv_iternext (hv)))
310	{	493	{
311	hes [i++] = he;	494	hes [i++] = he;
…		…
317		500
318	if (fast)	501	if (fast)
319	qsort (hes, count, sizeof (HE *), he_cmp_fast);	502	qsort (hes, count, sizeof (HE *), he_cmp_fast);
320	else	503	else
321	{	504	{
322	// hack to disable "use bytes"	505	// hack to forcefully disable "use bytes"
323	COP *oldcop = PL_curcop, cop;	506	COP cop = *PL_curcop;
324	cop.op_private = 0;	507	cop.op_private = 0;
		508
		509	ENTER;
		510	SAVETMPS;
		511
		512	SAVEVPTR (PL_curcop);
325	PL_curcop = &cop;	513	PL_curcop = &cop;
326		514
327	SAVETMPS;
328	qsort (hes, count, sizeof (HE *), he_cmp_slow);	515	qsort (hes, count, sizeof (HE *), he_cmp_slow);
		516
329	FREETMPS;	517	FREETMPS;
330		518	LEAVE;
331	PL_curcop = oldcop;
332	}	519	}
333		520
334	for (i = 0; i < count; ++i)	521	encode_nl (enc); ++enc->indent;
		522
		523	while (count--)
335	{	524	{
336	INDENT;	525	encode_indent (enc);
		526	he = hes [count];
337	encode_he (enc, hes [i]);	527	encode_hk (enc, he);
		528	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
338		529
339	if (i < count - 1)	530	if (count)
340	COMMA;	531	encode_comma (enc);
341	}	532	}
342		533
		534	encode_nl (enc); --enc->indent; encode_indent (enc);
		535	}
		536	}
		537	else
		538	{
		539	if (hv_iterinit (hv) || SvMAGICAL (hv))
		540	if ((he = hv_iternext (hv)))
343	NL;	541	{
		542	encode_nl (enc); ++enc->indent;
		543
		544	for (;;)
		545	{
		546	encode_indent (enc);
		547	encode_hk (enc, he);
		548	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		549
		550	if (!(he = hv_iternext (hv)))
		551	break;
		552
		553	encode_comma (enc);
		554	}
		555
		556	encode_nl (enc); --enc->indent; encode_indent (enc);
		557	}
		558	}
		559
		560	encode_ch (enc, '}');
		561	}
		562
		563	// encode objects, arrays and special \0=false and \1=true values.
		564	static void
		565	encode_rv (enc_t *enc, SV *sv)
		566	{
		567	svtype svt;
		568
		569	SvGETMAGIC (sv);
		570	svt = SvTYPE (sv);
		571
		572	if (expect_false (SvOBJECT (sv)))
		573	{
		574	HV *stash = !JSON_SLOW || json_boolean_stash
		575	? json_boolean_stash
		576	: gv_stashpv ("JSON::XS::Boolean", 1);
		577
		578	if (SvSTASH (sv) == stash)
		579	{
		580	if (SvIV (sv))
		581	encode_str (enc, "true", 4, 0);
		582	else
		583	encode_str (enc, "false", 5, 0);
344	}	584	}
345	else	585	else
346	{	586	{
347	SV *sv;	587	#if 0
348	HE *he = hv_iternext (hv);	588	if (0 && sv_derived_from (rv, "JSON::Literal"))
349
350	for (;;)
351	{	589	{
352	INDENT;	590	// not yet
353	encode_he (enc, he);
354
355	if (!(he = hv_iternext (hv)))
356	break;
357
358	COMMA;
359	}	591	}
360		592	#endif
		593	if (enc->json.flags & F_CONV_BLESSED)
361	NL;	594	{
		595	// we re-bless the reference to get overload and other niceties right
		596	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
		597
		598	if (to_json)
		599	{
		600	dSP;
		601
		602	ENTER; SAVETMPS; PUSHMARK (SP);
		603	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		604
		605	// calling with G_SCALAR ensures that we always get a 1 return value
		606	PUTBACK;
		607	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		608	SPAGAIN;
		609
		610	// catch this surprisingly common error
		611	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		612	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		613
		614	sv = POPs;
		615	PUTBACK;
		616
		617	encode_sv (enc, sv);
		618
		619	FREETMPS; LEAVE;
		620	}
		621	else if (enc->json.flags & F_ALLOW_BLESSED)
		622	encode_str (enc, "null", 4, 0);
		623	else
		624	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		625	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		626	}
		627	else if (enc->json.flags & F_ALLOW_BLESSED)
		628	encode_str (enc, "null", 4, 0);
		629	else
		630	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		631	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
362	}	632	}
363	}	633	}
		634	else if (svt == SVt_PVHV)
		635	encode_hv (enc, (HV *)sv);
		636	else if (svt == SVt_PVAV)
		637	encode_av (enc, (AV *)sv);
		638	else if (svt < SVt_PVAV)
		639	{
		640	STRLEN len = 0;
		641	char *pv = svt ? SvPV (sv, len) : 0;
364		642
365	--enc->indent; INDENT; encode_ch (enc, '}');	643	if (len == 1 && *pv == '1')
		644	encode_str (enc, "true", 4, 0);
		645	else if (len == 1 && *pv == '0')
		646	encode_str (enc, "false", 5, 0);
		647	else
		648	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		649	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		650	}
		651	else
		652	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		653	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
366	}	654	}
367		655
368	static void	656	static void
369	encode_sv (enc_t *enc, SV *sv)	657	encode_sv (enc_t *enc, SV *sv)
370	{	658	{
…		…
378	encode_str (enc, str, len, SvUTF8 (sv));	666	encode_str (enc, str, len, SvUTF8 (sv));
379	encode_ch (enc, '"');	667	encode_ch (enc, '"');
380	}	668	}
381	else if (SvNOKp (sv))	669	else if (SvNOKp (sv))
382	{	670	{
		671	// trust that perl will do the right thing w.r.t. JSON syntax.
383	need (enc, NV_DIG + 32);	672	need (enc, NV_DIG + 32);
384	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	673	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
385	enc->cur += strlen (enc->cur);	674	enc->cur += strlen (enc->cur);
386	}	675	}
387	else if (SvIOKp (sv))	676	else if (SvIOKp (sv))
388	{	677	{
		678	// we assume we can always read an IV as a UV and vice versa
		679	// we assume two's complement
		680	// we assume no aliasing issues in the union
		681	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		682	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		683	{
		684	// optimise the "small number case"
		685	// code will likely be branchless and use only a single multiplication
		686	// works for numbers up to 59074
		687	I32 i = SvIVX (sv);
		688	U32 u;
		689	char digit, nz = 0;
		690
389	need (enc, 64);	691	need (enc, 6);
		692
		693	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		694	u = i < 0 ? -i : i;
		695
		696	// convert to 4.28 fixed-point representation
		697	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		698
		699	// now output digit by digit, each time masking out the integer part
		700	// and multiplying by 5 while moving the decimal point one to the right,
		701	// resulting in a net multiplication by 10.
		702	// we always write the digit to memory but conditionally increment
		703	// the pointer, to enable the use of conditional move instructions.
		704	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		705	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		706	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		707	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		708	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		709	}
		710	else
		711	{
		712	// large integer, use the (rather slow) snprintf way.
		713	need (enc, IVUV_MAXCHARS);
390	enc->cur +=	714	enc->cur +=
391	SvIsUV(sv)	715	SvIsUV(sv)
392	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	716	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
393	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	717	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		718	}
394	}	719	}
395	else if (SvROK (sv))	720	else if (SvROK (sv))
396	{	721	encode_rv (enc, SvRV (sv));
397	if (!--enc->max_recurse)
398	croak ("data structure too deep (hit recursion limit)");
399
400	sv = SvRV (sv);
401
402	switch (SvTYPE (sv))
403	{
404	case SVt_PVAV: encode_av (enc, (AV *)sv); break;
405	case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
406
407	default:
408	croak ("JSON can only represent references to arrays or hashes");
409	}
410	}
411	else if (!SvOK (sv))	722	else if (!SvOK (sv))
412	encode_str (enc, "null", 4, 0);	723	encode_str (enc, "null", 4, 0);
413	else	724	else
414	croak ("encountered perl type that JSON cannot handle");	725	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
		726	SvPV_nolen (sv), SvFLAGS (sv));
415	}	727	}
416		728
417	static SV *	729	static SV *
418	encode_json (SV *scalar, UV flags)	730	encode_json (SV *scalar, JSON *json)
419	{	731	{
420	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
421	croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");
422
423	enc_t enc;	732	enc_t enc;
424	enc.flags = flags;	733
		734	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
		735	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
		736
		737	enc.json = *json;
425	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	738	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
426	enc.cur = SvPVX (enc.sv);	739	enc.cur = SvPVX (enc.sv);
427	enc.end = SvEND (enc.sv);	740	enc.end = SvEND (enc.sv);
428	enc.max_recurse = 0;
429	enc.indent = 0;	741	enc.indent = 0;
		742	enc.maxdepth = DEC_DEPTH (enc.json.flags);
		743	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		744	: enc.json.flags & F_LATIN1 ? 0x000100UL
		745	: 0x110000UL;
430		746
431	SvPOK_only (enc.sv);	747	SvPOK_only (enc.sv);
432	encode_sv (&enc, scalar);	748	encode_sv (&enc, scalar);
433		749
		750	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		751	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		752
434	if (!(flags & (F_ASCII | F_UTF8)))	753	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
435	SvUTF8_on (enc.sv);	754	SvUTF8_on (enc.sv);
436		755
437	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
438
439	if (enc.flags & F_SHRINK)	756	if (enc.json.flags & F_SHRINK)
440	shrink (enc.sv);	757	shrink (enc.sv);
441		758
442	return enc.sv;	759	return enc.sv;
443	}	760	}
444		761
445	/////////////////////////////////////////////////////////////////////////////	762	/////////////////////////////////////////////////////////////////////////////
		763	// decoder
446		764
447	#define WS \	765	// structure used for decoding JSON
		766	typedef struct
		767	{
		768	char *cur; // current parser pointer
		769	char *end; // end of input string
		770	const char *err; // parse error, if != 0
		771	JSON json;
		772	U32 depth; // recursion depth
		773	U32 maxdepth; // recursion depth limit
		774	} dec_t;
		775
		776	INLINE void
		777	decode_comment (dec_t *dec)
		778	{
		779	// only '#'-style comments allowed a.t.m.
		780
		781	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		782	++dec->cur;
		783	}
		784
		785	INLINE void
		786	decode_ws (dec_t *dec)
		787	{
448	for (;;) \	788	for (;;)
449	{ \	789	{
450	char ch = *dec->cur; \	790	char ch = *dec->cur;
		791
451	if (ch > 0x20 \	792	if (ch > 0x20)
		793	{
		794	if (expect_false (ch == '#'))
		795	{
		796	if (dec->json.flags & F_RELAXED)
		797	decode_comment (dec);
		798	else
		799	break;
		800	}
		801	else
		802	break;
		803	}
452	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	804	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
453	break; \	805	break; // parse error, but let higher level handle it, gives better error messages
		806
454	++dec->cur; \	807	++dec->cur;
455	}	808	}
		809	}
456		810
457	#define ERR(reason) SB dec->err = reason; goto fail; SE	811	#define ERR(reason) SB dec->err = reason; goto fail; SE
		812
458	#define EXPECT_CH(ch) SB \	813	#define EXPECT_CH(ch) SB \
459	if (*dec->cur != ch) \	814	if (*dec->cur != ch) \
460	ERR (# ch " expected"); \	815	ERR (# ch " expected"); \
461	++dec->cur; \	816	++dec->cur; \
462	SE	817	SE
463		818
		819	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		820	#define DEC_DEC_DEPTH --dec->depth
		821
464	static SV *decode_sv (dec_t *dec);	822	static SV *decode_sv (dec_t *dec);
465		823
466	static signed char decode_hexdigit[256];	824	static signed char decode_hexdigit[256];
467		825
468	static UV	826	static UV
469	decode_4hex (dec_t *dec)	827	decode_4hex (dec_t *dec)
470	{	828	{
471	signed char d1, d2, d3, d4;	829	signed char d1, d2, d3, d4;
		830	unsigned char *cur = (unsigned char *)dec->cur;
472		831
473	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	832	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
474	if (d1 < 0) ERR ("four hexadecimal digits expected");	833	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
475	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	834	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
476	if (d2 < 0) ERR ("four hexadecimal digits expected");	835	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
477	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
478	if (d3 < 0) ERR ("four hexadecimal digits expected");
479	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
480	if (d4 < 0) ERR ("four hexadecimal digits expected");
481		836
482	dec->cur += 4;	837	dec->cur += 4;
483		838
484	return ((UV)d1) << 12	839	return ((UV)d1) << 12
485	| ((UV)d2) << 8	840	| ((UV)d2) << 8
…		…
488		843
489	fail:	844	fail:
490	return (UV)-1;	845	return (UV)-1;
491	}	846	}
492		847
493	#define APPEND_GROW(n) SB \
494	if (cur + (n) >= end) \
495	{ \
496	STRLEN ofs = cur - SvPVX (sv); \
497	SvGROW (sv, ofs + (n) + 1); \
498	cur = SvPVX (sv) + ofs; \
499	end = SvEND (sv); \
500	} \
501	SE
502
503	#define APPEND_CH(ch) SB \
504	APPEND_GROW (1); \
505	*cur++ = (ch); \
506	SE
507
508	static SV *	848	static SV *
509	decode_str (dec_t *dec)	849	decode_str (dec_t *dec)
510	{	850	{
511	SV *sv = NEWSV (0,2);	851	SV *sv = 0;
512	int utf8 = 0;	852	int utf8 = 0;
513	char *cur = SvPVX (sv);	853	char *dec_cur = dec->cur;
514	char *end = SvEND (sv);
515		854
516	for (;;)	855	do
517	{	856	{
518	unsigned char ch = *(unsigned char *)dec->cur;	857	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		858	char *cur = buf;
519		859
520	if (ch == '"')	860	do
521	break;
522	else if (ch == '\\')
523	{	861	{
524	switch (*++dec->cur)	862	unsigned char ch = *(unsigned char *)dec_cur++;
		863
		864	if (expect_false (ch == '"'))
525	{	865	{
526	case '\\':	866	--dec_cur;
527	case '/':	867	break;
528	case '"': APPEND_CH (*dec->cur++); break;	868	}
529		869	else if (expect_false (ch == '\\'))
530	case 'b': APPEND_CH ('\010'); ++dec->cur; break;	870	{
531	case 't': APPEND_CH ('\011'); ++dec->cur; break;	871	switch (*dec_cur)
532	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
533	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
534	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
535
536	case 'u':
537	{	872	{
538	UV lo, hi;	873	case '\\':
539	++dec->cur;	874	case '/':
		875	case '"': *cur++ = *dec_cur++; break;
540		876
541	hi = decode_4hex (dec);	877	case 'b': ++dec_cur; *cur++ = '\010'; break;
542	if (hi == (UV)-1)	878	case 't': ++dec_cur; *cur++ = '\011'; break;
543	goto fail;	879	case 'n': ++dec_cur; *cur++ = '\012'; break;
		880	case 'f': ++dec_cur; *cur++ = '\014'; break;
		881	case 'r': ++dec_cur; *cur++ = '\015'; break;
544		882
545	// possibly a surrogate pair	883	case 'u':
546	if (hi >= 0xd800 && hi < 0xdc00)
547	{	884	{
548	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	885	UV lo, hi;
549	ERR ("missing low surrogate character in surrogate pair");	886	++dec_cur;
550		887
551	dec->cur += 2;	888	dec->cur = dec_cur;
552
553	lo = decode_4hex (dec);	889	hi = decode_4hex (dec);
		890	dec_cur = dec->cur;
554	if (lo == (UV)-1)	891	if (hi == (UV)-1)
555	goto fail;	892	goto fail;
556		893
		894	// possibly a surrogate pair
		895	if (hi >= 0xd800)
		896	if (hi < 0xdc00)
		897	{
		898	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		899	ERR ("missing low surrogate character in surrogate pair");
		900
		901	dec_cur += 2;
		902
		903	dec->cur = dec_cur;
		904	lo = decode_4hex (dec);
		905	dec_cur = dec->cur;
		906	if (lo == (UV)-1)
		907	goto fail;
		908
557	if (lo < 0xdc00 || lo >= 0xe000)	909	if (lo < 0xdc00 || lo >= 0xe000)
558	ERR ("surrogate pair expected");	910	ERR ("surrogate pair expected");
559		911
560	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	912	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		913	}
		914	else if (hi < 0xe000)
		915	ERR ("missing high surrogate character in surrogate pair");
		916
		917	if (hi >= 0x80)
		918	{
		919	utf8 = 1;
		920
		921	cur = encode_utf8 (cur, hi);
		922	}
		923	else
		924	*cur++ = hi;
561	}	925	}
562	else if (hi >= 0xdc00 && hi < 0xe000)
563	ERR ("missing high surrogate character in surrogate pair");
564
565	if (hi >= 0x80)
566	{	926	break;
567	utf8 = 1;
568		927
569	APPEND_GROW (4); // at most 4 bytes for 21 bits
570	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
571	}
572	else	928	default:
573	APPEND_CH (hi);	929	--dec_cur;
		930	ERR ("illegal backslash escape sequence in string");
574	}	931	}
575	break;	932	}
		933	else if (expect_true (ch >= 0x20 && ch < 0x80))
		934	*cur++ = ch;
		935	else if (ch >= 0x80)
		936	{
		937	STRLEN clen;
		938	UV uch;
576		939
577	default:
578	--dec->cur;	940	--dec_cur;
579	ERR ("illegal backslash escape sequence in string");	941
		942	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		943	if (clen == (STRLEN)-1)
		944	ERR ("malformed UTF-8 character in JSON string");
		945
		946	do
		947	*cur++ = *dec_cur++;
		948	while (--clen);
		949
		950	utf8 = 1;
		951	}
		952	else
		953	{
		954	--dec_cur;
		955
		956	if (!ch)
		957	ERR ("unexpected end of string while parsing JSON string");
		958	else
		959	ERR ("invalid character encountered while parsing JSON string");
580	}	960	}
581	}	961	}
582	else if (ch >= 0x20 && ch <= 0x7f)	962	while (cur < buf + SHORT_STRING_LEN);
583	APPEND_CH (*dec->cur++);	963
584	else if (ch >= 0x80)
585	{	964	{
586	STRLEN clen;	965	STRLEN len = cur - buf;
587	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
588	if (clen == (STRLEN)-1)
589	ERR ("malformed UTF-8 character in JSON string");
590		966
591	APPEND_GROW (clen);	967	if (sv)
592	do
593	{	968	{
594	*cur++ = *dec->cur++;	969	SvGROW (sv, SvCUR (sv) + len + 1);
		970	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		971	SvCUR_set (sv, SvCUR (sv) + len);
595	}	972	}
596	while (--clen);
597
598	utf8 = 1;
599	}
600	else if (dec->cur == dec->end)
601	ERR ("unexpected end of string while parsing json string");
602	else	973	else
603	ERR ("invalid character encountered");	974	sv = newSVpvn (buf, len);
604	}	975	}
		976	}
		977	while (*dec_cur != '"');
605		978
606	++dec->cur;	979	++dec_cur;
607		980
608	SvCUR_set (sv, cur - SvPVX (sv));	981	if (sv)
609		982	{
610	SvPOK_only (sv);	983	SvPOK_only (sv);
611	*SvEND (sv) = 0;	984	*SvEND (sv) = 0;
612		985
613	if (utf8)	986	if (utf8)
614	SvUTF8_on (sv);	987	SvUTF8_on (sv);
		988	}
		989	else
		990	sv = newSVpvn ("", 0);
615		991
616	if (dec->flags & F_SHRINK)	992	dec->cur = dec_cur;
617	shrink (sv);
618
619	return sv;	993	return sv;
620		994
621	fail:	995	fail:
622	SvREFCNT_dec (sv);	996	dec->cur = dec_cur;
623	return 0;	997	return 0;
624	}	998	}
625		999
626	static SV *	1000	static SV *
627	decode_num (dec_t *dec)	1001	decode_num (dec_t *dec)
…		…
685	is_nv = 1;	1059	is_nv = 1;
686	}	1060	}
687		1061
688	if (!is_nv)	1062	if (!is_nv)
689	{	1063	{
690	UV uv;	1064	int len = dec->cur - start;
691	int numtype = grok_number (start, dec->cur - start, &uv);	1065
692	if (numtype & IS_NUMBER_IN_UV)	1066	// special case the rather common 1..5-digit-int case
693	if (numtype & IS_NUMBER_NEG)	1067	if (*start == '-')
		1068	switch (len)
694	{	1069	{
695	if (uv < (UV)IV_MIN)	1070	case 2: return newSViv (-( start [1] - '0' * 1));
696	return newSViv (-(IV)uv);	1071	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		1072	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1073	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1074	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
697	}	1075	}
		1076	else
		1077	switch (len)
		1078	{
		1079	case 1: return newSViv ( start [0] - '0' * 1);
		1080	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1081	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1082	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1083	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1084	}
		1085
		1086	{
		1087	UV uv;
		1088	int numtype = grok_number (start, len, &uv);
		1089	if (numtype & IS_NUMBER_IN_UV)
		1090	if (numtype & IS_NUMBER_NEG)
		1091	{
		1092	if (uv < (UV)IV_MIN)
		1093	return newSViv (-(IV)uv);
		1094	}
698	else	1095	else
699	return newSVuv (uv);	1096	return newSVuv (uv);
700	}	1097	}
701		1098
		1099	len -= *start == '-' ? 1 : 0;
		1100
		1101	// does not fit into IV or UV, try NV
		1102	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1103	#if defined (LDBL_DIG)
		1104	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1105	#endif
		1106	)
		1107	// fits into NV without loss of precision
		1108	return newSVnv (Atof (start));
		1109
		1110	// everything else fails, convert it to a string
		1111	return newSVpvn (start, dec->cur - start);
		1112	}
		1113
		1114	// loss of precision here
702	return newSVnv (Atof (start));	1115	return newSVnv (Atof (start));
703		1116
704	fail:	1117	fail:
705	return 0;	1118	return 0;
706	}	1119	}
…		…
708	static SV *	1121	static SV *
709	decode_av (dec_t *dec)	1122	decode_av (dec_t *dec)
710	{	1123	{
711	AV *av = newAV ();	1124	AV *av = newAV ();
712		1125
713	WS;	1126	DEC_INC_DEPTH;
		1127	decode_ws (dec);
		1128
714	if (*dec->cur == ']')	1129	if (*dec->cur == ']')
715	++dec->cur;	1130	++dec->cur;
716	else	1131	else
717	for (;;)	1132	for (;;)
718	{	1133	{
…		…
722	if (!value)	1137	if (!value)
723	goto fail;	1138	goto fail;
724		1139
725	av_push (av, value);	1140	av_push (av, value);
726		1141
727	WS;	1142	decode_ws (dec);
728		1143
729	if (*dec->cur == ']')	1144	if (*dec->cur == ']')
730	{	1145	{
731	++dec->cur;	1146	++dec->cur;
732	break;	1147	break;
…		…
734		1149
735	if (*dec->cur != ',')	1150	if (*dec->cur != ',')
736	ERR (", or ] expected while parsing array");	1151	ERR (", or ] expected while parsing array");
737		1152
738	++dec->cur;	1153	++dec->cur;
		1154
		1155	decode_ws (dec);
		1156
		1157	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1158	{
		1159	++dec->cur;
		1160	break;
		1161	}
739	}	1162	}
740		1163
		1164	DEC_DEC_DEPTH;
741	return newRV_noinc ((SV *)av);	1165	return newRV_noinc ((SV *)av);
742		1166
743	fail:	1167	fail:
744	SvREFCNT_dec (av);	1168	SvREFCNT_dec (av);
		1169	DEC_DEC_DEPTH;
745	return 0;	1170	return 0;
746	}	1171	}
747		1172
748	static SV *	1173	static SV *
749	decode_hv (dec_t *dec)	1174	decode_hv (dec_t *dec)
750	{	1175	{
		1176	SV *sv;
751	HV *hv = newHV ();	1177	HV *hv = newHV ();
752		1178
753	WS;	1179	DEC_INC_DEPTH;
		1180	decode_ws (dec);
		1181
754	if (*dec->cur == '}')	1182	if (*dec->cur == '}')
755	++dec->cur;	1183	++dec->cur;
756	else	1184	else
757	for (;;)	1185	for (;;)
758	{	1186	{
759	SV *key, *value;
760
761	WS; EXPECT_CH ('"');	1187	EXPECT_CH ('"');
762		1188
763	key = decode_str (dec);	1189	// heuristic: assume that
764	if (!key)	1190	// a) decode_str + hv_store_ent are abysmally slow.
765	goto fail;	1191	// b) most hash keys are short, simple ascii text.
		1192	// => try to "fast-match" such strings to avoid
		1193	// the overhead of decode_str + hv_store_ent.
		1194	{
		1195	SV *value;
		1196	char *p = dec->cur;
		1197	char *e = p + 24; // only try up to 24 bytes
766		1198
767	WS; EXPECT_CH (':');	1199	for (;;)
768
769	value = decode_sv (dec);
770	if (!value)
771	{	1200	{
		1201	// the >= 0x80 is false on most architectures
		1202	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1203	{
		1204	// slow path, back up and use decode_str
		1205	SV *key = decode_str (dec);
		1206	if (!key)
		1207	goto fail;
		1208
		1209	decode_ws (dec); EXPECT_CH (':');
		1210
		1211	decode_ws (dec);
		1212	value = decode_sv (dec);
		1213	if (!value)
		1214	{
		1215	SvREFCNT_dec (key);
		1216	goto fail;
		1217	}
		1218
		1219	hv_store_ent (hv, key, value, 0);
772	SvREFCNT_dec (key);	1220	SvREFCNT_dec (key);
		1221
		1222	break;
		1223	}
		1224	else if (*p == '"')
		1225	{
		1226	// fast path, got a simple key
		1227	char *key = dec->cur;
		1228	int len = p - key;
		1229	dec->cur = p + 1;
		1230
		1231	decode_ws (dec); EXPECT_CH (':');
		1232
		1233	decode_ws (dec);
		1234	value = decode_sv (dec);
		1235	if (!value)
773	goto fail;	1236	goto fail;
		1237
		1238	hv_store (hv, key, len, value, 0);
		1239
		1240	break;
		1241	}
		1242
		1243	++p;
774	}	1244	}
775
776	//TODO: optimise
777	hv_store_ent (hv, key, value, 0);
778
779	WS;	1245	}
		1246
		1247	decode_ws (dec);
780		1248
781	if (*dec->cur == '}')	1249	if (*dec->cur == '}')
782	{	1250	{
783	++dec->cur;	1251	++dec->cur;
784	break;	1252	break;
…		…
786		1254
787	if (*dec->cur != ',')	1255	if (*dec->cur != ',')
788	ERR (", or } expected while parsing object/hash");	1256	ERR (", or } expected while parsing object/hash");
789		1257
790	++dec->cur;	1258	++dec->cur;
		1259
		1260	decode_ws (dec);
		1261
		1262	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1263	{
		1264	++dec->cur;
		1265	break;
		1266	}
791	}	1267	}
792		1268
		1269	DEC_DEC_DEPTH;
793	return newRV_noinc ((SV *)hv);	1270	sv = newRV_noinc ((SV *)hv);
		1271
		1272	// check filter callbacks
		1273	if (dec->json.flags & F_HOOK)
		1274	{
		1275	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1276	{
		1277	HE *cb, *he;
		1278
		1279	hv_iterinit (hv);
		1280	he = hv_iternext (hv);
		1281	hv_iterinit (hv);
		1282
		1283	// the next line creates a mortal sv each time its called.
		1284	// might want to optimise this for common cases.
		1285	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1286
		1287	if (cb)
		1288	{
		1289	dSP;
		1290	int count;
		1291
		1292	ENTER; SAVETMPS; PUSHMARK (SP);
		1293	XPUSHs (HeVAL (he));
		1294
		1295	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1296
		1297	if (count == 1)
		1298	{
		1299	sv = newSVsv (POPs);
		1300	FREETMPS; LEAVE;
		1301	return sv;
		1302	}
		1303
		1304	FREETMPS; LEAVE;
		1305	}
		1306	}
		1307
		1308	if (dec->json.cb_object)
		1309	{
		1310	dSP;
		1311	int count;
		1312
		1313	ENTER; SAVETMPS; PUSHMARK (SP);
		1314	XPUSHs (sv_2mortal (sv));
		1315
		1316	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1317
		1318	if (count == 1)
		1319	{
		1320	sv = newSVsv (POPs);
		1321	FREETMPS; LEAVE;
		1322	return sv;
		1323	}
		1324
		1325	SvREFCNT_inc (sv);
		1326	FREETMPS; LEAVE;
		1327	}
		1328	}
		1329
		1330	return sv;
794		1331
795	fail:	1332	fail:
796	SvREFCNT_dec (hv);	1333	SvREFCNT_dec (hv);
		1334	DEC_DEC_DEPTH;
797	return 0;	1335	return 0;
798	}	1336	}
799		1337
800	static SV *	1338	static SV *
801	decode_sv (dec_t *dec)	1339	decode_sv (dec_t *dec)
802	{	1340	{
803	WS;	1341	// the beauty of JSON: you need exactly one character lookahead
		1342	// to parse everything.
804	switch (*dec->cur)	1343	switch (*dec->cur)
805	{	1344	{
806	case '"': ++dec->cur; return decode_str (dec);	1345	case '"': ++dec->cur; return decode_str (dec);
807	case '[': ++dec->cur; return decode_av (dec);	1346	case '[': ++dec->cur; return decode_av (dec);
808	case '{': ++dec->cur; return decode_hv (dec);	1347	case '{': ++dec->cur; return decode_hv (dec);
809		1348
810	case '-':	1349	case '-':
811	case '0': case '1': case '2': case '3': case '4':	1350	case '0': case '1': case '2': case '3': case '4':
812	case '5': case '6': case '7': case '8': case '9':	1351	case '5': case '6': case '7': case '8': case '9':
813	return decode_num (dec);	1352	return decode_num (dec);
814		1353
815	case 't':	1354	case 't':
816	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1355	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
817	{	1356	{
818	dec->cur += 4;	1357	dec->cur += 4;
		1358	#if JSON_SLOW
		1359	json_true = get_bool ("JSON::XS::true");
		1360	#endif
819	return newSViv (1);	1361	return newSVsv (json_true);
820	}	1362	}
821	else	1363	else
822	ERR ("'true' expected");	1364	ERR ("'true' expected");
823		1365
824	break;	1366	break;
825		1367
826	case 'f':	1368	case 'f':
827	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1369	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
828	{	1370	{
829	dec->cur += 5;	1371	dec->cur += 5;
		1372	#if JSON_SLOW
		1373	json_false = get_bool ("JSON::XS::false");
		1374	#endif
830	return newSViv (0);	1375	return newSVsv (json_false);
831	}	1376	}
832	else	1377	else
833	ERR ("'false' expected");	1378	ERR ("'false' expected");
834		1379
835	break;	1380	break;
…		…
844	ERR ("'null' expected");	1389	ERR ("'null' expected");
845		1390
846	break;	1391	break;
847		1392
848	default:	1393	default:
849	ERR ("malformed json string, neither array, object, number, string or atom");	1394	ERR ("malformed JSON string, neither array, object, number, string or atom");
850	break;	1395	break;
851	}	1396	}
852		1397
853	fail:	1398	fail:
854	return 0;	1399	return 0;
855	}	1400	}
856		1401
857	static SV *	1402	static SV *
858	decode_json (SV *string, UV flags)	1403	decode_json (SV *string, JSON *json, STRLEN *offset_return)
859	{	1404	{
		1405	dec_t dec;
		1406	STRLEN offset;
860	SV *sv;	1407	SV *sv;
861		1408
		1409	SvGETMAGIC (string);
		1410	SvUPGRADE (string, SVt_PV);
		1411
		1412	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1413	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1414	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1415
862	if (flags & F_UTF8)	1416	if (json->flags & F_UTF8)
863	sv_utf8_downgrade (string, 0);	1417	sv_utf8_downgrade (string, 0);
864	else	1418	else
865	sv_utf8_upgrade (string);	1419	sv_utf8_upgrade (string);
866		1420
867	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1421	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
868		1422
869	dec_t dec;	1423	dec.json = *json;
870	dec.flags = flags;
871	dec.cur = SvPVX (string);	1424	dec.cur = SvPVX (string);
872	dec.end = SvEND (string);	1425	dec.end = SvEND (string);
873	dec.err = 0;	1426	dec.err = 0;
		1427	dec.depth = 0;
		1428	dec.maxdepth = DEC_DEPTH (dec.json.flags);
874		1429
		1430	if (dec.json.cb_object || dec.json.cb_sk_object)
		1431	dec.json.flags |= F_HOOK;
		1432
		1433	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1434
		1435	decode_ws (&dec);
875	sv = decode_sv (&dec);	1436	sv = decode_sv (&dec);
876		1437
		1438	if (!(offset_return || !sv))
		1439	{
		1440	// check for trailing garbage
		1441	decode_ws (&dec);
		1442
		1443	if (*dec.cur)
		1444	{
		1445	dec.err = "garbage after JSON object";
		1446	SvREFCNT_dec (sv);
		1447	sv = 0;
		1448	}
		1449	}
		1450
		1451	if (offset_return || !sv)
		1452	{
		1453	offset = dec.json.flags & F_UTF8
		1454	? dec.cur - SvPVX (string)
		1455	: utf8_distance (dec.cur, SvPVX (string));
		1456
		1457	if (offset_return)
		1458	*offset_return = offset;
		1459	}
		1460
877	if (!sv)	1461	if (!sv)
878	{	1462	{
879	IV offset = dec.flags & F_UTF8
880	? dec.cur - SvPVX (string)
881	: utf8_distance (dec.cur, SvPVX (string));
882	SV *uni = sv_newmortal ();	1463	SV *uni = sv_newmortal ();
		1464
883	// horrible hack to silence warning inside pv_uni_display	1465	// horrible hack to silence warning inside pv_uni_display
884	COP cop;	1466	COP cop = *PL_curcop;
885	memset (&cop, 0, sizeof (cop));
886	cop.cop_warnings = pWARN_NONE;	1467	cop.cop_warnings = pWARN_NONE;
		1468	ENTER;
887	SAVEVPTR (PL_curcop);	1469	SAVEVPTR (PL_curcop);
888	PL_curcop = &cop;	1470	PL_curcop = &cop;
889
890	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1471	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
		1472	LEAVE;
		1473
891	croak ("%s, at character offset %d (%s)",	1474	croak ("%s, at character offset %d [\"%s\"]",
892	dec.err,	1475	dec.err,
893	(int)offset,	1476	(int)offset,
894	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1477	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
895	}	1478	}
896		1479
897	sv = sv_2mortal (sv);	1480	sv = sv_2mortal (sv);
898		1481
899	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1482	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
900	croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)");	1483	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
901		1484
902	return sv;	1485	return sv;
903	}	1486	}
904		1487
		1488	/////////////////////////////////////////////////////////////////////////////
		1489	// incremental parser
		1490
		1491	static void
		1492	incr_parse (JSON *self)
		1493	{
		1494	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1495
		1496	for (;;)
		1497	{
		1498	//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
		1499	switch (self->incr_mode)
		1500	{
		1501	// only used for intiial whitespace skipping
		1502	case INCR_M_WS:
		1503	for (;;)
		1504	{
		1505	if (*p > 0x20)
		1506	{
		1507	self->incr_mode = INCR_M_JSON;
		1508	goto incr_m_json;
		1509	}
		1510	else if (!*p)
		1511	goto interrupt;
		1512
		1513	++p;
		1514	}
		1515
		1516	// skip a single char inside a string (for \\-processing)
		1517	case INCR_M_BS:
		1518	if (!*p)
		1519	goto interrupt;
		1520
		1521	++p;
		1522	self->incr_mode = INCR_M_STR;
		1523	goto incr_m_str;
		1524
		1525	// inside a string
		1526	case INCR_M_STR:
		1527	incr_m_str:
		1528	for (;;)
		1529	{
		1530	if (*p == '"')
		1531	{
		1532	++p;
		1533	self->incr_mode = INCR_M_JSON;
		1534
		1535	if (!self->incr_nest)
		1536	goto interrupt;
		1537
		1538	goto incr_m_json;
		1539	}
		1540	else if (*p == '\\')
		1541	{
		1542	++p; // "virtually" consumes character after \
		1543
		1544	if (!*p) // if at end of string we have to switch modes
		1545	{
		1546	self->incr_mode = INCR_M_BS;
		1547	goto interrupt;
		1548	}
		1549	}
		1550	else if (!*p)
		1551	goto interrupt;
		1552
		1553	++p;
		1554	}
		1555
		1556	// after initial ws, outside string
		1557	case INCR_M_JSON:
		1558	incr_m_json:
		1559	for (;;)
		1560	{
		1561	switch (*p++)
		1562	{
		1563	case 0:
		1564	--p;
		1565	goto interrupt;
		1566
		1567	case 0x09:
		1568	case 0x0a:
		1569	case 0x0d:
		1570	case 0x20:
		1571	if (!self->incr_nest)
		1572	{
		1573	--p; // do not eat the whitespace, let the next round do it
		1574	goto interrupt;
		1575	}
		1576	break;
		1577
		1578	case '"':
		1579	self->incr_mode = INCR_M_STR;
		1580	goto incr_m_str;
		1581
		1582	case '[':
		1583	case '{':
		1584	++self->incr_nest;
		1585	break;
		1586
		1587	case ']':
		1588	case '}':
		1589	if (!--self->incr_nest)
		1590	goto interrupt;
		1591	}
		1592	}
		1593	}
		1594
		1595	modechange:
		1596	;
		1597	}
		1598
		1599	interrupt:
		1600	self->incr_pos = p - SvPVX (self->incr_text);
		1601	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
		1602	}
		1603
		1604	/////////////////////////////////////////////////////////////////////////////
		1605	// XS interface functions
		1606
905	MODULE = JSON::XS PACKAGE = JSON::XS	1607	MODULE = JSON::XS PACKAGE = JSON::XS
906		1608
907	BOOT:	1609	BOOT:
908	{	1610	{
909	int i;	1611	int i;
910		1612
911	memset (decode_hexdigit, 0xff, 256);
912	for (i = 10; i--; )	1613	for (i = 0; i < 256; ++i)
913	decode_hexdigit ['0' + i] = i;	1614	decode_hexdigit [i] =
		1615	i >= '0' && i <= '9' ? i - '0'
		1616	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1617	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1618	: -1;
914		1619
915	for (i = 7; i--; )
916	{
917	decode_hexdigit ['a' + i] = 10 + i;
918	decode_hexdigit ['A' + i] = 10 + i;
919	}
920
921	json_stash = gv_stashpv ("JSON::XS", 1);	1620	json_stash = gv_stashpv ("JSON::XS" , 1);
		1621	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1622
		1623	json_true = get_bool ("JSON::XS::true");
		1624	json_false = get_bool ("JSON::XS::false");
922	}	1625	}
923		1626
924	PROTOTYPES: DISABLE	1627	PROTOTYPES: DISABLE
925		1628
926	SV *new (char *dummy)	1629	void CLONE (...)
927	CODE:	1630	CODE:
928	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1631	json_stash = 0;
		1632	json_boolean_stash = 0;
		1633
		1634	void new (char *klass)
		1635	PPCODE:
		1636	{
		1637	SV *pv = NEWSV (0, sizeof (JSON));
		1638	SvPOK_only (pv);
		1639	Zero (SvPVX (pv), 1, JSON);
		1640	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1641	XPUSHs (sv_2mortal (sv_bless (
		1642	newRV_noinc (pv),
		1643	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1644	)));
		1645	}
		1646
		1647	void ascii (JSON *self, int enable = 1)
		1648	ALIAS:
		1649	ascii = F_ASCII
		1650	latin1 = F_LATIN1
		1651	utf8 = F_UTF8
		1652	indent = F_INDENT
		1653	canonical = F_CANONICAL
		1654	space_before = F_SPACE_BEFORE
		1655	space_after = F_SPACE_AFTER
		1656	pretty = F_PRETTY
		1657	allow_nonref = F_ALLOW_NONREF
		1658	shrink = F_SHRINK
		1659	allow_blessed = F_ALLOW_BLESSED
		1660	convert_blessed = F_CONV_BLESSED
		1661	relaxed = F_RELAXED
		1662	PPCODE:
		1663	{
		1664	if (enable)
		1665	self->flags |= ix;
		1666	else
		1667	self->flags &= ~ix;
		1668
		1669	XPUSHs (ST (0));
		1670	}
		1671
		1672	void get_ascii (JSON *self)
		1673	ALIAS:
		1674	get_ascii = F_ASCII
		1675	get_latin1 = F_LATIN1
		1676	get_utf8 = F_UTF8
		1677	get_indent = F_INDENT
		1678	get_canonical = F_CANONICAL
		1679	get_space_before = F_SPACE_BEFORE
		1680	get_space_after = F_SPACE_AFTER
		1681	get_allow_nonref = F_ALLOW_NONREF
		1682	get_shrink = F_SHRINK
		1683	get_allow_blessed = F_ALLOW_BLESSED
		1684	get_convert_blessed = F_CONV_BLESSED
		1685	get_relaxed = F_RELAXED
		1686	PPCODE:
		1687	XPUSHs (boolSV (self->flags & ix));
		1688
		1689	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1690	PPCODE:
		1691	{
		1692	UV log2 = 0;
		1693
		1694	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1695
		1696	while ((1UL << log2) < max_depth)
		1697	++log2;
		1698
		1699	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1700
		1701	XPUSHs (ST (0));
		1702	}
		1703
		1704	U32 get_max_depth (JSON *self)
		1705	CODE:
		1706	RETVAL = DEC_DEPTH (self->flags);
929	OUTPUT:	1707	OUTPUT:
930	RETVAL	1708	RETVAL
931		1709
932	SV *ascii (SV *self, int enable = 1)	1710	void max_size (JSON *self, UV max_size = 0)
933	ALIAS:	1711	PPCODE:
934	ascii = F_ASCII	1712	{
935	utf8 = F_UTF8	1713	UV log2 = 0;
936	indent = F_INDENT	1714
937	canonical = F_CANONICAL	1715	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
938	space_before = F_SPACE_BEFORE	1716	if (max_size == 1) max_size = 2;
939	space_after = F_SPACE_AFTER	1717
940	json_rpc = F_JSON_RPC	1718	while ((1UL << log2) < max_size)
941	pretty = F_PRETTY	1719	++log2;
942	allow_nonref = F_ALLOW_NONREF	1720
943	shrink = F_SHRINK	1721	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1722
		1723	XPUSHs (ST (0));
		1724	}
		1725
		1726	int get_max_size (JSON *self)
944	CODE:	1727	CODE:
945	{	1728	RETVAL = DEC_SIZE (self->flags);
946	UV *uv = SvJSON (self);
947	if (enable)
948	*uv |= ix;
949	else
950	*uv &= ~ix;
951
952	RETVAL = newSVsv (self);
953	}
954	OUTPUT:	1729	OUTPUT:
955	RETVAL	1730	RETVAL
956		1731
957	void encode (SV *self, SV *scalar)	1732	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
958	PPCODE:	1733	PPCODE:
959	XPUSHs (encode_json (scalar, *SvJSON (self)));	1734	{
		1735	SvREFCNT_dec (self->cb_object);
		1736	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
960		1737
961	void decode (SV *self, SV *jsonstr)	1738	XPUSHs (ST (0));
		1739	}
		1740
		1741	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
962	PPCODE:	1742	PPCODE:
		1743	{
		1744	if (!self->cb_sk_object)
		1745	self->cb_sk_object = newHV ();
		1746
		1747	if (SvOK (cb))
		1748	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1749	else
		1750	{
		1751	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1752
		1753	if (!HvKEYS (self->cb_sk_object))
		1754	{
		1755	SvREFCNT_dec (self->cb_sk_object);
		1756	self->cb_sk_object = 0;
		1757	}
		1758	}
		1759
		1760	XPUSHs (ST (0));
		1761	}
		1762
		1763	void encode (JSON *self, SV *scalar)
		1764	PPCODE:
		1765	XPUSHs (encode_json (scalar, self));
		1766
		1767	void decode (JSON *self, SV *jsonstr)
		1768	PPCODE:
963	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1769	XPUSHs (decode_json (jsonstr, self, 0));
		1770
		1771	void decode_prefix (JSON *self, SV *jsonstr)
		1772	PPCODE:
		1773	{
		1774	STRLEN offset;
		1775	EXTEND (SP, 2);
		1776	PUSHs (decode_json (jsonstr, self, &offset));
		1777	PUSHs (sv_2mortal (newSVuv (offset)));
		1778	}
		1779
		1780	void incr_parse (JSON *self, SV *jsonstr = 0)
		1781	PPCODE:
		1782	{
		1783	if (!self->incr_text)
		1784	self->incr_text = newSVpvn ("", 0);
		1785
		1786	// append data, if any
		1787	if (jsonstr)
		1788	{
		1789	if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text))
		1790	{
		1791	/* utf-8-ness differs, need to upgrade */
		1792	sv_utf8_upgrade (self->incr_text);
		1793
		1794	if (self->incr_pos)
		1795	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1796	- (U8 *)SvPVX (self->incr_text);
		1797	}
		1798
		1799	{
		1800	STRLEN len;
		1801	const char *str = SvPV (jsonstr, len);
		1802	SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1);
		1803	Move (str, SvEND (self->incr_text), len, char);
		1804	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1805	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1806	}
		1807	}
		1808
		1809	if (GIMME_V != G_VOID)
		1810	do
		1811	{
		1812	STRLEN offset;
		1813
		1814	if (!INCR_DONE (self))
		1815	{
		1816	incr_parse (self);
		1817	if (!INCR_DONE (self))
		1818	break;
		1819	}
		1820
		1821	XPUSHs (decode_json (self->incr_text, self, &offset));
		1822
		1823	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
		1824	self->incr_pos -= offset;
		1825	self->incr_nest = 0;
		1826	self->incr_mode = 0;
		1827	}
		1828	while (GIMME_V == G_ARRAY);
		1829	}
		1830
		1831	SV *incr_text (JSON *self)
		1832	ATTRS: lvalue
		1833	CODE:
		1834	{
		1835	if (self->incr_pos)
		1836	croak ("incr_text can not be called when the incremental parser already started parsing");
		1837
		1838	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		1839	}
		1840	OUTPUT:
		1841	RETVAL
		1842
		1843	void incr_skip (JSON *self)
		1844	CODE:
		1845	{
		1846	if (self->incr_pos)
		1847	{
		1848	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1849	self->incr_pos = 0;
		1850	self->incr_nest = 0;
		1851	self->incr_mode = 0;
		1852	}
		1853	}
		1854
		1855	void DESTROY (JSON *self)
		1856	CODE:
		1857	SvREFCNT_dec (self->cb_sk_object);
		1858	SvREFCNT_dec (self->cb_object);
		1859	SvREFCNT_dec (self->incr_text);
964		1860
965	PROTOTYPES: ENABLE	1861	PROTOTYPES: ENABLE
966		1862
967	void to_json (SV *scalar)	1863	void encode_json (SV *scalar)
		1864	ALIAS:
		1865	to_json_ = 0
		1866	encode_json = F_UTF8
968	PPCODE:	1867	PPCODE:
		1868	{
		1869	JSON json = { F_DEFAULT | ix };
969	XPUSHs (encode_json (scalar, F_UTF8));	1870	XPUSHs (encode_json (scalar, &json));
		1871	}
970		1872
971	void from_json (SV *jsonstr)	1873	void decode_json (SV *jsonstr)
		1874	ALIAS:
		1875	from_json_ = 0
		1876	decode_json = F_UTF8
972	PPCODE:	1877	PPCODE:
		1878	{
		1879	JSON json = { F_DEFAULT | ix };
973	XPUSHs (decode_json (jsonstr, F_UTF8));	1880	XPUSHs (decode_json (jsonstr, &json, 0));
		1881	}
974		1882
		1883

Diff Legend

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