[ViewVC] Diff of: cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.53 by root, Tue Jul 10 15:45:34 2007 UTC vs.
Revision 1.92 by root, Mon Sep 22 07:29:29 2008 UTC

…		…
1	#include "EXTERN.h"	1	#include "EXTERN.h"
2	#include "perl.h"	2	#include "perl.h"
3	#include "XSUB.h"	3	#include "XSUB.h"
4		4
5	#include "assert.h"	5	#include <assert.h>
6	#include "string.h"	6	#include <string.h>
7	#include "stdlib.h"	7	#include <stdlib.h>
8	#include "stdio.h"	8	#include <stdio.h>
		9	#include <limits.h>
		10	#include <float.h>
9		11
10	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	12	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
11	# define snprintf _snprintf // C compilers have this in stdio.h	13	# define snprintf _snprintf // C compilers have this in stdio.h
12	#endif	14	#endif
13		15
14	// some old perls do not have this, try to make it work, no	16	// some old perls do not have this, try to make it work, no
15	// guarentees, though. if it breaks, you get to keep the pieces.	17	// guarentees, though. if it breaks, you get to keep the pieces.
16	#ifndef UTF8_MAXBYTES	18	#ifndef UTF8_MAXBYTES
17	# define UTF8_MAXBYTES 13	19	# define UTF8_MAXBYTES 13
18	#endif	20	#endif
		21
		22	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2)
19		23
20	#define F_ASCII 0x00000001UL	24	#define F_ASCII 0x00000001UL
21	#define F_LATIN1 0x00000002UL	25	#define F_LATIN1 0x00000002UL
22	#define F_UTF8 0x00000004UL	26	#define F_UTF8 0x00000004UL
23	#define F_INDENT 0x00000008UL	27	#define F_INDENT 0x00000008UL
26	#define F_SPACE_AFTER 0x00000040UL	30	#define F_SPACE_AFTER 0x00000040UL
27	#define F_ALLOW_NONREF 0x00000100UL	31	#define F_ALLOW_NONREF 0x00000100UL
28	#define F_SHRINK 0x00000200UL	32	#define F_SHRINK 0x00000200UL
29	#define F_ALLOW_BLESSED 0x00000400UL	33	#define F_ALLOW_BLESSED 0x00000400UL
30	#define F_CONV_BLESSED 0x00000800UL	34	#define F_CONV_BLESSED 0x00000800UL
31	#define F_MAXDEPTH 0xf8000000UL
32	#define S_MAXDEPTH 27
33	#define F_MAXSIZE 0x01f00000UL	35	#define F_RELAXED 0x00001000UL
34	#define S_MAXSIZE 20	36	#define F_ALLOW_UNKNOWN 0x00002000UL
35	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing	37	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
36		38
37	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
38	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
39
40	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER	39	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER
41	#define F_DEFAULT (9UL << S_MAXDEPTH)
42		40
43	#define INIT_SIZE 32 // initial scalar size to be allocated	41	#define INIT_SIZE 32 // initial scalar size to be allocated
44	#define INDENT_STEP 3 // spaces per indentation level	42	#define INDENT_STEP 3 // spaces per indentation level
45		43
46	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size	44	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
47		45
48	#define SB do {	46	#define SB do {
49	#define SE } while (0)	47	#define SE } while (0)
50		48
51	#if __GNUC__ >= 3	49	#if __GNUC__ >= 3
52	# define expect(expr,value) __builtin_expect ((expr),(value))	50	# define expect(expr,value) __builtin_expect ((expr), (value))
53	# define inline inline	51	# define INLINE static inline
54	#else	52	#else
55	# define expect(expr,value) (expr)	53	# define expect(expr,value) (expr)
56	# define inline static	54	# define INLINE static
57	#endif	55	#endif
58		56
59	#define expect_false(expr) expect ((expr) != 0, 0)	57	#define expect_false(expr) expect ((expr) != 0, 0)
60	#define expect_true(expr) expect ((expr) != 0, 1)	58	#define expect_true(expr) expect ((expr) != 0, 1)
61		59
		60	#define IN_RANGE_INC(type,val,beg,end) \
		61	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		62	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		63
		64	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
		65
		66	#ifdef USE_ITHREADS
		67	# define JSON_SLOW 1
		68	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		69	#else
		70	# define JSON_SLOW 0
		71	# define JSON_STASH json_stash
		72	#endif
		73
62	static HV json_stash, json_boolean_stash; // JSON::XS::	74	static HV json_stash, json_boolean_stash; // JSON::XS::
63	static SV json_true, json_false;	75	static SV json_true, json_false;
64		76
		77	enum {
		78	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		79	INCR_M_STR, // inside string
		80	INCR_M_BS, // inside backslash
		81	INCR_M_JSON // outside anything, count nesting
		82	};
		83
		84	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		85
65	typedef struct {	86	typedef struct {
66	U32 flags;	87	U32 flags;
		88	U32 max_depth;
		89	STRLEN max_size;
		90
67	SV *cb_object;	91	SV *cb_object;
68	HV *cb_sk_object;	92	HV *cb_sk_object;
		93
		94	// for the incremental parser
		95	SV *incr_text; // the source text so far
		96	STRLEN incr_pos; // the current offset into the text
		97	int incr_nest; // {[]}-nesting level
		98	unsigned char incr_mode;
69	} JSON;	99	} JSON;
		100
		101	INLINE void
		102	json_init (JSON *json)
		103	{
		104	Zero (json, 1, JSON);
		105	json->max_depth = 512;
		106	}
70		107
71	/////////////////////////////////////////////////////////////////////////////	108	/////////////////////////////////////////////////////////////////////////////
72	// utility functions	109	// utility functions
73		110
74	inline void	111	INLINE SV *
		112	get_bool (const char *name)
		113	{
		114	SV *sv = get_sv (name, 1);
		115
		116	SvREADONLY_on (sv);
		117	SvREADONLY_on (SvRV (sv));
		118
		119	return sv;
		120	}
		121
		122	INLINE void
75	shrink (SV *sv)	123	shrink (SV *sv)
76	{	124	{
77	sv_utf8_downgrade (sv, 1);	125	sv_utf8_downgrade (sv, 1);
		126
78	if (SvLEN (sv) > SvCUR (sv) + 1)	127	if (SvLEN (sv) > SvCUR (sv) + 1)
79	{	128	{
80	#ifdef SvPV_shrink_to_cur	129	#ifdef SvPV_shrink_to_cur
81	SvPV_shrink_to_cur (sv);	130	SvPV_shrink_to_cur (sv);
82	#elif defined (SvPV_renew)	131	#elif defined (SvPV_renew)
…		…
88	// decode an utf-8 character and return it, or (UV)-1 in	137	// decode an utf-8 character and return it, or (UV)-1 in
89	// case of an error.	138	// case of an error.
90	// we special-case "safe" characters from U+80 .. U+7FF,	139	// we special-case "safe" characters from U+80 .. U+7FF,
91	// but use the very good perl function to parse anything else.	140	// but use the very good perl function to parse anything else.
92	// note that we never call this function for a ascii codepoints	141	// note that we never call this function for a ascii codepoints
93	inline UV	142	INLINE UV
94	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	143	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
95	{	144	{
96	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))	145	if (expect_true (len >= 2
97	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	146	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
98	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	147	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
99	{	148	{
100	*clen = 2;	149	*clen = 2;
101	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);	150	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
102	}	151	}
103	else	152	else
104	{	153	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
105	*clen = (STRLEN)-1;	154	}
106	return (UV)-1;	155
107	}	156	// likewise for encoding, also never called for ascii codepoints
		157	// this function takes advantage of this fact, although current gccs
		158	// seem to optimise the check for >= 0x80 away anyways
		159	INLINE unsigned char *
		160	encode_utf8 (unsigned char *s, UV ch)
		161	{
		162	if (expect_false (ch < 0x000080))
		163	*s++ = ch;
		164	else if (expect_true (ch < 0x000800))
		165	*s++ = 0xc0 \| ( ch >> 6),
		166	*s++ = 0x80 \| ( ch & 0x3f);
		167	else if ( ch < 0x010000)
		168	*s++ = 0xe0 \| ( ch >> 12),
		169	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		170	*s++ = 0x80 \| ( ch & 0x3f);
		171	else if ( ch < 0x110000)
		172	*s++ = 0xf0 \| ( ch >> 18),
		173	*s++ = 0x80 \| ((ch >> 12) & 0x3f),
		174	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		175	*s++ = 0x80 \| ( ch & 0x3f);
		176
		177	return s;
108	}	178	}
109		179
110	/////////////////////////////////////////////////////////////////////////////	180	/////////////////////////////////////////////////////////////////////////////
111	// encoder	181	// encoder
112		182
…		…
116	char *cur; // SvPVX (sv) + current output position	186	char *cur; // SvPVX (sv) + current output position
117	char *end; // SvEND (sv)	187	char *end; // SvEND (sv)
118	SV *sv; // result scalar	188	SV *sv; // result scalar
119	JSON json;	189	JSON json;
120	U32 indent; // indentation level	190	U32 indent; // indentation level
121	U32 maxdepth; // max. indentation/recursion level	191	UV limit; // escape character values >= this value when encoding
122	} enc_t;	192	} enc_t;
123		193
124	inline void	194	INLINE void
125	need (enc_t *enc, STRLEN len)	195	need (enc_t *enc, STRLEN len)
126	{	196	{
127	if (expect_false (enc->cur + len >= enc->end))	197	if (expect_false (enc->cur + len >= enc->end))
128	{	198	{
129	STRLEN cur = enc->cur - SvPVX (enc->sv);	199	STRLEN cur = enc->cur - SvPVX (enc->sv);
…		…
131	enc->cur = SvPVX (enc->sv) + cur;	201	enc->cur = SvPVX (enc->sv) + cur;
132	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	202	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
133	}	203	}
134	}	204	}
135		205
136	inline void	206	INLINE void
137	encode_ch (enc_t *enc, char ch)	207	encode_ch (enc_t *enc, char ch)
138	{	208	{
139	need (enc, 1);	209	need (enc, 1);
140	*enc->cur++ = ch;	210	*enc->cur++ = ch;
141	}	211	}
…		…
195	{	265	{
196	uch = ch;	266	uch = ch;
197	clen = 1;	267	clen = 1;
198	}	268	}
199		269
200	if (uch > 0x10FFFFUL)	270	if (uch < 0x80/0x20/ \|\| uch >= enc->limit)
201	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
202
203	if (uch < 0x80 \|\| enc->json.flags & F_ASCII \|\| (enc->json.flags & F_LATIN1 && uch > 0xFF))
204	{	271	{
205	if (uch > 0xFFFFUL)	272	if (uch >= 0x10000UL)
206	{	273	{
		274	if (uch >= 0x110000UL)
		275	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		276
207	need (enc, len += 11);	277	need (enc, len += 11);
208	sprintf (enc->cur, "\\u%04x\\u%04x",	278	sprintf (enc->cur, "\\u%04x\\u%04x",
209	(int)((uch - 0x10000) / 0x400 + 0xD800),	279	(int)((uch - 0x10000) / 0x400 + 0xD800),
210	(int)((uch - 0x10000) % 0x400 + 0xDC00));	280	(int)((uch - 0x10000) % 0x400 + 0xDC00));
211	enc->cur += 12;	281	enc->cur += 12;
212	}	282	}
213	else	283	else
214	{	284	{
215	static char hexdigit [16] = "0123456789abcdef";
216	need (enc, len += 5);	285	need (enc, len += 5);
217	*enc->cur++ = '\\';	286	*enc->cur++ = '\\';
218	*enc->cur++ = 'u';	287	*enc->cur++ = 'u';
219	*enc->cur++ = hexdigit [ uch >> 12 ];	288	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
220	*enc->cur++ = hexdigit [(uch >> 8) & 15];	289	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
221	*enc->cur++ = hexdigit [(uch >> 4) & 15];	290	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
222	*enc->cur++ = hexdigit [(uch >> 0) & 15];	291	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
223	}	292	}
224		293
225	str += clen;	294	str += clen;
226	}	295	}
227	else if (enc->json.flags & F_LATIN1)	296	else if (enc->json.flags & F_LATIN1)
…		…
239	while (--clen);	308	while (--clen);
240	}	309	}
241	else	310	else
242	{	311	{
243	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	312	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
244	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	313	enc->cur = encode_utf8 (enc->cur, uch);
245	++str;	314	++str;
246	}	315	}
247	}	316	}
248	}	317	}
249	}	318	}
250		319
251	--len;	320	--len;
252	}	321	}
253	}	322	}
254		323
255	inline void	324	INLINE void
256	encode_indent (enc_t *enc)	325	encode_indent (enc_t *enc)
257	{	326	{
258	if (enc->json.flags & F_INDENT)	327	if (enc->json.flags & F_INDENT)
259	{	328	{
260	int spaces = enc->indent * INDENT_STEP;	329	int spaces = enc->indent * INDENT_STEP;
…		…
263	memset (enc->cur, ' ', spaces);	332	memset (enc->cur, ' ', spaces);
264	enc->cur += spaces;	333	enc->cur += spaces;
265	}	334	}
266	}	335	}
267		336
268	inline void	337	INLINE void
269	encode_space (enc_t *enc)	338	encode_space (enc_t *enc)
270	{	339	{
271	need (enc, 1);	340	need (enc, 1);
272	encode_ch (enc, ' ');	341	encode_ch (enc, ' ');
273	}	342	}
274		343
275	inline void	344	INLINE void
276	encode_nl (enc_t *enc)	345	encode_nl (enc_t *enc)
277	{	346	{
278	if (enc->json.flags & F_INDENT)	347	if (enc->json.flags & F_INDENT)
279	{	348	{
280	need (enc, 1);	349	need (enc, 1);
281	encode_ch (enc, '\n');	350	encode_ch (enc, '\n');
282	}	351	}
283	}	352	}
284		353
285	inline void	354	INLINE void
286	encode_comma (enc_t *enc)	355	encode_comma (enc_t *enc)
287	{	356	{
288	encode_ch (enc, ',');	357	encode_ch (enc, ',');
289		358
290	if (enc->json.flags & F_INDENT)	359	if (enc->json.flags & F_INDENT)
…		…
298	static void	367	static void
299	encode_av (enc_t enc, AV av)	368	encode_av (enc_t enc, AV av)
300	{	369	{
301	int i, len = av_len (av);	370	int i, len = av_len (av);
302		371
303	if (enc->indent >= enc->maxdepth)	372	if (enc->indent >= enc->json.max_depth)
304	croak ("data structure too deep (hit recursion limit)");	373	croak (ERR_NESTING_EXCEEDED);
305		374
306	encode_ch (enc, '['); encode_nl (enc);	375	encode_ch (enc, '[');
307	++enc->indent;	376
		377	if (len >= 0)
		378	{
		379	encode_nl (enc); ++enc->indent;
308		380
309	for (i = 0; i <= len; ++i)	381	for (i = 0; i <= len; ++i)
310	{	382	{
		383	SV **svp = av_fetch (av, i, 0);
		384
311	encode_indent (enc);	385	encode_indent (enc);
312	encode_sv (enc, *av_fetch (av, i, 0));
313		386
		387	if (svp)
		388	encode_sv (enc, *svp);
		389	else
		390	encode_str (enc, "null", 4, 0);
		391
314	if (i < len)	392	if (i < len)
315	encode_comma (enc);	393	encode_comma (enc);
316	}	394	}
317		395
		396	encode_nl (enc); --enc->indent; encode_indent (enc);
		397	}
		398
318	encode_nl (enc);	399	encode_ch (enc, ']');
319
320	--enc->indent;
321	encode_indent (enc); encode_ch (enc, ']');
322	}	400	}
323		401
324	static void	402	static void
325	encode_he (enc_t enc, HE he)	403	encode_hk (enc_t enc, HE he)
326	{	404	{
327	encode_ch (enc, '"');	405	encode_ch (enc, '"');
328		406
329	if (HeKLEN (he) == HEf_SVKEY)	407	if (HeKLEN (he) == HEf_SVKEY)
330	{	408	{
…		…
343	encode_ch (enc, '"');	421	encode_ch (enc, '"');
344		422
345	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);	423	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
346	encode_ch (enc, ':');	424	encode_ch (enc, ':');
347	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);	425	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
348	encode_sv (enc, HeVAL (he));
349	}	426	}
350		427
351	// compare hash entries, used when all keys are bytestrings	428	// compare hash entries, used when all keys are bytestrings
352	static int	429	static int
353	he_cmp_fast (const void a_, const void b_)	430	he_cmp_fast (const void a_, const void b_)
…		…
358	HE b = (HE **)b_;	435	HE b = (HE **)b_;
359		436
360	STRLEN la = HeKLEN (a);	437	STRLEN la = HeKLEN (a);
361	STRLEN lb = HeKLEN (b);	438	STRLEN lb = HeKLEN (b);
362		439
363	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	440	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
364	cmp = la - lb;	441	cmp = lb - la;
365		442
366	return cmp;	443	return cmp;
367	}	444	}
368		445
369	// compare hash entries, used when some keys are sv's or utf-x	446	// compare hash entries, used when some keys are sv's or utf-x
370	static int	447	static int
371	he_cmp_slow (const void a, const void b)	448	he_cmp_slow (const void a, const void b)
372	{	449	{
373	return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));	450	return sv_cmp (HeSVKEY_force ((HE )b), HeSVKEY_force ((HE **)a));
374	}	451	}
375		452
376	static void	453	static void
377	encode_hv (enc_t enc, HV hv)	454	encode_hv (enc_t enc, HV hv)
378	{	455	{
379	int count, i;	456	HE *he;
380		457
381	if (enc->indent >= enc->maxdepth)	458	if (enc->indent >= enc->json.max_depth)
382	croak ("data structure too deep (hit recursion limit)");	459	croak (ERR_NESTING_EXCEEDED);
383		460
384	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	461	encode_ch (enc, '{');
385		462
386	if ((count = hv_iterinit (hv)))
387	{
388	// for canonical output we have to sort by keys first	463	// for canonical output we have to sort by keys first
389	// actually, this is mostly due to the stupid so-called	464	// actually, this is mostly due to the stupid so-called
390	// security workaround added somewhere in 5.8.x.	465	// security workaround added somewhere in 5.8.x
391	// that randomises hash orderings	466	// that randomises hash orderings
392	if (enc->json.flags & F_CANONICAL)	467	if (enc->json.flags & F_CANONICAL)
		468	{
		469	int count = hv_iterinit (hv);
		470
		471	if (SvMAGICAL (hv))
393	{	472	{
		473	// need to count by iterating. could improve by dynamically building the vector below
		474	// but I don't care for the speed of this special case.
		475	// note also that we will run into undefined behaviour when the two iterations
		476	// do not result in the same count, something I might care for in some later release.
		477
		478	count = 0;
		479	while (hv_iternext (hv))
		480	++count;
		481
		482	hv_iterinit (hv);
		483	}
		484
		485	if (count)
		486	{
394	int fast = 1;	487	int i, fast = 1;
395	HE *he;
396	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	488	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
397	HE *hes = _alloca (count sizeof (HE));	489	HE *hes = _alloca (count sizeof (HE));
398	#else	490	#else
399	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode	491	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
400	#endif	492	#endif
…		…
427		519
428	FREETMPS;	520	FREETMPS;
429	LEAVE;	521	LEAVE;
430	}	522	}
431		523
432	for (i = 0; i < count; ++i)	524	encode_nl (enc); ++enc->indent;
		525
		526	while (count--)
433	{	527	{
434	encode_indent (enc);	528	encode_indent (enc);
		529	he = hes [count];
435	encode_he (enc, hes [i]);	530	encode_hk (enc, he);
		531	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
436		532
437	if (i < count - 1)	533	if (count)
438	encode_comma (enc);	534	encode_comma (enc);
439	}	535	}
440		536
441	encode_nl (enc);	537	encode_nl (enc); --enc->indent; encode_indent (enc);
442	}	538	}
		539	}
443	else	540	else
		541	{
		542	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))
		543	if ((he = hv_iternext (hv)))
444	{	544	{
445	HE *he = hv_iternext (hv);	545	encode_nl (enc); ++enc->indent;
446		546
447	for (;;)	547	for (;;)
448	{	548	{
449	encode_indent (enc);	549	encode_indent (enc);
450	encode_he (enc, he);	550	encode_hk (enc, he);
		551	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
451		552
452	if (!(he = hv_iternext (hv)))	553	if (!(he = hv_iternext (hv)))
453	break;	554	break;
454		555
455	encode_comma (enc);	556	encode_comma (enc);
456	}	557	}
457		558
458	encode_nl (enc);	559	encode_nl (enc); --enc->indent; encode_indent (enc);
459	}	560	}
460	}	561	}
461		562
462	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	563	encode_ch (enc, '}');
463	}	564	}
464		565
465	// encode objects, arrays and special \0=false and \1=true values.	566	// encode objects, arrays and special \0=false and \1=true values.
466	static void	567	static void
467	encode_rv (enc_t enc, SV sv)	568	encode_rv (enc_t enc, SV sv)
…		…
471	SvGETMAGIC (sv);	572	SvGETMAGIC (sv);
472	svt = SvTYPE (sv);	573	svt = SvTYPE (sv);
473		574
474	if (expect_false (SvOBJECT (sv)))	575	if (expect_false (SvOBJECT (sv)))
475	{	576	{
		577	HV *stash = !JSON_SLOW \|\| json_boolean_stash
		578	? json_boolean_stash
		579	: gv_stashpv ("JSON::XS::Boolean", 1);
		580
476	if (SvSTASH (sv) == json_boolean_stash)	581	if (SvSTASH (sv) == stash)
477	{	582	{
478	if (SvIV (sv))	583	if (SvIV (sv))
479	encode_str (enc, "true", 4, 0);	584	encode_str (enc, "true", 4, 0);
480	else	585	else
481	encode_str (enc, "false", 5, 0);	586	encode_str (enc, "false", 5, 0);
…		…
499		604
500	ENTER; SAVETMPS; PUSHMARK (SP);	605	ENTER; SAVETMPS; PUSHMARK (SP);
501	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));	606	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
502		607
503	// calling with G_SCALAR ensures that we always get a 1 return value	608	// calling with G_SCALAR ensures that we always get a 1 return value
504	// check anyways.
505	PUTBACK;	609	PUTBACK;
506	assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR));	610	call_sv ((SV *)GvCV (to_json), G_SCALAR);
507	SPAGAIN;	611	SPAGAIN;
508		612
509	// catch this surprisingly common error	613	// catch this surprisingly common error
510	if (SvROK (TOPs) && SvRV (TOPs) == sv)	614	if (SvROK (TOPs) && SvRV (TOPs) == sv)
511	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));	615	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
…		…
541		645
542	if (len == 1 && *pv == '1')	646	if (len == 1 && *pv == '1')
543	encode_str (enc, "true", 4, 0);	647	encode_str (enc, "true", 4, 0);
544	else if (len == 1 && *pv == '0')	648	else if (len == 1 && *pv == '0')
545	encode_str (enc, "false", 5, 0);	649	encode_str (enc, "false", 5, 0);
		650	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		651	encode_str (enc, "null", 4, 0);
546	else	652	else
547	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	653	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
548	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	654	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
549	}	655	}
		656	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		657	encode_str (enc, "null", 4, 0);
550	else	658	else
551	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	659	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
552	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	660	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
553	}	661	}
554		662
…		…
572	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	680	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
573	enc->cur += strlen (enc->cur);	681	enc->cur += strlen (enc->cur);
574	}	682	}
575	else if (SvIOKp (sv))	683	else if (SvIOKp (sv))
576	{	684	{
577	// we assume we can always read an IV as a UV	685	// we assume we can always read an IV as a UV and vice versa
578	if (SvUV (sv) & ~(UV)0x7fff)	686	// we assume two's complement
579	{	687	// we assume no aliasing issues in the union
580	// large integer, use the (rather slow) snprintf way.	688	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
581	need (enc, sizeof (UV) * 3);	689	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
582	enc->cur +=
583	SvIsUV(sv)
584	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
585	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
586	}
587	else
588	{	690	{
589	// optimise the "small number case"	691	// optimise the "small number case"
590	// code will likely be branchless and use only a single multiplication	692	// code will likely be branchless and use only a single multiplication
		693	// works for numbers up to 59074
591	I32 i = SvIV (sv);	694	I32 i = SvIVX (sv);
592	U32 u;	695	U32 u;
593	char digit, nz = 0;	696	char digit, nz = 0;
594		697
595	need (enc, 6);	698	need (enc, 6);
596		699
…		…
602		705
603	// now output digit by digit, each time masking out the integer part	706	// now output digit by digit, each time masking out the integer part
604	// and multiplying by 5 while moving the decimal point one to the right,	707	// and multiplying by 5 while moving the decimal point one to the right,
605	// resulting in a net multiplication by 10.	708	// resulting in a net multiplication by 10.
606	// we always write the digit to memory but conditionally increment	709	// we always write the digit to memory but conditionally increment
607	// the pointer, to ease the usage of conditional move instructions.	710	// the pointer, to enable the use of conditional move instructions.
608	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffff) 5;	711	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffffUL) 5;
609	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffff) 5;	712	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffffUL) 5;
610	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffff) 5;	713	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffffUL) 5;
611	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffff) 5;	714	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffffUL) 5;
612	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'	715	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
613	}	716	}
		717	else
		718	{
		719	// large integer, use the (rather slow) snprintf way.
		720	need (enc, IVUV_MAXCHARS);
		721	enc->cur +=
		722	SvIsUV(sv)
		723	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
		724	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		725	}
614	}	726	}
615	else if (SvROK (sv))	727	else if (SvROK (sv))
616	encode_rv (enc, SvRV (sv));	728	encode_rv (enc, SvRV (sv));
617	else if (!SvOK (sv))	729	else if (!SvOK (sv) \|\| enc->json.flags & F_ALLOW_UNKNOWN)
618	encode_str (enc, "null", 4, 0);	730	encode_str (enc, "null", 4, 0);
619	else	731	else
620	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	732	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
621	SvPV_nolen (sv), SvFLAGS (sv));	733	SvPV_nolen (sv), SvFLAGS (sv));
622	}	734	}
…		…
632	enc.json = *json;	744	enc.json = *json;
633	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	745	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
634	enc.cur = SvPVX (enc.sv);	746	enc.cur = SvPVX (enc.sv);
635	enc.end = SvEND (enc.sv);	747	enc.end = SvEND (enc.sv);
636	enc.indent = 0;	748	enc.indent = 0;
637	enc.maxdepth = DEC_DEPTH (enc.json.flags);	749	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		750	: enc.json.flags & F_LATIN1 ? 0x000100UL
		751	: 0x110000UL;
638		752
639	SvPOK_only (enc.sv);	753	SvPOK_only (enc.sv);
640	encode_sv (&enc, scalar);	754	encode_sv (&enc, scalar);
641		755
642	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	756	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
…		…
663	JSON json;	777	JSON json;
664	U32 depth; // recursion depth	778	U32 depth; // recursion depth
665	U32 maxdepth; // recursion depth limit	779	U32 maxdepth; // recursion depth limit
666	} dec_t;	780	} dec_t;
667		781
668	inline void	782	INLINE void
		783	decode_comment (dec_t *dec)
		784	{
		785	// only '#'-style comments allowed a.t.m.
		786
		787	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)
		788	++dec->cur;
		789	}
		790
		791	INLINE void
669	decode_ws (dec_t *dec)	792	decode_ws (dec_t *dec)
670	{	793	{
671	for (;;)	794	for (;;)
672	{	795	{
673	char ch = *dec->cur;	796	char ch = *dec->cur;
674		797
675	if (ch > 0x20	798	if (ch > 0x20)
		799	{
		800	if (expect_false (ch == '#'))
		801	{
		802	if (dec->json.flags & F_RELAXED)
		803	decode_comment (dec);
		804	else
		805	break;
		806	}
		807	else
		808	break;
		809	}
676	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	810	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
677	break;	811	break; // parse error, but let higher level handle it, gives better error messages
678		812
679	++dec->cur;	813	++dec->cur;
680	}	814	}
681	}	815	}
682		816
…		…
686	if (*dec->cur != ch) \	820	if (*dec->cur != ch) \
687	ERR (# ch " expected"); \	821	ERR (# ch " expected"); \
688	++dec->cur; \	822	++dec->cur; \
689	SE	823	SE
690		824
691	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	825	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
692	#define DEC_DEC_DEPTH --dec->depth	826	#define DEC_DEC_DEPTH --dec->depth
693		827
694	static SV decode_sv (dec_t dec);	828	static SV decode_sv (dec_t dec);
695		829
696	static signed char decode_hexdigit[256];	830	static signed char decode_hexdigit[256];
…		…
788		922
789	if (hi >= 0x80)	923	if (hi >= 0x80)
790	{	924	{
791	utf8 = 1;	925	utf8 = 1;
792		926
793	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	927	cur = encode_utf8 (cur, hi);
794	}	928	}
795	else	929	else
796	*cur++ = hi;	930	*cur++ = hi;
797	}	931	}
798	break;	932	break;
…		…
800	default:	934	default:
801	--dec_cur;	935	--dec_cur;
802	ERR ("illegal backslash escape sequence in string");	936	ERR ("illegal backslash escape sequence in string");
803	}	937	}
804	}	938	}
805	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	939	else if (expect_true (ch >= 0x20 && ch < 0x80))
806	*cur++ = ch;	940	*cur++ = ch;
807	else if (ch >= 0x80)	941	else if (ch >= 0x80)
808	{	942	{
809	STRLEN clen;	943	STRLEN clen;
810	UV uch;	944	UV uch;
…		…
931	is_nv = 1;	1065	is_nv = 1;
932	}	1066	}
933		1067
934	if (!is_nv)	1068	if (!is_nv)
935	{	1069	{
		1070	int len = dec->cur - start;
		1071
936	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1072	// special case the rather common 1..5-digit-int case
937	if (*start == '-')	1073	if (*start == '-')
938	switch (dec->cur - start)	1074	switch (len)
939	{	1075	{
940	case 2: return newSViv (-( start [1] - '0' * 1));	1076	case 2: return newSViv (-( start [1] - '0' * 1));
941	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1077	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
942	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));	1078	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
943	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));	1079	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1080	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
944	}	1081	}
945	else	1082	else
946	switch (dec->cur - start)	1083	switch (len)
947	{	1084	{
948	case 1: return newSViv ( start [0] - '0' * 1);	1085	case 1: return newSViv ( start [0] - '0' * 1);
949	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1086	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
950	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);	1087	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
951	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);	1088	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1089	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
952	}	1090	}
953		1091
954	{	1092	{
955	UV uv;	1093	UV uv;
956	int numtype = grok_number (start, dec->cur - start, &uv);	1094	int numtype = grok_number (start, len, &uv);
957	if (numtype & IS_NUMBER_IN_UV)	1095	if (numtype & IS_NUMBER_IN_UV)
958	if (numtype & IS_NUMBER_NEG)	1096	if (numtype & IS_NUMBER_NEG)
959	{	1097	{
960	if (uv < (UV)IV_MIN)	1098	if (uv < (UV)IV_MIN)
961	return newSViv (-(IV)uv);	1099	return newSViv (-(IV)uv);
962	}	1100	}
963	else	1101	else
964	return newSVuv (uv);	1102	return newSVuv (uv);
965
966	// here would likely be the place for bigint support
967	}	1103	}
968	}
969		1104
970	// if we ever support bigint or bigfloat, this is the place for bigfloat	1105	len -= *start == '-' ? 1 : 0;
		1106
		1107	// does not fit into IV or UV, try NV
		1108	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1109	#if defined (LDBL_DIG)
		1110	\|\| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1111	#endif
		1112	)
		1113	// fits into NV without loss of precision
		1114	return newSVnv (Atof (start));
		1115
		1116	// everything else fails, convert it to a string
		1117	return newSVpvn (start, dec->cur - start);
		1118	}
		1119
		1120	// loss of precision here
971	return newSVnv (Atof (start));	1121	return newSVnv (Atof (start));
972		1122
973	fail:	1123	fail:
974	return 0;	1124	return 0;
975	}	1125	}
…		…
1005		1155
1006	if (*dec->cur != ',')	1156	if (*dec->cur != ',')
1007	ERR (", or ] expected while parsing array");	1157	ERR (", or ] expected while parsing array");
1008		1158
1009	++dec->cur;	1159	++dec->cur;
		1160
		1161	decode_ws (dec);
		1162
		1163	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1164	{
		1165	++dec->cur;
		1166	break;
		1167	}
1010	}	1168	}
1011		1169
1012	DEC_DEC_DEPTH;	1170	DEC_DEC_DEPTH;
1013	return newRV_noinc ((SV *)av);	1171	return newRV_noinc ((SV *)av);
1014		1172
…		…
1030	if (*dec->cur == '}')	1188	if (*dec->cur == '}')
1031	++dec->cur;	1189	++dec->cur;
1032	else	1190	else
1033	for (;;)	1191	for (;;)
1034	{	1192	{
1035	decode_ws (dec); EXPECT_CH ('"');	1193	EXPECT_CH ('"');
1036		1194
1037	// heuristic: assume that	1195	// heuristic: assume that
1038	// a) decode_str + hv_store_ent are abysmally slow.	1196	// a) decode_str + hv_store_ent are abysmally slow.
1039	// b) most hash keys are short, simple ascii text.	1197	// b) most hash keys are short, simple ascii text.
1040	// => try to "fast-match" such strings to avoid	1198	// => try to "fast-match" such strings to avoid
…		…
1044	char *p = dec->cur;	1202	char *p = dec->cur;
1045	char *e = p + 24; // only try up to 24 bytes	1203	char *e = p + 24; // only try up to 24 bytes
1046		1204
1047	for (;;)	1205	for (;;)
1048	{	1206	{
1049	// the >= 0x80 is true on most architectures	1207	// the >= 0x80 is false on most architectures
1050	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')	1208	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
1051	{	1209	{
1052	// slow path, back up and use decode_str	1210	// slow path, back up and use decode_str
1053	SV *key = decode_str (dec);	1211	SV *key = decode_str (dec);
1054	if (!key)	1212	if (!key)
1055	goto fail;	1213	goto fail;
1056		1214
1057	decode_ws (dec); EXPECT_CH (':');	1215	decode_ws (dec); EXPECT_CH (':');
1058		1216
		1217	decode_ws (dec);
1059	value = decode_sv (dec);	1218	value = decode_sv (dec);
1060	if (!value)	1219	if (!value)
1061	{	1220	{
1062	SvREFCNT_dec (key);	1221	SvREFCNT_dec (key);
1063	goto fail;	1222	goto fail;
…		…
1075	int len = p - key;	1234	int len = p - key;
1076	dec->cur = p + 1;	1235	dec->cur = p + 1;
1077		1236
1078	decode_ws (dec); EXPECT_CH (':');	1237	decode_ws (dec); EXPECT_CH (':');
1079		1238
		1239	decode_ws (dec);
1080	value = decode_sv (dec);	1240	value = decode_sv (dec);
1081	if (!value)	1241	if (!value)
1082	goto fail;	1242	goto fail;
1083		1243
1084	hv_store (hv, key, len, value, 0);	1244	hv_store (hv, key, len, value, 0);
…		…
1100		1260
1101	if (*dec->cur != ',')	1261	if (*dec->cur != ',')
1102	ERR (", or } expected while parsing object/hash");	1262	ERR (", or } expected while parsing object/hash");
1103		1263
1104	++dec->cur;	1264	++dec->cur;
		1265
		1266	decode_ws (dec);
		1267
		1268	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1269	{
		1270	++dec->cur;
		1271	break;
		1272	}
1105	}	1273	}
1106		1274
1107	DEC_DEC_DEPTH;	1275	DEC_DEC_DEPTH;
1108	sv = newRV_noinc ((SV *)hv);	1276	sv = newRV_noinc ((SV *)hv);
1109		1277
…		…
1174	}	1342	}
1175		1343
1176	static SV *	1344	static SV *
1177	decode_sv (dec_t *dec)	1345	decode_sv (dec_t *dec)
1178	{	1346	{
1179	decode_ws (dec);
1180
1181	// the beauty of JSON: you need exactly one character lookahead	1347	// the beauty of JSON: you need exactly one character lookahead
1182	// to parse anything.	1348	// to parse everything.
1183	switch (*dec->cur)	1349	switch (*dec->cur)
1184	{	1350	{
1185	case '"': ++dec->cur; return decode_str (dec);	1351	case '"': ++dec->cur; return decode_str (dec);
1186	case '[': ++dec->cur; return decode_av (dec);	1352	case '[': ++dec->cur; return decode_av (dec);
1187	case '{': ++dec->cur; return decode_hv (dec);	1353	case '{': ++dec->cur; return decode_hv (dec);
1188		1354
1189	case '-':	1355	case '-':
1190	case '0': case '1': case '2': case '3': case '4':	1356	case '0': case '1': case '2': case '3': case '4':
1191	case '5': case '6': case '7': case '8': case '9':	1357	case '5': case '6': case '7': case '8': case '9':
1192	return decode_num (dec);	1358	return decode_num (dec);
1193		1359
1194	case 't':	1360	case 't':
1195	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1361	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1196	{	1362	{
1197	dec->cur += 4;	1363	dec->cur += 4;
		1364	#if JSON_SLOW
		1365	json_true = get_bool ("JSON::XS::true");
		1366	#endif
1198	return SvREFCNT_inc (json_true);	1367	return newSVsv (json_true);
1199	}	1368	}
1200	else	1369	else
1201	ERR ("'true' expected");	1370	ERR ("'true' expected");
1202		1371
1203	break;	1372	break;
1204		1373
1205	case 'f':	1374	case 'f':
1206	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1375	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1207	{	1376	{
1208	dec->cur += 5;	1377	dec->cur += 5;
		1378	#if JSON_SLOW
		1379	json_false = get_bool ("JSON::XS::false");
		1380	#endif
1209	return SvREFCNT_inc (json_false);	1381	return newSVsv (json_false);
1210	}	1382	}
1211	else	1383	else
1212	ERR ("'false' expected");	1384	ERR ("'false' expected");
1213		1385
1214	break;	1386	break;
…		…
1232	fail:	1404	fail:
1233	return 0;	1405	return 0;
1234	}	1406	}
1235		1407
1236	static SV *	1408	static SV *
1237	decode_json (SV string, JSON json, UV *offset_return)	1409	decode_json (SV string, JSON json, STRLEN *offset_return)
1238	{	1410	{
1239	dec_t dec;	1411	dec_t dec;
1240	UV offset;	1412	STRLEN offset;
1241	SV *sv;	1413	SV *sv;
1242		1414
1243	SvGETMAGIC (string);	1415	SvGETMAGIC (string);
1244	SvUPGRADE (string, SVt_PV);	1416	SvUPGRADE (string, SVt_PV);
1245		1417
1246	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))	1418	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1419	* assertion with -DDEBUGGING, although SvCUR is documented to
		1420	* return the xpv_cur field which certainly exists after upgrading.
		1421	* according to nicholas clark, calling SvPOK fixes this.
		1422	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1423	* and hope for the best.
		1424	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1425	* assertion business is seriously broken, try yet another workaround
		1426	* for the broken -DDEBUGGING.
		1427	*/
		1428	#ifdef DEBUGGING
		1429	offset = SvOK (string) ? sv_len (string) : 0;
		1430	#else
		1431	offset = SvCUR (string);
		1432	#endif
		1433
		1434	if (offset > json->max_size && json->max_size)
1247	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",	1435	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1248	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));	1436	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
1249		1437
1250	if (json->flags & F_UTF8)	1438	if (json->flags & F_UTF8)
1251	sv_utf8_downgrade (string, 0);	1439	sv_utf8_downgrade (string, 0);
1252	else	1440	else
1253	sv_utf8_upgrade (string);	1441	sv_utf8_upgrade (string);
1254		1442
1255	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1443	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1256		1444
1257	dec.json = *json;	1445	dec.json = *json;
1258	dec.cur = SvPVX (string);	1446	dec.cur = SvPVX (string);
1259	dec.end = SvEND (string);	1447	dec.end = SvEND (string);
1260	dec.err = 0;	1448	dec.err = 0;
1261	dec.depth = 0;	1449	dec.depth = 0;
1262	dec.maxdepth = DEC_DEPTH (dec.json.flags);
1263		1450
1264	if (dec.json.cb_object \|\| dec.json.cb_sk_object)	1451	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
1265	dec.json.flags \|= F_HOOK;	1452	dec.json.flags \|= F_HOOK;
1266		1453
1267	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1454	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1455
		1456	decode_ws (&dec);
1268	sv = decode_sv (&dec);	1457	sv = decode_sv (&dec);
1269		1458
1270	if (!(offset_return \|\| !sv))	1459	if (!(offset_return \|\| !sv))
1271	{	1460	{
1272	// check for trailing garbage	1461	// check for trailing garbage
…		…
1316		1505
1317	return sv;	1506	return sv;
1318	}	1507	}
1319		1508
1320	/////////////////////////////////////////////////////////////////////////////	1509	/////////////////////////////////////////////////////////////////////////////
		1510	// incremental parser
		1511
		1512	static void
		1513	incr_parse (JSON *self)
		1514	{
		1515	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1516
		1517	for (;;)
		1518	{
		1519	//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
		1520	switch (self->incr_mode)
		1521	{
		1522	// only used for intiial whitespace skipping
		1523	case INCR_M_WS:
		1524	for (;;)
		1525	{
		1526	if (*p > 0x20)
		1527	{
		1528	self->incr_mode = INCR_M_JSON;
		1529	goto incr_m_json;
		1530	}
		1531	else if (!*p)
		1532	goto interrupt;
		1533
		1534	++p;
		1535	}
		1536
		1537	// skip a single char inside a string (for \\-processing)
		1538	case INCR_M_BS:
		1539	if (!*p)
		1540	goto interrupt;
		1541
		1542	++p;
		1543	self->incr_mode = INCR_M_STR;
		1544	goto incr_m_str;
		1545
		1546	// inside a string
		1547	case INCR_M_STR:
		1548	incr_m_str:
		1549	for (;;)
		1550	{
		1551	if (*p == '"')
		1552	{
		1553	++p;
		1554	self->incr_mode = INCR_M_JSON;
		1555
		1556	if (!self->incr_nest)
		1557	goto interrupt;
		1558
		1559	goto incr_m_json;
		1560	}
		1561	else if (*p == '\\')
		1562	{
		1563	++p; // "virtually" consumes character after \
		1564
		1565	if (!*p) // if at end of string we have to switch modes
		1566	{
		1567	self->incr_mode = INCR_M_BS;
		1568	goto interrupt;
		1569	}
		1570	}
		1571	else if (!*p)
		1572	goto interrupt;
		1573
		1574	++p;
		1575	}
		1576
		1577	// after initial ws, outside string
		1578	case INCR_M_JSON:
		1579	incr_m_json:
		1580	for (;;)
		1581	{
		1582	switch (*p++)
		1583	{
		1584	case 0:
		1585	--p;
		1586	goto interrupt;
		1587
		1588	case 0x09:
		1589	case 0x0a:
		1590	case 0x0d:
		1591	case 0x20:
		1592	if (!self->incr_nest)
		1593	{
		1594	--p; // do not eat the whitespace, let the next round do it
		1595	goto interrupt;
		1596	}
		1597	break;
		1598
		1599	case '"':
		1600	self->incr_mode = INCR_M_STR;
		1601	goto incr_m_str;
		1602
		1603	case '[':
		1604	case '{':
		1605	if (++self->incr_nest > self->max_depth)
		1606	croak (ERR_NESTING_EXCEEDED);
		1607	break;
		1608
		1609	case ']':
		1610	case '}':
		1611	if (--self->incr_nest <= 0)
		1612	goto interrupt;
		1613	}
		1614	}
		1615	}
		1616
		1617	modechange:
		1618	;
		1619	}
		1620
		1621	interrupt:
		1622	self->incr_pos = p - SvPVX (self->incr_text);
		1623	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
		1624	}
		1625
		1626	/////////////////////////////////////////////////////////////////////////////
1321	// XS interface functions	1627	// XS interface functions
1322		1628
1323	MODULE = JSON::XS PACKAGE = JSON::XS	1629	MODULE = JSON::XS PACKAGE = JSON::XS
1324		1630
1325	BOOT:	1631	BOOT:
…		…
1334	: -1;	1640	: -1;
1335		1641
1336	json_stash = gv_stashpv ("JSON::XS" , 1);	1642	json_stash = gv_stashpv ("JSON::XS" , 1);
1337	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);	1643	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1338		1644
1339	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );	1645	json_true = get_bool ("JSON::XS::true");
1340	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1646	json_false = get_bool ("JSON::XS::false");
1341	}	1647	}
1342		1648
1343	PROTOTYPES: DISABLE	1649	PROTOTYPES: DISABLE
		1650
		1651	void CLONE (...)
		1652	CODE:
		1653	json_stash = 0;
		1654	json_boolean_stash = 0;
1344		1655
1345	void new (char *klass)	1656	void new (char *klass)
1346	PPCODE:	1657	PPCODE:
1347	{	1658	{
1348	SV *pv = NEWSV (0, sizeof (JSON));	1659	SV *pv = NEWSV (0, sizeof (JSON));
1349	SvPOK_only (pv);	1660	SvPOK_only (pv);
1350	Zero (SvPVX (pv), 1, JSON);	1661	json_init ((JSON *)SvPVX (pv));
1351	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	1662	XPUSHs (sv_2mortal (sv_bless (
1352	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash)));	1663	newRV_noinc (pv),
		1664	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1665	)));
1353	}	1666	}
1354		1667
1355	void ascii (JSON *self, int enable = 1)	1668	void ascii (JSON *self, int enable = 1)
1356	ALIAS:	1669	ALIAS:
1357	ascii = F_ASCII	1670	ascii = F_ASCII
…		…
1364	pretty = F_PRETTY	1677	pretty = F_PRETTY
1365	allow_nonref = F_ALLOW_NONREF	1678	allow_nonref = F_ALLOW_NONREF
1366	shrink = F_SHRINK	1679	shrink = F_SHRINK
1367	allow_blessed = F_ALLOW_BLESSED	1680	allow_blessed = F_ALLOW_BLESSED
1368	convert_blessed = F_CONV_BLESSED	1681	convert_blessed = F_CONV_BLESSED
		1682	relaxed = F_RELAXED
		1683	allow_unknown = F_ALLOW_UNKNOWN
1369	PPCODE:	1684	PPCODE:
1370	{	1685	{
1371	if (enable)	1686	if (enable)
1372	self->flags \|= ix;	1687	self->flags \|= ix;
1373	else	1688	else
1374	self->flags &= ~ix;	1689	self->flags &= ~ix;
1375		1690
1376	XPUSHs (ST (0));	1691	XPUSHs (ST (0));
1377	}	1692	}
1378		1693
1379	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1694	void get_ascii (JSON *self)
		1695	ALIAS:
		1696	get_ascii = F_ASCII
		1697	get_latin1 = F_LATIN1
		1698	get_utf8 = F_UTF8
		1699	get_indent = F_INDENT
		1700	get_canonical = F_CANONICAL
		1701	get_space_before = F_SPACE_BEFORE
		1702	get_space_after = F_SPACE_AFTER
		1703	get_allow_nonref = F_ALLOW_NONREF
		1704	get_shrink = F_SHRINK
		1705	get_allow_blessed = F_ALLOW_BLESSED
		1706	get_convert_blessed = F_CONV_BLESSED
		1707	get_relaxed = F_RELAXED
		1708	get_allow_unknown = F_ALLOW_UNKNOWN
1380	PPCODE:	1709	PPCODE:
1381	{	1710	XPUSHs (boolSV (self->flags & ix));
1382	UV log2 = 0;
1383		1711
1384	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	1712	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1385		1713	PPCODE:
1386	while ((1UL << log2) < max_depth)	1714	self->max_depth = max_depth;
1387	++log2;
1388
1389	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1390
1391	XPUSHs (ST (0));	1715	XPUSHs (ST (0));
1392	}
1393		1716
		1717	U32 get_max_depth (JSON *self)
		1718	CODE:
		1719	RETVAL = self->max_depth;
		1720	OUTPUT:
		1721	RETVAL
		1722
1394	void max_size (JSON *self, UV max_size = 0)	1723	void max_size (JSON *self, U32 max_size = 0)
1395	PPCODE:	1724	PPCODE:
1396	{	1725	self->max_size = max_size;
1397	UV log2 = 0;
1398
1399	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1400	if (max_size == 1) max_size = 2;
1401
1402	while ((1UL << log2) < max_size)
1403	++log2;
1404
1405	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
1406
1407	XPUSHs (ST (0));	1726	XPUSHs (ST (0));
1408	}	1727
		1728	int get_max_size (JSON *self)
		1729	CODE:
		1730	RETVAL = self->max_size;
		1731	OUTPUT:
		1732	RETVAL
1409		1733
1410	void filter_json_object (JSON self, SV cb = &PL_sv_undef)	1734	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1411	PPCODE:	1735	PPCODE:
1412	{	1736	{
1413	SvREFCNT_dec (self->cb_object);	1737	SvREFCNT_dec (self->cb_object);
…		…
1447	XPUSHs (decode_json (jsonstr, self, 0));	1771	XPUSHs (decode_json (jsonstr, self, 0));
1448		1772
1449	void decode_prefix (JSON self, SV jsonstr)	1773	void decode_prefix (JSON self, SV jsonstr)
1450	PPCODE:	1774	PPCODE:
1451	{	1775	{
1452	UV offset;	1776	STRLEN offset;
1453	EXTEND (SP, 2);	1777	EXTEND (SP, 2);
1454	PUSHs (decode_json (jsonstr, self, &offset));	1778	PUSHs (decode_json (jsonstr, self, &offset));
1455	PUSHs (sv_2mortal (newSVuv (offset)));	1779	PUSHs (sv_2mortal (newSVuv (offset)));
		1780	}
		1781
		1782	void incr_parse (JSON self, SV jsonstr = 0)
		1783	PPCODE:
		1784	{
		1785	if (!self->incr_text)
		1786	self->incr_text = newSVpvn ("", 0);
		1787
		1788	// append data, if any
		1789	if (jsonstr)
		1790	{
		1791	if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text))
		1792	{
		1793	/* utf-8-ness differs, need to upgrade */
		1794	sv_utf8_upgrade (self->incr_text);
		1795
		1796	if (self->incr_pos)
		1797	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1798	- (U8 *)SvPVX (self->incr_text);
		1799	}
		1800
		1801	{
		1802	STRLEN len;
		1803	const char *str = SvPV (jsonstr, len);
		1804	SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1);
		1805	Move (str, SvEND (self->incr_text), len, char);
		1806	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1807	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1808	}
		1809	}
		1810
		1811	if (GIMME_V != G_VOID)
		1812	do
		1813	{
		1814	STRLEN offset;
		1815
		1816	if (!INCR_DONE (self))
		1817	{
		1818	incr_parse (self);
		1819
		1820	if (self->incr_pos > self->max_size && self->max_size)
		1821	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1822	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1823
		1824	if (!INCR_DONE (self))
		1825	break;
		1826	}
		1827
		1828	XPUSHs (decode_json (self->incr_text, self, &offset));
		1829
		1830	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
		1831	self->incr_pos -= offset;
		1832	self->incr_nest = 0;
		1833	self->incr_mode = 0;
		1834	}
		1835	while (GIMME_V == G_ARRAY);
		1836	}
		1837
		1838	SV incr_text (JSON self)
		1839	ATTRS: lvalue
		1840	CODE:
		1841	{
		1842	if (self->incr_pos)
		1843	croak ("incr_text can not be called when the incremental parser already started parsing");
		1844
		1845	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		1846	}
		1847	OUTPUT:
		1848	RETVAL
		1849
		1850	void incr_skip (JSON *self)
		1851	CODE:
		1852	{
		1853	if (self->incr_pos)
		1854	{
		1855	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1856	self->incr_pos = 0;
		1857	self->incr_nest = 0;
		1858	self->incr_mode = 0;
		1859	}
		1860	}
		1861
		1862	void incr_reset (JSON *self)
		1863	CODE:
		1864	{
		1865	SvREFCNT_dec (self->incr_text);
		1866	self->incr_text = 0;
		1867	self->incr_pos = 0;
		1868	self->incr_nest = 0;
		1869	self->incr_mode = 0;
1456	}	1870	}
1457		1871
1458	void DESTROY (JSON *self)	1872	void DESTROY (JSON *self)
1459	CODE:	1873	CODE:
1460	SvREFCNT_dec (self->cb_sk_object);	1874	SvREFCNT_dec (self->cb_sk_object);
1461	SvREFCNT_dec (self->cb_object);	1875	SvREFCNT_dec (self->cb_object);
		1876	SvREFCNT_dec (self->incr_text);
1462		1877
1463	PROTOTYPES: ENABLE	1878	PROTOTYPES: ENABLE
1464		1879
1465	void to_json (SV *scalar)	1880	void encode_json (SV *scalar)
		1881	ALIAS:
		1882	to_json_ = 0
		1883	encode_json = F_UTF8
1466	PPCODE:	1884	PPCODE:
1467	{	1885	{
1468	JSON json = { F_DEFAULT \| F_UTF8 };	1886	JSON json;
		1887	json_init (&json);
		1888	json.flags \|= ix;
1469	XPUSHs (encode_json (scalar, &json));	1889	XPUSHs (encode_json (scalar, &json));
1470	}	1890	}
1471		1891
1472	void from_json (SV *jsonstr)	1892	void decode_json (SV *jsonstr)
		1893	ALIAS:
		1894	from_json_ = 0
		1895	decode_json = F_UTF8
1473	PPCODE:	1896	PPCODE:
1474	{	1897	{
1475	JSON json = { F_DEFAULT \| F_UTF8 };	1898	JSON json;
		1899	json_init (&json);
		1900	json.flags \|= ix;
1476	XPUSHs (decode_json (jsonstr, &json, 0));	1901	XPUSHs (decode_json (jsonstr, &json, 0));
1477	}	1902	}
1478		1903
		1904

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.53 by root, Tue Jul 10 15:45:34 2007 UTC vs. Revision 1.92 by root, Mon Sep 22 07:29:29 2008 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.53 by root, Tue Jul 10 15:45:34 2007 UTC vs.
Revision 1.92 by root, Mon Sep 22 07:29:29 2008 UTC