ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.14 by root, Sat Mar 24 22:55:43 2007 UTC vs.
Revision 1.80 by root, Wed Mar 26 01:32:53 2008 UTC

1	#include "EXTERN.h"	1	#include "EXTERN.h"
2	#include "perl.h"	2	#include "perl.h"
3	#include "XSUB.h"	3	#include "XSUB.h"
4		4
5	#include "assert.h"	5	#include <assert.h>
6	#include "string.h"	6	#include <string.h>
7	#include "stdlib.h"	7	#include <stdlib.h>
		8	#include <stdio.h>
		9	#include <limits.h>
		10	#include <float.h>
8		11
		12	#if defined(__BORLANDC__) || defined(_MSC_VER)
		13	# define snprintf _snprintf // C compilers have this in stdio.h
		14	#endif
		15
		16	// some old perls do not have this, try to make it work, no
		17	// guarentees, though. if it breaks, you get to keep the pieces.
		18	#ifndef UTF8_MAXBYTES
		19	# define UTF8_MAXBYTES 13
		20	#endif
		21
		22	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2)
		23
9	#define F_ASCII 0x00000001	24	#define F_ASCII 0x00000001UL
		25	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	26	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	27	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	28	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	29	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	30	#define F_SPACE_AFTER 0x00000040UL
15	#define F_ALLOW_NONREF 0x00000080	31	#define F_ALLOW_NONREF 0x00000100UL
16	#define F_SHRINK 0x00000100	32	#define F_SHRINK 0x00000200UL
		33	#define F_ALLOW_BLESSED 0x00000400UL
		34	#define F_CONV_BLESSED 0x00000800UL
		35	#define F_RELAXED 0x00001000UL
		36
		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 ))
17		45
18	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	46	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
19	#define F_DEFAULT 0	47	#define F_DEFAULT (9UL << S_MAXDEPTH)
20		48
21	#define INIT_SIZE 32 // initial scalar size to be allocated	49	#define INIT_SIZE 32 // initial scalar size to be allocated
22	#define INDENT_STEP 3 // spaces per indentation level	50	#define INDENT_STEP 3 // spaces per indentation level
23		51
24	#define UTF8_MAX_LEN 11 // for perls UTF-X: max. number of octets per character
25	#define SHORT_STRING_LEN 512 // special-case strings of up to this size	52	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
26		53
27	#define SB do {	54	#define SB do {
28	#define SE } while (0)	55	#define SE } while (0)
29		56
		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
30	static HV *json_stash; // JSON::XS::	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;
31		103
32	/////////////////////////////////////////////////////////////////////////////	104	/////////////////////////////////////////////////////////////////////////////
33	// utility functions	105	// utility functions
34		106
35	static UV *	107	INLINE void
36	SvJSON (SV *sv)
37	{
38	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
39	croak ("object is not of type JSON::XS");
40
41	return &SvUVX (SvRV (sv));
42	}
43
44	static void
45	shrink (SV *sv)	108	shrink (SV *sv)
46	{	109	{
47	sv_utf8_downgrade (sv, 1);	110	sv_utf8_downgrade (sv, 1);
48	if (SvLEN (sv) > SvCUR (sv) + 1)	111	if (SvLEN (sv) > SvCUR (sv) + 1)
49	{	112	{
…		…
58	// decode an utf-8 character and return it, or (UV)-1 in	121	// decode an utf-8 character and return it, or (UV)-1 in
59	// case of an error.	122	// case of an error.
60	// we special-case "safe" characters from U+80 .. U+7FF,	123	// we special-case "safe" characters from U+80 .. U+7FF,
61	// but use the very good perl function to parse anything else.	124	// but use the very good perl function to parse anything else.
62	// note that we never call this function for a ascii codepoints	125	// note that we never call this function for a ascii codepoints
63	static UV	126	INLINE UV
64	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)	127	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
65	{	128	{
66	if (s[0] > 0xdf || s[0] < 0xc2)	129	if (expect_true (len >= 2
67	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	130	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
68	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	131	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
69	{	132	{
70	*clen = 2;	133	*clen = 2;
71	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);	134	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
72	}	135	}
73	else	136	else
74	return (UV)-1;	137	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		138	}
		139
		140	// likewise for encoding, also never called for ascii codepoints
		141	// this function takes advantage of this fact, although current gccs
		142	// seem to optimise the check for >= 0x80 away anyways
		143	INLINE unsigned char *
		144	encode_utf8 (unsigned char *s, UV ch)
		145	{
		146	if (expect_false (ch < 0x000080))
		147	*s++ = ch;
		148	else if (expect_true (ch < 0x000800))
		149	*s++ = 0xc0 | ( ch >> 6),
		150	*s++ = 0x80 | ( ch & 0x3f);
		151	else if ( ch < 0x010000)
		152	*s++ = 0xe0 | ( ch >> 12),
		153	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		154	*s++ = 0x80 | ( ch & 0x3f);
		155	else if ( ch < 0x110000)
		156	*s++ = 0xf0 | ( ch >> 18),
		157	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		158	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		159	*s++ = 0x80 | ( ch & 0x3f);
		160
		161	return s;
75	}	162	}
76		163
77	/////////////////////////////////////////////////////////////////////////////	164	/////////////////////////////////////////////////////////////////////////////
78	// encoder	165	// encoder
79		166
…		…
81	typedef struct	168	typedef struct
82	{	169	{
83	char *cur; // SvPVX (sv) + current output position	170	char *cur; // SvPVX (sv) + current output position
84	char *end; // SvEND (sv)	171	char *end; // SvEND (sv)
85	SV *sv; // result scalar	172	SV *sv; // result scalar
86	UV flags; // F_*	173	JSON json;
87	int indent; // indentation level	174	U32 indent; // indentation level
88	int max_depth; // max. recursion level	175	U32 maxdepth; // max. indentation/recursion level
		176	UV limit; // escape character values >= this value when encoding
89	} enc_t;	177	} enc_t;
90		178
91	static void	179	INLINE void
92	need (enc_t *enc, STRLEN len)	180	need (enc_t *enc, STRLEN len)
93	{	181	{
94	if (enc->cur + len >= enc->end)	182	if (expect_false (enc->cur + len >= enc->end))
95	{	183	{
96	STRLEN cur = enc->cur - SvPVX (enc->sv);	184	STRLEN cur = enc->cur - SvPVX (enc->sv);
97	SvGROW (enc->sv, cur + len + 1);	185	SvGROW (enc->sv, cur + len + 1);
98	enc->cur = SvPVX (enc->sv) + cur;	186	enc->cur = SvPVX (enc->sv) + cur;
99	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	187	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
100	}	188	}
101	}	189	}
102		190
103	static void	191	INLINE void
104	encode_ch (enc_t *enc, char ch)	192	encode_ch (enc_t *enc, char ch)
105	{	193	{
106	need (enc, 1);	194	need (enc, 1);
107	*enc->cur++ = ch;	195	*enc->cur++ = ch;
108	}	196	}
…		…
116		204
117	while (str < end)	205	while (str < end)
118	{	206	{
119	unsigned char ch = *(unsigned char *)str;	207	unsigned char ch = *(unsigned char *)str;
120		208
121	if (ch >= 0x20 && ch < 0x80) // most common case	209	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
122	{	210	{
123	if (ch == '"') // but with slow exceptions	211	if (expect_false (ch == '"')) // but with slow exceptions
124	{	212	{
125	need (enc, len += 1);	213	need (enc, len += 1);
126	*enc->cur++ = '\\';	214	*enc->cur++ = '\\';
127	*enc->cur++ = '"';	215	*enc->cur++ = '"';
128	}	216	}
129	else if (ch == '\\')	217	else if (expect_false (ch == '\\'))
130	{	218	{
131	need (enc, len += 1);	219	need (enc, len += 1);
132	*enc->cur++ = '\\';	220	*enc->cur++ = '\\';
133	*enc->cur++ = '\\';	221	*enc->cur++ = '\\';
134	}	222	}
…		…
152	STRLEN clen;	240	STRLEN clen;
153	UV uch;	241	UV uch;
154		242
155	if (is_utf8)	243	if (is_utf8)
156	{	244	{
157	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
158	uch = decode_utf8 (str, end - str, &clen);	245	uch = decode_utf8 (str, end - str, &clen);
159	if (clen == (STRLEN)-1)	246	if (clen == (STRLEN)-1)
160	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	247	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
161	}	248	}
162	else	249	else
163	{	250	{
164	uch = ch;	251	uch = ch;
165	clen = 1;	252	clen = 1;
166	}	253	}
167		254
168	if (uch > 0x10FFFFUL)	255	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
169	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
170
171	if (uch < 0x80 || enc->flags & F_ASCII)
172	{	256	{
173	if (uch > 0xFFFFUL)	257	if (uch >= 0x10000UL)
174	{	258	{
		259	if (uch >= 0x110000UL)
		260	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		261
175	need (enc, len += 11);	262	need (enc, len += 11);
176	sprintf (enc->cur, "\\u%04x\\u%04x",	263	sprintf (enc->cur, "\\u%04x\\u%04x",
177	(int)((uch - 0x10000) / 0x400 + 0xD800),	264	(int)((uch - 0x10000) / 0x400 + 0xD800),
178	(int)((uch - 0x10000) % 0x400 + 0xDC00));	265	(int)((uch - 0x10000) % 0x400 + 0xDC00));
179	enc->cur += 12;	266	enc->cur += 12;
…		…
190	*enc->cur++ = hexdigit [(uch >> 0) & 15];	277	*enc->cur++ = hexdigit [(uch >> 0) & 15];
191	}	278	}
192		279
193	str += clen;	280	str += clen;
194	}	281	}
		282	else if (enc->json.flags & F_LATIN1)
		283	{
		284	*enc->cur++ = uch;
		285	str += clen;
		286	}
195	else if (is_utf8)	287	else if (is_utf8)
196	{	288	{
197	need (enc, len += clen);	289	need (enc, len += clen);
198	do	290	do
199	{	291	{
…		…
201	}	293	}
202	while (--clen);	294	while (--clen);
203	}	295	}
204	else	296	else
205	{	297	{
206	need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed	298	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
207	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	299	enc->cur = encode_utf8 (enc->cur, uch);
208	++str;	300	++str;
209	}	301	}
210	}	302	}
211	}	303	}
212	}	304	}
213		305
214	--len;	306	--len;
215	}	307	}
216	}	308	}
217		309
218	static void	310	INLINE void
219	encode_indent (enc_t *enc)	311	encode_indent (enc_t *enc)
220	{	312	{
221	if (enc->flags & F_INDENT)	313	if (enc->json.flags & F_INDENT)
222	{	314	{
223	int spaces = enc->indent * INDENT_STEP;	315	int spaces = enc->indent * INDENT_STEP;
224		316
225	need (enc, spaces);	317	need (enc, spaces);
226	memset (enc->cur, ' ', spaces);	318	memset (enc->cur, ' ', spaces);
227	enc->cur += spaces;	319	enc->cur += spaces;
228	}	320	}
229	}	321	}
230		322
231	static void	323	INLINE void
232	encode_space (enc_t *enc)	324	encode_space (enc_t *enc)
233	{	325	{
234	need (enc, 1);	326	need (enc, 1);
235	encode_ch (enc, ' ');	327	encode_ch (enc, ' ');
236	}	328	}
237		329
238	static void	330	INLINE void
239	encode_nl (enc_t *enc)	331	encode_nl (enc_t *enc)
240	{	332	{
241	if (enc->flags & F_INDENT)	333	if (enc->json.flags & F_INDENT)
242	{	334	{
243	need (enc, 1);	335	need (enc, 1);
244	encode_ch (enc, '\n');	336	encode_ch (enc, '\n');
245	}	337	}
246	}	338	}
247		339
248	static void	340	INLINE void
249	encode_comma (enc_t *enc)	341	encode_comma (enc_t *enc)
250	{	342	{
251	encode_ch (enc, ',');	343	encode_ch (enc, ',');
252		344
253	if (enc->flags & F_INDENT)	345	if (enc->json.flags & F_INDENT)
254	encode_nl (enc);	346	encode_nl (enc);
255	else if (enc->flags & F_SPACE_AFTER)	347	else if (enc->json.flags & F_SPACE_AFTER)
256	encode_space (enc);	348	encode_space (enc);
257	}	349	}
258		350
259	static void encode_sv (enc_t *enc, SV *sv);	351	static void encode_sv (enc_t *enc, SV *sv);
260		352
261	static void	353	static void
262	encode_av (enc_t *enc, AV *av)	354	encode_av (enc_t *enc, AV *av)
263	{	355	{
264	int i, len = av_len (av);	356	int i, len = av_len (av);
265		357
266	encode_ch (enc, '['); encode_nl (enc);	358	if (enc->indent >= enc->maxdepth)
267	++enc->indent;	359	croak ("data structure too deep (hit recursion limit)");
268		360
		361	encode_ch (enc, '[');
		362
		363	if (len >= 0)
		364	{
		365	encode_nl (enc); ++enc->indent;
		366
269	for (i = 0; i <= len; ++i)	367	for (i = 0; i <= len; ++i)
270	{	368	{
		369	SV **svp = av_fetch (av, i, 0);
		370
271	encode_indent (enc);	371	encode_indent (enc);
272	encode_sv (enc, *av_fetch (av, i, 0));
273		372
		373	if (svp)
		374	encode_sv (enc, *svp);
		375	else
		376	encode_str (enc, "null", 4, 0);
		377
274	if (i < len)	378	if (i < len)
275	encode_comma (enc);	379	encode_comma (enc);
276	}	380	}
277		381
		382	encode_nl (enc); --enc->indent; encode_indent (enc);
		383	}
		384
278	encode_nl (enc);	385	encode_ch (enc, ']');
279
280	--enc->indent;
281	encode_indent (enc); encode_ch (enc, ']');
282	}	386	}
283		387
284	static void	388	static void
285	encode_he (enc_t *enc, HE *he)	389	encode_hk (enc_t *enc, HE *he)
286	{	390	{
287	encode_ch (enc, '"');	391	encode_ch (enc, '"');
288		392
289	if (HeKLEN (he) == HEf_SVKEY)	393	if (HeKLEN (he) == HEf_SVKEY)
290	{	394	{
…		…
300	else	404	else
301	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	405	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
302		406
303	encode_ch (enc, '"');	407	encode_ch (enc, '"');
304		408
305	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	409	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
306	encode_ch (enc, ':');	410	encode_ch (enc, ':');
307	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	411	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
308	encode_sv (enc, HeVAL (he));
309	}	412	}
310		413
311	// compare hash entries, used when all keys are bytestrings	414	// compare hash entries, used when all keys are bytestrings
312	static int	415	static int
313	he_cmp_fast (const void *a_, const void *b_)	416	he_cmp_fast (const void *a_, const void *b_)
…		…
318	HE *b = *(HE **)b_;	421	HE *b = *(HE **)b_;
319		422
320	STRLEN la = HeKLEN (a);	423	STRLEN la = HeKLEN (a);
321	STRLEN lb = HeKLEN (b);	424	STRLEN lb = HeKLEN (b);
322		425
323	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	426	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
324	cmp = la - lb;	427	cmp = lb - la;
325		428
326	return cmp;	429	return cmp;
327	}	430	}
328		431
329	// compare hash entries, used when some keys are sv's or utf-x	432	// compare hash entries, used when some keys are sv's or utf-x
330	static int	433	static int
331	he_cmp_slow (const void *a, const void *b)	434	he_cmp_slow (const void *a, const void *b)
332	{	435	{
333	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	436	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
334	}	437	}
335		438
336	static void	439	static void
337	encode_hv (enc_t *enc, HV *hv)	440	encode_hv (enc_t *enc, HV *hv)
338	{	441	{
339	int count, i;	442	HE *he;
340		443
341	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	444	if (enc->indent >= enc->maxdepth)
		445	croak ("data structure too deep (hit recursion limit)");
342		446
343	if ((count = hv_iterinit (hv)))	447	encode_ch (enc, '{');
344	{	448
345	// for canonical output we have to sort by keys first	449	// for canonical output we have to sort by keys first
346	// actually, this is mostly due to the stupid so-called	450	// actually, this is mostly due to the stupid so-called
347	// security workaround added somewhere in 5.8.x.	451	// security workaround added somewhere in 5.8.x.
348	// that randomises hash orderings	452	// that randomises hash orderings
349	if (enc->flags & F_CANONICAL)	453	if (enc->json.flags & F_CANONICAL)
		454	{
		455	int count = hv_iterinit (hv);
		456
		457	if (SvMAGICAL (hv))
350	{	458	{
		459	// need to count by iterating. could improve by dynamically building the vector below
		460	// but I don't care for the speed of this special case.
		461	// note also that we will run into undefined behaviour when the two iterations
		462	// do not result in the same count, something I might care for in some later release.
		463
		464	count = 0;
		465	while (hv_iternext (hv))
		466	++count;
		467
		468	hv_iterinit (hv);
		469	}
		470
		471	if (count)
		472	{
		473	int i, fast = 1;
		474	#if defined(__BORLANDC__) || defined(_MSC_VER)
		475	HE **hes = _alloca (count * sizeof (HE));
		476	#else
351	HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode	477	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
352	int fast = 1;	478	#endif
353		479
354	i = 0;	480	i = 0;
355	while ((he = hv_iternext (hv)))	481	while ((he = hv_iternext (hv)))
356	{	482	{
357	hes [i++] = he;	483	hes [i++] = he;
…		…
379		505
380	FREETMPS;	506	FREETMPS;
381	LEAVE;	507	LEAVE;
382	}	508	}
383		509
384	for (i = 0; i < count; ++i)	510	encode_nl (enc); ++enc->indent;
		511
		512	while (count--)
385	{	513	{
386	encode_indent (enc);	514	encode_indent (enc);
		515	he = hes [count];
387	encode_he (enc, hes [i]);	516	encode_hk (enc, he);
		517	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
388		518
389	if (i < count - 1)	519	if (count)
390	encode_comma (enc);	520	encode_comma (enc);
391	}	521	}
392		522
		523	encode_nl (enc); --enc->indent; encode_indent (enc);
		524	}
		525	}
		526	else
		527	{
		528	if (hv_iterinit (hv) || SvMAGICAL (hv))
		529	if ((he = hv_iternext (hv)))
		530	{
		531	encode_nl (enc); ++enc->indent;
		532
		533	for (;;)
		534	{
393	encode_nl (enc);	535	encode_indent (enc);
		536	encode_hk (enc, he);
		537	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		538
		539	if (!(he = hv_iternext (hv)))
		540	break;
		541
		542	encode_comma (enc);
		543	}
		544
		545	encode_nl (enc); --enc->indent; encode_indent (enc);
		546	}
		547	}
		548
		549	encode_ch (enc, '}');
		550	}
		551
		552	// encode objects, arrays and special \0=false and \1=true values.
		553	static void
		554	encode_rv (enc_t *enc, SV *sv)
		555	{
		556	svtype svt;
		557
		558	SvGETMAGIC (sv);
		559	svt = SvTYPE (sv);
		560
		561	if (expect_false (SvOBJECT (sv)))
		562	{
		563	HV *stash = !JSON_SLOW || json_boolean_stash
		564	? json_boolean_stash
		565	: gv_stashpv ("JSON::XS::Boolean", 1);
		566
		567	if (SvSTASH (sv) == stash)
		568	{
		569	if (SvIV (sv))
		570	encode_str (enc, "true", 4, 0);
		571	else
		572	encode_str (enc, "false", 5, 0);
394	}	573	}
395	else	574	else
396	{	575	{
397	SV *sv;	576	#if 0
398	HE *he = hv_iternext (hv);	577	if (0 && sv_derived_from (rv, "JSON::Literal"))
399
400	for (;;)
401	{	578	{
402	encode_indent (enc);	579	// not yet
403	encode_he (enc, he);
404
405	if (!(he = hv_iternext (hv)))
406	break;
407
408	encode_comma (enc);
409	}	580	}
		581	#endif
		582	if (enc->json.flags & F_CONV_BLESSED)
		583	{
		584	// we re-bless the reference to get overload and other niceties right
		585	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
410		586
		587	if (to_json)
		588	{
		589	dSP;
		590
		591	ENTER; SAVETMPS; PUSHMARK (SP);
		592	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		593
		594	// calling with G_SCALAR ensures that we always get a 1 return value
		595	PUTBACK;
		596	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		597	SPAGAIN;
		598
		599	// catch this surprisingly common error
		600	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		601	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		602
		603	sv = POPs;
		604	PUTBACK;
		605
411	encode_nl (enc);	606	encode_sv (enc, sv);
		607
		608	FREETMPS; LEAVE;
		609	}
		610	else if (enc->json.flags & F_ALLOW_BLESSED)
		611	encode_str (enc, "null", 4, 0);
		612	else
		613	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		614	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		615	}
		616	else if (enc->json.flags & F_ALLOW_BLESSED)
		617	encode_str (enc, "null", 4, 0);
		618	else
		619	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		620	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
412	}	621	}
413	}	622	}
		623	else if (svt == SVt_PVHV)
		624	encode_hv (enc, (HV *)sv);
		625	else if (svt == SVt_PVAV)
		626	encode_av (enc, (AV *)sv);
		627	else if (svt < SVt_PVAV)
		628	{
		629	STRLEN len = 0;
		630	char *pv = svt ? SvPV (sv, len) : 0;
414		631
415	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	632	if (len == 1 && *pv == '1')
		633	encode_str (enc, "true", 4, 0);
		634	else if (len == 1 && *pv == '0')
		635	encode_str (enc, "false", 5, 0);
		636	else
		637	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		638	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		639	}
		640	else
		641	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		642	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
416	}	643	}
417		644
418	static void	645	static void
419	encode_sv (enc_t *enc, SV *sv)	646	encode_sv (enc_t *enc, SV *sv)
420	{	647	{
…		…
428	encode_str (enc, str, len, SvUTF8 (sv));	655	encode_str (enc, str, len, SvUTF8 (sv));
429	encode_ch (enc, '"');	656	encode_ch (enc, '"');
430	}	657	}
431	else if (SvNOKp (sv))	658	else if (SvNOKp (sv))
432	{	659	{
		660	// trust that perl will do the right thing w.r.t. JSON syntax.
433	need (enc, NV_DIG + 32);	661	need (enc, NV_DIG + 32);
434	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	662	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
435	enc->cur += strlen (enc->cur);	663	enc->cur += strlen (enc->cur);
436	}	664	}
437	else if (SvIOKp (sv))	665	else if (SvIOKp (sv))
438	{	666	{
		667	// we assume we can always read an IV as a UV and vice versa
		668	// we assume two's complement
		669	// we assume no aliasing issues in the union
		670	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		671	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		672	{
		673	// optimise the "small number case"
		674	// code will likely be branchless and use only a single multiplication
		675	// works for numbers up to 59074
		676	I32 i = SvIVX (sv);
		677	U32 u;
		678	char digit, nz = 0;
		679
439	need (enc, 64);	680	need (enc, 6);
		681
		682	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		683	u = i < 0 ? -i : i;
		684
		685	// convert to 4.28 fixed-point representation
		686	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		687
		688	// now output digit by digit, each time masking out the integer part
		689	// and multiplying by 5 while moving the decimal point one to the right,
		690	// resulting in a net multiplication by 10.
		691	// we always write the digit to memory but conditionally increment
		692	// the pointer, to enable the use of conditional move instructions.
		693	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		694	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		695	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		696	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		697	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		698	}
		699	else
		700	{
		701	// large integer, use the (rather slow) snprintf way.
		702	need (enc, IVUV_MAXCHARS);
440	enc->cur +=	703	enc->cur +=
441	SvIsUV(sv)	704	SvIsUV(sv)
442	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	705	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
443	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	706	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		707	}
444	}	708	}
445	else if (SvROK (sv))	709	else if (SvROK (sv))
446	{	710	encode_rv (enc, SvRV (sv));
447	SV *rv = SvRV (sv);
448
449	if (enc->indent >= enc->max_depth)
450	croak ("data structure too deep (hit recursion limit)");
451
452	switch (SvTYPE (rv))
453	{
454	case SVt_PVAV: encode_av (enc, (AV *)rv); break;
455	case SVt_PVHV: encode_hv (enc, (HV *)rv); break;
456
457	default:
458	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
459	SvPV_nolen (sv));
460	}
461	}
462	else if (!SvOK (sv))	711	else if (!SvOK (sv))
463	encode_str (enc, "null", 4, 0);	712	encode_str (enc, "null", 4, 0);
464	else	713	else
465	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	714	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
466	SvPV_nolen (sv), SvFLAGS (sv));	715	SvPV_nolen (sv), SvFLAGS (sv));
467	}	716	}
468		717
469	static SV *	718	static SV *
470	encode_json (SV *scalar, UV flags)	719	encode_json (SV *scalar, JSON *json)
471	{	720	{
		721	enc_t enc;
		722
472	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	723	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
473	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	724	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
474		725
475	enc_t enc;	726	enc.json = *json;
476	enc.flags = flags;
477	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	727	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
478	enc.cur = SvPVX (enc.sv);	728	enc.cur = SvPVX (enc.sv);
479	enc.end = SvEND (enc.sv);	729	enc.end = SvEND (enc.sv);
480	enc.indent = 0;	730	enc.indent = 0;
481	enc.max_depth = 0x7fffffffUL;	731	enc.maxdepth = DEC_DEPTH (enc.json.flags);
		732	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		733	: enc.json.flags & F_LATIN1 ? 0x000100UL
		734	: 0x110000UL;
482		735
483	SvPOK_only (enc.sv);	736	SvPOK_only (enc.sv);
484	encode_sv (&enc, scalar);	737	encode_sv (&enc, scalar);
485		738
		739	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		740	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		741
486	if (!(flags & (F_ASCII | F_UTF8)))	742	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
487	SvUTF8_on (enc.sv);	743	SvUTF8_on (enc.sv);
488		744
489	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
490
491	if (enc.flags & F_SHRINK)	745	if (enc.json.flags & F_SHRINK)
492	shrink (enc.sv);	746	shrink (enc.sv);
493		747
494	return enc.sv;	748	return enc.sv;
495	}	749	}
496		750
…		…
501	typedef struct	755	typedef struct
502	{	756	{
503	char *cur; // current parser pointer	757	char *cur; // current parser pointer
504	char *end; // end of input string	758	char *end; // end of input string
505	const char *err; // parse error, if != 0	759	const char *err; // parse error, if != 0
506	UV flags; // F_*	760	JSON json;
		761	U32 depth; // recursion depth
		762	U32 maxdepth; // recursion depth limit
507	} dec_t;	763	} dec_t;
508		764
509	static void	765	INLINE void
		766	decode_comment (dec_t *dec)
		767	{
		768	// only '#'-style comments allowed a.t.m.
		769
		770	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		771	++dec->cur;
		772	}
		773
		774	INLINE void
510	decode_ws (dec_t *dec)	775	decode_ws (dec_t *dec)
511	{	776	{
512	for (;;)	777	for (;;)
513	{	778	{
514	char ch = *dec->cur;	779	char ch = *dec->cur;
515		780
516	if (ch > 0x20	781	if (ch > 0x20)
		782	{
		783	if (expect_false (ch == '#'))
		784	{
		785	if (dec->json.flags & F_RELAXED)
		786	decode_comment (dec);
		787	else
		788	break;
		789	}
		790	else
		791	break;
		792	}
517	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	793	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
518	break;	794	break; // parse error, but let higher level handle it, gives better error messages
519		795
520	++dec->cur;	796	++dec->cur;
521	}	797	}
522	}	798	}
523		799
524	#define ERR(reason) SB dec->err = reason; goto fail; SE	800	#define ERR(reason) SB dec->err = reason; goto fail; SE
		801
525	#define EXPECT_CH(ch) SB \	802	#define EXPECT_CH(ch) SB \
526	if (*dec->cur != ch) \	803	if (*dec->cur != ch) \
527	ERR (# ch " expected"); \	804	ERR (# ch " expected"); \
528	++dec->cur; \	805	++dec->cur; \
529	SE	806	SE
530		807
		808	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		809	#define DEC_DEC_DEPTH --dec->depth
		810
531	static SV *decode_sv (dec_t *dec);	811	static SV *decode_sv (dec_t *dec);
532		812
533	static signed char decode_hexdigit[256];	813	static signed char decode_hexdigit[256];
534		814
535	static UV	815	static UV
536	decode_4hex (dec_t *dec)	816	decode_4hex (dec_t *dec)
537	{	817	{
538	signed char d1, d2, d3, d4;	818	signed char d1, d2, d3, d4;
539	unsigned char *cur = (unsigned char *)dec->cur;	819	unsigned char *cur = (unsigned char *)dec->cur;
540		820
541	d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected");	821	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
542	d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected");	822	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
543	d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected");	823	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
544	d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected");	824	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
545		825
546	dec->cur += 4;	826	dec->cur += 4;
547		827
548	return ((UV)d1) << 12	828	return ((UV)d1) << 12
549	| ((UV)d2) << 8	829	| ((UV)d2) << 8
…		…
557	static SV *	837	static SV *
558	decode_str (dec_t *dec)	838	decode_str (dec_t *dec)
559	{	839	{
560	SV *sv = 0;	840	SV *sv = 0;
561	int utf8 = 0;	841	int utf8 = 0;
		842	char *dec_cur = dec->cur;
562		843
563	do	844	do
564	{	845	{
565	char buf [SHORT_STRING_LEN + UTF8_MAX_LEN];	846	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
566	char *cur = buf;	847	char *cur = buf;
567		848
568	do	849	do
569	{	850	{
570	unsigned char ch = *(unsigned char *)dec->cur++;	851	unsigned char ch = *(unsigned char *)dec_cur++;
571		852
572	if (ch == '"')	853	if (expect_false (ch == '"'))
573	{	854	{
574	--dec->cur;	855	--dec_cur;
575	break;	856	break;
576	}	857	}
577	else if (ch == '\\')	858	else if (expect_false (ch == '\\'))
578	{	859	{
579	switch (*dec->cur)	860	switch (*dec_cur)
580	{	861	{
581	case '\\':	862	case '\\':
582	case '/':	863	case '/':
583	case '"': *cur++ = *dec->cur++; break;	864	case '"': *cur++ = *dec_cur++; break;
584		865
585	case 'b': ++dec->cur; *cur++ = '\010'; break;	866	case 'b': ++dec_cur; *cur++ = '\010'; break;
586	case 't': ++dec->cur; *cur++ = '\011'; break;	867	case 't': ++dec_cur; *cur++ = '\011'; break;
587	case 'n': ++dec->cur; *cur++ = '\012'; break;	868	case 'n': ++dec_cur; *cur++ = '\012'; break;
588	case 'f': ++dec->cur; *cur++ = '\014'; break;	869	case 'f': ++dec_cur; *cur++ = '\014'; break;
589	case 'r': ++dec->cur; *cur++ = '\015'; break;	870	case 'r': ++dec_cur; *cur++ = '\015'; break;
590		871
591	case 'u':	872	case 'u':
592	{	873	{
593	UV lo, hi;	874	UV lo, hi;
594	++dec->cur;	875	++dec_cur;
595		876
		877	dec->cur = dec_cur;
596	hi = decode_4hex (dec);	878	hi = decode_4hex (dec);
		879	dec_cur = dec->cur;
597	if (hi == (UV)-1)	880	if (hi == (UV)-1)
598	goto fail;	881	goto fail;
599		882
600	// possibly a surrogate pair	883	// possibly a surrogate pair
601	if (hi >= 0xd800)	884	if (hi >= 0xd800)
602	if (hi < 0xdc00)	885	if (hi < 0xdc00)
603	{	886	{
604	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	887	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
605	ERR ("missing low surrogate character in surrogate pair");	888	ERR ("missing low surrogate character in surrogate pair");
606		889
607	dec->cur += 2;	890	dec_cur += 2;
608		891
		892	dec->cur = dec_cur;
609	lo = decode_4hex (dec);	893	lo = decode_4hex (dec);
		894	dec_cur = dec->cur;
610	if (lo == (UV)-1)	895	if (lo == (UV)-1)
611	goto fail;	896	goto fail;
612		897
613	if (lo < 0xdc00 || lo >= 0xe000)	898	if (lo < 0xdc00 || lo >= 0xe000)
614	ERR ("surrogate pair expected");	899	ERR ("surrogate pair expected");
…		…
620		905
621	if (hi >= 0x80)	906	if (hi >= 0x80)
622	{	907	{
623	utf8 = 1;	908	utf8 = 1;
624		909
625	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	910	cur = encode_utf8 (cur, hi);
626	}	911	}
627	else	912	else
628	*cur++ = hi;	913	*cur++ = hi;
629	}	914	}
630	break;	915	break;
631		916
632	default:	917	default:
633	--dec->cur;	918	--dec_cur;
634	ERR ("illegal backslash escape sequence in string");	919	ERR ("illegal backslash escape sequence in string");
635	}	920	}
636	}	921	}
637	else if (ch >= 0x20 && ch <= 0x7f)	922	else if (expect_true (ch >= 0x20 && ch < 0x80))
638	*cur++ = ch;	923	*cur++ = ch;
639	else if (ch >= 0x80)	924	else if (ch >= 0x80)
640	{	925	{
641	--dec->cur;
642
643	STRLEN clen;	926	STRLEN clen;
		927	UV uch;
		928
		929	--dec_cur;
		930
644	UV uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	931	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
645	if (clen == (STRLEN)-1)	932	if (clen == (STRLEN)-1)
646	ERR ("malformed UTF-8 character in JSON string");	933	ERR ("malformed UTF-8 character in JSON string");
647		934
648	do	935	do
649	{
650	*cur++ = *dec->cur++;	936	*cur++ = *dec_cur++;
651	}
652	while (--clen);	937	while (--clen);
653		938
654	utf8 = 1;	939	utf8 = 1;
655	}	940	}
656	else if (!ch)
657	ERR ("unexpected end of string while parsing json string");
658	else	941	else
		942	{
		943	--dec_cur;
		944
		945	if (!ch)
		946	ERR ("unexpected end of string while parsing JSON string");
		947	else
659	ERR ("invalid character encountered");	948	ERR ("invalid character encountered while parsing JSON string");
660		949	}
661	}	950	}
662	while (cur < buf + SHORT_STRING_LEN);	951	while (cur < buf + SHORT_STRING_LEN);
663		952
		953	{
664	STRLEN len = cur - buf;	954	STRLEN len = cur - buf;
665		955
666	if (sv)	956	if (sv)
667	{	957	{
668	SvGROW (sv, SvCUR (sv) + len + 1);	958	SvGROW (sv, SvCUR (sv) + len + 1);
669	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	959	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
670	SvCUR_set (sv, SvCUR (sv) + len);	960	SvCUR_set (sv, SvCUR (sv) + len);
671	}	961	}
672	else	962	else
673	sv = newSVpvn (buf, len);	963	sv = newSVpvn (buf, len);
674	}	964	}
		965	}
675	while (*dec->cur != '"');	966	while (*dec_cur != '"');
676		967
677	++dec->cur;	968	++dec_cur;
678		969
679	if (sv)	970	if (sv)
680	{	971	{
681	SvPOK_only (sv);	972	SvPOK_only (sv);
682	*SvEND (sv) = 0;	973	*SvEND (sv) = 0;
…		…
685	SvUTF8_on (sv);	976	SvUTF8_on (sv);
686	}	977	}
687	else	978	else
688	sv = newSVpvn ("", 0);	979	sv = newSVpvn ("", 0);
689		980
		981	dec->cur = dec_cur;
690	return sv;	982	return sv;
691		983
692	fail:	984	fail:
		985	dec->cur = dec_cur;
693	return 0;	986	return 0;
694	}	987	}
695		988
696	static SV *	989	static SV *
697	decode_num (dec_t *dec)	990	decode_num (dec_t *dec)
…		…
755	is_nv = 1;	1048	is_nv = 1;
756	}	1049	}
757		1050
758	if (!is_nv)	1051	if (!is_nv)
759	{	1052	{
760	UV uv;	1053	int len = dec->cur - start;
761	int numtype = grok_number (start, dec->cur - start, &uv);	1054
762	if (numtype & IS_NUMBER_IN_UV)	1055	// special case the rather common 1..5-digit-int case
763	if (numtype & IS_NUMBER_NEG)	1056	if (*start == '-')
		1057	switch (len)
764	{	1058	{
765	if (uv < (UV)IV_MIN)	1059	case 2: return newSViv (-( start [1] - '0' * 1));
766	return newSViv (-(IV)uv);	1060	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		1061	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1062	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1063	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
767	}	1064	}
		1065	else
		1066	switch (len)
		1067	{
		1068	case 1: return newSViv ( start [0] - '0' * 1);
		1069	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1070	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1071	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1072	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1073	}
		1074
		1075	{
		1076	UV uv;
		1077	int numtype = grok_number (start, len, &uv);
		1078	if (numtype & IS_NUMBER_IN_UV)
		1079	if (numtype & IS_NUMBER_NEG)
		1080	{
		1081	if (uv < (UV)IV_MIN)
		1082	return newSViv (-(IV)uv);
		1083	}
768	else	1084	else
769	return newSVuv (uv);	1085	return newSVuv (uv);
770	}	1086	}
771		1087
		1088	len -= *start == '-' ? 1 : 0;
		1089
		1090	// does not fit into IV or UV, try NV
		1091	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1092	#if defined (LDBL_DIG)
		1093	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1094	#endif
		1095	)
		1096	// fits into NV without loss of precision
		1097	return newSVnv (Atof (start));
		1098
		1099	// everything else fails, convert it to a string
		1100	return newSVpvn (start, dec->cur - start);
		1101	}
		1102
		1103	// loss of precision here
772	return newSVnv (Atof (start));	1104	return newSVnv (Atof (start));
773		1105
774	fail:	1106	fail:
775	return 0;	1107	return 0;
776	}	1108	}
…		…
778	static SV *	1110	static SV *
779	decode_av (dec_t *dec)	1111	decode_av (dec_t *dec)
780	{	1112	{
781	AV *av = newAV ();	1113	AV *av = newAV ();
782		1114
		1115	DEC_INC_DEPTH;
783	decode_ws (dec);	1116	decode_ws (dec);
		1117
784	if (*dec->cur == ']')	1118	if (*dec->cur == ']')
785	++dec->cur;	1119	++dec->cur;
786	else	1120	else
787	for (;;)	1121	for (;;)
788	{	1122	{
…		…
804		1138
805	if (*dec->cur != ',')	1139	if (*dec->cur != ',')
806	ERR (", or ] expected while parsing array");	1140	ERR (", or ] expected while parsing array");
807		1141
808	++dec->cur;	1142	++dec->cur;
		1143
		1144	decode_ws (dec);
		1145
		1146	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1147	{
		1148	++dec->cur;
		1149	break;
		1150	}
809	}	1151	}
810		1152
		1153	DEC_DEC_DEPTH;
811	return newRV_noinc ((SV *)av);	1154	return newRV_noinc ((SV *)av);
812		1155
813	fail:	1156	fail:
814	SvREFCNT_dec (av);	1157	SvREFCNT_dec (av);
		1158	DEC_DEC_DEPTH;
815	return 0;	1159	return 0;
816	}	1160	}
817		1161
818	static SV *	1162	static SV *
819	decode_hv (dec_t *dec)	1163	decode_hv (dec_t *dec)
820	{	1164	{
		1165	SV *sv;
821	HV *hv = newHV ();	1166	HV *hv = newHV ();
822		1167
		1168	DEC_INC_DEPTH;
823	decode_ws (dec);	1169	decode_ws (dec);
		1170
824	if (*dec->cur == '}')	1171	if (*dec->cur == '}')
825	++dec->cur;	1172	++dec->cur;
826	else	1173	else
827	for (;;)	1174	for (;;)
828	{	1175	{
		1176	EXPECT_CH ('"');
		1177
		1178	// heuristic: assume that
		1179	// a) decode_str + hv_store_ent are abysmally slow.
		1180	// b) most hash keys are short, simple ascii text.
		1181	// => try to "fast-match" such strings to avoid
		1182	// the overhead of decode_str + hv_store_ent.
		1183	{
829	SV *key, *value;	1184	SV *value;
		1185	char *p = dec->cur;
		1186	char *e = p + 24; // only try up to 24 bytes
830		1187
831	decode_ws (dec); EXPECT_CH ('"');	1188	for (;;)
832
833	key = decode_str (dec);
834	if (!key)
835	goto fail;
836
837	decode_ws (dec); EXPECT_CH (':');
838
839	value = decode_sv (dec);
840	if (!value)
841	{	1189	{
		1190	// the >= 0x80 is false on most architectures
		1191	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1192	{
		1193	// slow path, back up and use decode_str
		1194	SV *key = decode_str (dec);
		1195	if (!key)
		1196	goto fail;
		1197
		1198	decode_ws (dec); EXPECT_CH (':');
		1199
		1200	decode_ws (dec);
		1201	value = decode_sv (dec);
		1202	if (!value)
		1203	{
		1204	SvREFCNT_dec (key);
		1205	goto fail;
		1206	}
		1207
		1208	hv_store_ent (hv, key, value, 0);
842	SvREFCNT_dec (key);	1209	SvREFCNT_dec (key);
		1210
		1211	break;
		1212	}
		1213	else if (*p == '"')
		1214	{
		1215	// fast path, got a simple key
		1216	char *key = dec->cur;
		1217	int len = p - key;
		1218	dec->cur = p + 1;
		1219
		1220	decode_ws (dec); EXPECT_CH (':');
		1221
		1222	decode_ws (dec);
		1223	value = decode_sv (dec);
		1224	if (!value)
843	goto fail;	1225	goto fail;
		1226
		1227	hv_store (hv, key, len, value, 0);
		1228
		1229	break;
		1230	}
		1231
		1232	++p;
844	}	1233	}
845		1234	}
846	//TODO: optimise
847	hv_store_ent (hv, key, value, 0);
848		1235
849	decode_ws (dec);	1236	decode_ws (dec);
850		1237
851	if (*dec->cur == '}')	1238	if (*dec->cur == '}')
852	{	1239	{
…		…
856		1243
857	if (*dec->cur != ',')	1244	if (*dec->cur != ',')
858	ERR (", or } expected while parsing object/hash");	1245	ERR (", or } expected while parsing object/hash");
859		1246
860	++dec->cur;	1247	++dec->cur;
		1248
		1249	decode_ws (dec);
		1250
		1251	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1252	{
		1253	++dec->cur;
		1254	break;
		1255	}
861	}	1256	}
862		1257
		1258	DEC_DEC_DEPTH;
863	return newRV_noinc ((SV *)hv);	1259	sv = newRV_noinc ((SV *)hv);
		1260
		1261	// check filter callbacks
		1262	if (dec->json.flags & F_HOOK)
		1263	{
		1264	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1265	{
		1266	HE *cb, *he;
		1267
		1268	hv_iterinit (hv);
		1269	he = hv_iternext (hv);
		1270	hv_iterinit (hv);
		1271
		1272	// the next line creates a mortal sv each time its called.
		1273	// might want to optimise this for common cases.
		1274	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1275
		1276	if (cb)
		1277	{
		1278	dSP;
		1279	int count;
		1280
		1281	ENTER; SAVETMPS; PUSHMARK (SP);
		1282	XPUSHs (HeVAL (he));
		1283
		1284	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1285
		1286	if (count == 1)
		1287	{
		1288	sv = newSVsv (POPs);
		1289	FREETMPS; LEAVE;
		1290	return sv;
		1291	}
		1292
		1293	FREETMPS; LEAVE;
		1294	}
		1295	}
		1296
		1297	if (dec->json.cb_object)
		1298	{
		1299	dSP;
		1300	int count;
		1301
		1302	ENTER; SAVETMPS; PUSHMARK (SP);
		1303	XPUSHs (sv_2mortal (sv));
		1304
		1305	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1306
		1307	if (count == 1)
		1308	{
		1309	sv = newSVsv (POPs);
		1310	FREETMPS; LEAVE;
		1311	return sv;
		1312	}
		1313
		1314	SvREFCNT_inc (sv);
		1315	FREETMPS; LEAVE;
		1316	}
		1317	}
		1318
		1319	return sv;
864		1320
865	fail:	1321	fail:
866	SvREFCNT_dec (hv);	1322	SvREFCNT_dec (hv);
		1323	DEC_DEC_DEPTH;
867	return 0;	1324	return 0;
868	}	1325	}
869		1326
870	static SV *	1327	static SV *
871	decode_sv (dec_t *dec)	1328	decode_sv (dec_t *dec)
872	{	1329	{
873	decode_ws (dec);	1330	// the beauty of JSON: you need exactly one character lookahead
		1331	// to parse everything.
874	switch (*dec->cur)	1332	switch (*dec->cur)
875	{	1333	{
876	case '"': ++dec->cur; return decode_str (dec);	1334	case '"': ++dec->cur; return decode_str (dec);
877	case '[': ++dec->cur; return decode_av (dec);	1335	case '[': ++dec->cur; return decode_av (dec);
878	case '{': ++dec->cur; return decode_hv (dec);	1336	case '{': ++dec->cur; return decode_hv (dec);
879		1337
880	case '-':	1338	case '-':
881	case '0': case '1': case '2': case '3': case '4':	1339	case '0': case '1': case '2': case '3': case '4':
882	case '5': case '6': case '7': case '8': case '9':	1340	case '5': case '6': case '7': case '8': case '9':
883	return decode_num (dec);	1341	return decode_num (dec);
884		1342
885	case 't':	1343	case 't':
886	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1344	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
887	{	1345	{
888	dec->cur += 4;	1346	dec->cur += 4;
889	return newSViv (1);	1347	#if JSON_SLOW
		1348	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);
		1349	#endif
		1350	return SvREFCNT_inc (json_true);
890	}	1351	}
891	else	1352	else
892	ERR ("'true' expected");	1353	ERR ("'true' expected");
893		1354
894	break;	1355	break;
895		1356
896	case 'f':	1357	case 'f':
897	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1358	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
898	{	1359	{
899	dec->cur += 5;	1360	dec->cur += 5;
900	return newSViv (0);	1361	#if JSON_SLOW
		1362	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1363	#endif
		1364	return SvREFCNT_inc (json_false);
901	}	1365	}
902	else	1366	else
903	ERR ("'false' expected");	1367	ERR ("'false' expected");
904		1368
905	break;	1369	break;
…		…
914	ERR ("'null' expected");	1378	ERR ("'null' expected");
915		1379
916	break;	1380	break;
917		1381
918	default:	1382	default:
919	ERR ("malformed json string, neither array, object, number, string or atom");	1383	ERR ("malformed JSON string, neither array, object, number, string or atom");
920	break;	1384	break;
921	}	1385	}
922		1386
923	fail:	1387	fail:
924	return 0;	1388	return 0;
925	}	1389	}
926		1390
927	static SV *	1391	static SV *
928	decode_json (SV *string, UV flags)	1392	decode_json (SV *string, JSON *json, STRLEN *offset_return)
929	{	1393	{
		1394	dec_t dec;
		1395	STRLEN offset;
930	SV *sv;	1396	SV *sv;
931		1397
		1398	SvGETMAGIC (string);
		1399	SvUPGRADE (string, SVt_PV);
		1400
		1401	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1402	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1403	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1404
932	if (flags & F_UTF8)	1405	if (json->flags & F_UTF8)
933	sv_utf8_downgrade (string, 0);	1406	sv_utf8_downgrade (string, 0);
934	else	1407	else
935	sv_utf8_upgrade (string);	1408	sv_utf8_upgrade (string);
936		1409
937	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1410	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
938		1411
939	dec_t dec;	1412	dec.json = *json;
940	dec.flags = flags;
941	dec.cur = SvPVX (string);	1413	dec.cur = SvPVX (string);
942	dec.end = SvEND (string);	1414	dec.end = SvEND (string);
943	dec.err = 0;	1415	dec.err = 0;
		1416	dec.depth = 0;
		1417	dec.maxdepth = DEC_DEPTH (dec.json.flags);
944		1418
		1419	if (dec.json.cb_object || dec.json.cb_sk_object)
		1420	dec.json.flags |= F_HOOK;
		1421
		1422	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1423
		1424	decode_ws (&dec);
945	sv = decode_sv (&dec);	1425	sv = decode_sv (&dec);
946		1426
		1427	if (!(offset_return || !sv))
		1428	{
		1429	// check for trailing garbage
		1430	decode_ws (&dec);
		1431
		1432	if (*dec.cur)
		1433	{
		1434	dec.err = "garbage after JSON object";
		1435	SvREFCNT_dec (sv);
		1436	sv = 0;
		1437	}
		1438	}
		1439
		1440	if (offset_return || !sv)
		1441	{
		1442	offset = dec.json.flags & F_UTF8
		1443	? dec.cur - SvPVX (string)
		1444	: utf8_distance (dec.cur, SvPVX (string));
		1445
		1446	if (offset_return)
		1447	*offset_return = offset;
		1448	}
		1449
947	if (!sv)	1450	if (!sv)
948	{	1451	{
949	IV offset = dec.flags & F_UTF8
950	? dec.cur - SvPVX (string)
951	: utf8_distance (dec.cur, SvPVX (string));
952	SV *uni = sv_newmortal ();	1452	SV *uni = sv_newmortal ();
953		1453
954	// horrible hack to silence warning inside pv_uni_display	1454	// horrible hack to silence warning inside pv_uni_display
955	COP cop = *PL_curcop;	1455	COP cop = *PL_curcop;
956	cop.cop_warnings = pWARN_NONE;	1456	cop.cop_warnings = pWARN_NONE;
…		…
958	SAVEVPTR (PL_curcop);	1458	SAVEVPTR (PL_curcop);
959	PL_curcop = &cop;	1459	PL_curcop = &cop;
960	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1460	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
961	LEAVE;	1461	LEAVE;
962		1462
963	croak ("%s, at character offset %d (%s)",	1463	croak ("%s, at character offset %d [\"%s\"]",
964	dec.err,	1464	dec.err,
965	(int)offset,	1465	(int)offset,
966	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1466	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
967	}	1467	}
968		1468
969	sv = sv_2mortal (sv);	1469	sv = sv_2mortal (sv);
970		1470
971	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1471	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
972	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1472	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
973		1473
974	return sv;	1474	return sv;
		1475	}
		1476
		1477	/////////////////////////////////////////////////////////////////////////////
		1478	// incremental parser
		1479
		1480	static void
		1481	incr_parse (JSON *self)
		1482	{
		1483	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1484
		1485	for (;;)
		1486	{
		1487	//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
		1488	switch (self->incr_mode)
		1489	{
		1490	// only used for intiial whitespace skipping
		1491	case INCR_M_WS:
		1492	for (;;)
		1493	{
		1494	if (*p > 0x20)
		1495	{
		1496	self->incr_mode = INCR_M_JSON;
		1497	goto incr_m_json;
		1498	}
		1499	else if (!*p)
		1500	goto interrupt;
		1501
		1502	++p;
		1503	}
		1504
		1505	// skip a single char inside a string (for \\-processing)
		1506	case INCR_M_BS:
		1507	if (!*p)
		1508	goto interrupt;
		1509
		1510	++p;
		1511	self->incr_mode = INCR_M_STR;
		1512	goto incr_m_str;
		1513
		1514	// inside a string
		1515	case INCR_M_STR:
		1516	incr_m_str:
		1517	for (;;)
		1518	{
		1519	if (*p == '"')
		1520	{
		1521	++p;
		1522	self->incr_mode = INCR_M_JSON;
		1523
		1524	if (!self->incr_nest)
		1525	goto interrupt;
		1526
		1527	goto incr_m_json;
		1528	}
		1529	else if (*p == '\\')
		1530	{
		1531	++p; // "virtually" consumes character after \
		1532
		1533	if (!*p) // if at end of string we have to switch modes
		1534	{
		1535	self->incr_mode = INCR_M_BS;
		1536	goto interrupt;
		1537	}
		1538	}
		1539	else if (!*p)
		1540	goto interrupt;
		1541
		1542	++p;
		1543	}
		1544
		1545	// after initial ws, outside string
		1546	case INCR_M_JSON:
		1547	incr_m_json:
		1548	for (;;)
		1549	{
		1550	switch (*p++)
		1551	{
		1552	case 0:
		1553	--p;
		1554	goto interrupt;
		1555
		1556	case 0x09:
		1557	case 0x0a:
		1558	case 0x0d:
		1559	case 0x20:
		1560	if (!self->incr_nest)
		1561	{
		1562	--p; // do not eat the whitespace, let the next round do it
		1563	goto interrupt;
		1564	}
		1565	break;
		1566
		1567	case '"':
		1568	self->incr_mode = INCR_M_STR;
		1569	goto incr_m_str;
		1570
		1571	case '[':
		1572	case '{':
		1573	++self->incr_nest;
		1574	break;
		1575
		1576	case ']':
		1577	case '}':
		1578	if (!--self->incr_nest)
		1579	goto interrupt;
		1580	}
		1581	}
		1582	}
		1583
		1584	modechange:
		1585	;
		1586	}
		1587
		1588	interrupt:
		1589	self->incr_pos = p - SvPVX (self->incr_text);
		1590	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
975	}	1591	}
976		1592
977	/////////////////////////////////////////////////////////////////////////////	1593	/////////////////////////////////////////////////////////////////////////////
978	// XS interface functions	1594	// XS interface functions
979		1595
…		…
981		1597
982	BOOT:	1598	BOOT:
983	{	1599	{
984	int i;	1600	int i;
985		1601
986	memset (decode_hexdigit, 0xff, 256);
987	for (i = 10; i--; )	1602	for (i = 0; i < 256; ++i)
988	decode_hexdigit ['0' + i] = i;	1603	decode_hexdigit [i] =
		1604	i >= '0' && i <= '9' ? i - '0'
		1605	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1606	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1607	: -1;
989		1608
990	for (i = 7; i--; )
991	{
992	decode_hexdigit ['a' + i] = 10 + i;
993	decode_hexdigit ['A' + i] = 10 + i;
994	}
995
996	json_stash = gv_stashpv ("JSON::XS", 1);	1609	json_stash = gv_stashpv ("JSON::XS" , 1);
		1610	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1611
		1612	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1613	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
997	}	1614	}
998		1615
999	PROTOTYPES: DISABLE	1616	PROTOTYPES: DISABLE
1000		1617
1001	SV *new (char *dummy)	1618	void CLONE (...)
1002	CODE:	1619	CODE:
1003	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1620	json_stash = 0;
		1621	json_boolean_stash = 0;
		1622
		1623	void new (char *klass)
		1624	PPCODE:
		1625	{
		1626	SV *pv = NEWSV (0, sizeof (JSON));
		1627	SvPOK_only (pv);
		1628	Zero (SvPVX (pv), 1, JSON);
		1629	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1630	XPUSHs (sv_2mortal (sv_bless (
		1631	newRV_noinc (pv),
		1632	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1633	)));
		1634	}
		1635
		1636	void ascii (JSON *self, int enable = 1)
		1637	ALIAS:
		1638	ascii = F_ASCII
		1639	latin1 = F_LATIN1
		1640	utf8 = F_UTF8
		1641	indent = F_INDENT
		1642	canonical = F_CANONICAL
		1643	space_before = F_SPACE_BEFORE
		1644	space_after = F_SPACE_AFTER
		1645	pretty = F_PRETTY
		1646	allow_nonref = F_ALLOW_NONREF
		1647	shrink = F_SHRINK
		1648	allow_blessed = F_ALLOW_BLESSED
		1649	convert_blessed = F_CONV_BLESSED
		1650	relaxed = F_RELAXED
		1651	PPCODE:
		1652	{
		1653	if (enable)
		1654	self->flags |= ix;
		1655	else
		1656	self->flags &= ~ix;
		1657
		1658	XPUSHs (ST (0));
		1659	}
		1660
		1661	void get_ascii (JSON *self)
		1662	ALIAS:
		1663	get_ascii = F_ASCII
		1664	get_latin1 = F_LATIN1
		1665	get_utf8 = F_UTF8
		1666	get_indent = F_INDENT
		1667	get_canonical = F_CANONICAL
		1668	get_space_before = F_SPACE_BEFORE
		1669	get_space_after = F_SPACE_AFTER
		1670	get_allow_nonref = F_ALLOW_NONREF
		1671	get_shrink = F_SHRINK
		1672	get_allow_blessed = F_ALLOW_BLESSED
		1673	get_convert_blessed = F_CONV_BLESSED
		1674	get_relaxed = F_RELAXED
		1675	PPCODE:
		1676	XPUSHs (boolSV (self->flags & ix));
		1677
		1678	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1679	PPCODE:
		1680	{
		1681	UV log2 = 0;
		1682
		1683	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1684
		1685	while ((1UL << log2) < max_depth)
		1686	++log2;
		1687
		1688	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1689
		1690	XPUSHs (ST (0));
		1691	}
		1692
		1693	U32 get_max_depth (JSON *self)
		1694	CODE:
		1695	RETVAL = DEC_DEPTH (self->flags);
1004	OUTPUT:	1696	OUTPUT:
1005	RETVAL	1697	RETVAL
1006		1698
1007	SV *ascii (SV *self, int enable = 1)	1699	void max_size (JSON *self, UV max_size = 0)
1008	ALIAS:	1700	PPCODE:
1009	ascii = F_ASCII	1701	{
1010	utf8 = F_UTF8	1702	UV log2 = 0;
1011	indent = F_INDENT	1703
1012	canonical = F_CANONICAL	1704	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1013	space_before = F_SPACE_BEFORE	1705	if (max_size == 1) max_size = 2;
1014	space_after = F_SPACE_AFTER	1706
1015	pretty = F_PRETTY	1707	while ((1UL << log2) < max_size)
1016	allow_nonref = F_ALLOW_NONREF	1708	++log2;
1017	shrink = F_SHRINK	1709
		1710	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1711
		1712	XPUSHs (ST (0));
		1713	}
		1714
		1715	int get_max_size (JSON *self)
1018	CODE:	1716	CODE:
1019	{	1717	RETVAL = DEC_SIZE (self->flags);
1020	UV *uv = SvJSON (self);
1021	if (enable)
1022	*uv |= ix;
1023	else
1024	*uv &= ~ix;
1025
1026	RETVAL = newSVsv (self);
1027	}
1028	OUTPUT:	1718	OUTPUT:
1029	RETVAL	1719	RETVAL
1030		1720
1031	void encode (SV *self, SV *scalar)	1721	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1032	PPCODE:	1722	PPCODE:
1033	XPUSHs (encode_json (scalar, *SvJSON (self)));	1723	{
		1724	SvREFCNT_dec (self->cb_object);
		1725	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
1034		1726
1035	void decode (SV *self, SV *jsonstr)	1727	XPUSHs (ST (0));
		1728	}
		1729
		1730	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1036	PPCODE:	1731	PPCODE:
		1732	{
		1733	if (!self->cb_sk_object)
		1734	self->cb_sk_object = newHV ();
		1735
		1736	if (SvOK (cb))
		1737	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1738	else
		1739	{
		1740	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1741
		1742	if (!HvKEYS (self->cb_sk_object))
		1743	{
		1744	SvREFCNT_dec (self->cb_sk_object);
		1745	self->cb_sk_object = 0;
		1746	}
		1747	}
		1748
		1749	XPUSHs (ST (0));
		1750	}
		1751
		1752	void encode (JSON *self, SV *scalar)
		1753	PPCODE:
		1754	XPUSHs (encode_json (scalar, self));
		1755
		1756	void decode (JSON *self, SV *jsonstr)
		1757	PPCODE:
1037	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1758	XPUSHs (decode_json (jsonstr, self, 0));
		1759
		1760	void decode_prefix (JSON *self, SV *jsonstr)
		1761	PPCODE:
		1762	{
		1763	STRLEN offset;
		1764	EXTEND (SP, 2);
		1765	PUSHs (decode_json (jsonstr, self, &offset));
		1766	PUSHs (sv_2mortal (newSVuv (offset)));
		1767	}
		1768
		1769	void incr_parse (JSON *self, SV *jsonstr = 0)
		1770	PPCODE:
		1771	{
		1772	if (!self->incr_text)
		1773	self->incr_text = newSVpvn ("", 0);
		1774
		1775	// append data, if any
		1776	if (jsonstr)
		1777	{
		1778	if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text))
		1779	{
		1780	/* utf-8-ness differs, need to upgrade */
		1781	sv_utf8_upgrade (self->incr_text);
		1782
		1783	if (self->incr_pos)
		1784	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1785	- (U8 *)SvPVX (self->incr_text);
		1786	}
		1787
		1788	{
		1789	STRLEN len;
		1790	const char *str = SvPV (jsonstr, len);
		1791	SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1);
		1792	Move (str, SvEND (self->incr_text), len, char);
		1793	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1794	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1795	}
		1796	}
		1797
		1798	if (GIMME_V != G_VOID)
		1799	do
		1800	{
		1801	STRLEN offset;
		1802
		1803	incr_parse (self);
		1804
		1805	if (!INCR_DONE (self))
		1806	break;
		1807
		1808	XPUSHs (decode_json (self->incr_text, self, &offset));
		1809
		1810	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
		1811	self->incr_pos -= offset;
		1812	self->incr_nest = 0;
		1813	self->incr_mode = 0;
		1814	}
		1815	while (GIMME_V == G_ARRAY);
		1816	}
		1817
		1818	SV *incr_text (JSON *self)
		1819	ATTRS: lvalue
		1820	CODE:
		1821	{
		1822	if (self->incr_pos)
		1823	croak ("incr_text can only be called after a successful incr_parse call in scalar context");
		1824
		1825	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		1826	}
		1827	OUTPUT:
		1828	RETVAL
		1829
		1830	void incr_skip (JSON *self)
		1831	CODE:
		1832	{
		1833	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1834	self->incr_pos = 0;
		1835	self->incr_nest = 0;
		1836	self->incr_mode = 0;
		1837	}
		1838
		1839	void DESTROY (JSON *self)
		1840	CODE:
		1841	SvREFCNT_dec (self->cb_sk_object);
		1842	SvREFCNT_dec (self->cb_object);
		1843	SvREFCNT_dec (self->incr_text);
1038		1844
1039	PROTOTYPES: ENABLE	1845	PROTOTYPES: ENABLE
1040		1846
1041	void to_json (SV *scalar)	1847	void encode_json (SV *scalar)
		1848	ALIAS:
		1849	to_json_ = 0
		1850	encode_json = F_UTF8
1042	PPCODE:	1851	PPCODE:
		1852	{
		1853	JSON json = { F_DEFAULT | ix };
1043	XPUSHs (encode_json (scalar, F_UTF8));	1854	XPUSHs (encode_json (scalar, &json));
		1855	}
1044		1856
1045	void from_json (SV *jsonstr)	1857	void decode_json (SV *jsonstr)
		1858	ALIAS:
		1859	from_json_ = 0
		1860	decode_json = F_UTF8
1046	PPCODE:	1861	PPCODE:
		1862	{
		1863	JSON json = { F_DEFAULT | ix };
1047	XPUSHs (decode_json (jsonstr, F_UTF8));	1864	XPUSHs (decode_json (jsonstr, &json, 0));
		1865	}
1048		1866
		1867

Diff Legend

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