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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.63 by root, Mon Aug 27 01:49:01 2007 UTC vs.
Revision 1.108 by root, Tue Aug 17 23:27:36 2010 UTC

…		…
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>
9	#include <float.h>	10	#include <float.h>
10		11
11	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	12	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
12	# define snprintf _snprintf // C compilers have this in stdio.h	13	# define snprintf _snprintf // C compilers have this in stdio.h
13	#endif	14	#endif
14		15
15	// 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
16	// guarentees, though. if it breaks, you get to keep the pieces.	17	// guarantees, though. if it breaks, you get to keep the pieces.
17	#ifndef UTF8_MAXBYTES	18	#ifndef UTF8_MAXBYTES
18	# define UTF8_MAXBYTES 13	19	# define UTF8_MAXBYTES 13
19	#endif	20	#endif
		21
		22	// three extra for rounding, sign, and end of string
		23	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
20		24
21	#define F_ASCII 0x00000001UL	25	#define F_ASCII 0x00000001UL
22	#define F_LATIN1 0x00000002UL	26	#define F_LATIN1 0x00000002UL
23	#define F_UTF8 0x00000004UL	27	#define F_UTF8 0x00000004UL
24	#define F_INDENT 0x00000008UL	28	#define F_INDENT 0x00000008UL
…		…
28	#define F_ALLOW_NONREF 0x00000100UL	32	#define F_ALLOW_NONREF 0x00000100UL
29	#define F_SHRINK 0x00000200UL	33	#define F_SHRINK 0x00000200UL
30	#define F_ALLOW_BLESSED 0x00000400UL	34	#define F_ALLOW_BLESSED 0x00000400UL
31	#define F_CONV_BLESSED 0x00000800UL	35	#define F_CONV_BLESSED 0x00000800UL
32	#define F_RELAXED 0x00001000UL	36	#define F_RELAXED 0x00001000UL
33		37	#define F_ALLOW_UNKNOWN 0x00002000UL
34	#define F_MAXDEPTH 0xf8000000UL
35	#define S_MAXDEPTH 27
36	#define F_MAXSIZE 0x01f00000UL
37	#define S_MAXSIZE 20
38	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing	38	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
39		39
40	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
41	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
42
43	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER	40	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER
44	#define F_DEFAULT (9UL << S_MAXDEPTH)
45		41
46	#define INIT_SIZE 32 // initial scalar size to be allocated	42	#define INIT_SIZE 32 // initial scalar size to be allocated
47	#define INDENT_STEP 3 // spaces per indentation level	43	#define INDENT_STEP 3 // spaces per indentation level
48		44
49	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size	45	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
50		46
		47	#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
		48
51	#define SB do {	49	#define SB do {
52	#define SE } while (0)	50	#define SE } while (0)
53		51
54	#if __GNUC__ >= 3	52	#if __GNUC__ >= 3
55	# define expect(expr,value) __builtin_expect ((expr),(value))	53	# define expect(expr,value) __builtin_expect ((expr), (value))
56	# define inline inline	54	# define INLINE static inline
57	#else	55	#else
58	# define expect(expr,value) (expr)	56	# define expect(expr,value) (expr)
59	# define inline static	57	# define INLINE static
60	#endif	58	#endif
61		59
62	#define expect_false(expr) expect ((expr) != 0, 0)	60	#define expect_false(expr) expect ((expr) != 0, 0)
63	#define expect_true(expr) expect ((expr) != 0, 1)	61	#define expect_true(expr) expect ((expr) != 0, 1)
		62
		63	#define IN_RANGE_INC(type,val,beg,end) \
		64	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		65	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		66
		67	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
64		68
65	#ifdef USE_ITHREADS	69	#ifdef USE_ITHREADS
66	# define JSON_SLOW 1	70	# define JSON_SLOW 1
67	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))	71	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
68	#else	72	#else
…		…
71	#endif	75	#endif
72		76
73	static HV json_stash, json_boolean_stash; // JSON::XS::	77	static HV json_stash, json_boolean_stash; // JSON::XS::
74	static SV json_true, json_false;	78	static SV json_true, json_false;
75		79
		80	enum {
		81	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		82	INCR_M_STR, // inside string
		83	INCR_M_BS, // inside backslash
		84	INCR_M_C0, // inside comment in initial whitespace sequence
		85	INCR_M_C1, // inside comment in other places
		86	INCR_M_JSON // outside anything, count nesting
		87	};
		88
		89	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		90
76	typedef struct {	91	typedef struct {
77	U32 flags;	92	U32 flags;
		93	U32 max_depth;
		94	STRLEN max_size;
		95
78	SV *cb_object;	96	SV *cb_object;
79	HV *cb_sk_object;	97	HV *cb_sk_object;
		98
		99	// for the incremental parser
		100	SV *incr_text; // the source text so far
		101	STRLEN incr_pos; // the current offset into the text
		102	int incr_nest; // {[]}-nesting level
		103	unsigned char incr_mode;
80	} JSON;	104	} JSON;
		105
		106	INLINE void
		107	json_init (JSON *json)
		108	{
		109	Zero (json, 1, JSON);
		110	json->max_depth = 512;
		111	}
81		112
82	/////////////////////////////////////////////////////////////////////////////	113	/////////////////////////////////////////////////////////////////////////////
83	// utility functions	114	// utility functions
84		115
85	inline void	116	INLINE SV *
		117	get_bool (const char *name)
		118	{
		119	SV *sv = get_sv (name, 1);
		120
		121	SvREADONLY_on (sv);
		122	SvREADONLY_on (SvRV (sv));
		123
		124	return sv;
		125	}
		126
		127	INLINE void
86	shrink (SV *sv)	128	shrink (SV *sv)
87	{	129	{
88	sv_utf8_downgrade (sv, 1);	130	sv_utf8_downgrade (sv, 1);
		131
89	if (SvLEN (sv) > SvCUR (sv) + 1)	132	if (SvLEN (sv) > SvCUR (sv) + 1)
90	{	133	{
91	#ifdef SvPV_shrink_to_cur	134	#ifdef SvPV_shrink_to_cur
92	SvPV_shrink_to_cur (sv);	135	SvPV_shrink_to_cur (sv);
93	#elif defined (SvPV_renew)	136	#elif defined (SvPV_renew)
…		…
99	// decode an utf-8 character and return it, or (UV)-1 in	142	// decode an utf-8 character and return it, or (UV)-1 in
100	// case of an error.	143	// case of an error.
101	// we special-case "safe" characters from U+80 .. U+7FF,	144	// we special-case "safe" characters from U+80 .. U+7FF,
102	// but use the very good perl function to parse anything else.	145	// but use the very good perl function to parse anything else.
103	// note that we never call this function for a ascii codepoints	146	// note that we never call this function for a ascii codepoints
104	inline UV	147	INLINE UV
105	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	148	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
106	{	149	{
107	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))	150	if (expect_true (len >= 2
108	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	151	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
109	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	152	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
110	{	153	{
111	*clen = 2;	154	*clen = 2;
112	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);	155	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
113	}	156	}
114	else	157	else
115	{	158	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
116	*clen = (STRLEN)-1;	159	}
117	return (UV)-1;	160
		161	// likewise for encoding, also never called for ascii codepoints
		162	// this function takes advantage of this fact, although current gccs
		163	// seem to optimise the check for >= 0x80 away anyways
		164	INLINE unsigned char *
		165	encode_utf8 (unsigned char *s, UV ch)
		166	{
		167	if (expect_false (ch < 0x000080))
		168	*s++ = ch;
		169	else if (expect_true (ch < 0x000800))
		170	*s++ = 0xc0 \| ( ch >> 6),
		171	*s++ = 0x80 \| ( ch & 0x3f);
		172	else if ( ch < 0x010000)
		173	*s++ = 0xe0 \| ( ch >> 12),
		174	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		175	*s++ = 0x80 \| ( ch & 0x3f);
		176	else if ( ch < 0x110000)
		177	*s++ = 0xf0 \| ( ch >> 18),
		178	*s++ = 0x80 \| ((ch >> 12) & 0x3f),
		179	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		180	*s++ = 0x80 \| ( ch & 0x3f);
		181
		182	return s;
		183	}
		184
		185	// convert offset pointer to character index, sv must be string
		186	static STRLEN
		187	ptr_to_index (SV sv, char offset)
		188	{
		189	return SvUTF8 (sv)
		190	? utf8_distance (offset, SvPVX (sv))
		191	: offset - SvPVX (sv);
		192	}
		193
		194	/////////////////////////////////////////////////////////////////////////////
		195	// fp hell
		196
		197	// scan a group of digits, and a trailing exponent
		198	static void
		199	json_atof_scan1 (const char s, NV accum, int *expo, int postdp, int maxdepth)
		200	{
		201	UV uaccum = 0;
		202	int eaccum = 0;
		203
		204	// if we recurse too deep, skip all remaining digits
		205	// to avoid a stack overflow attack
		206	if (expect_false (--maxdepth <= 0))
		207	while (((U8)*s - '0') < 10)
		208	++s;
		209
		210	for (;;)
118	}	211	{
119	}	212	U8 dig = (U8)*s - '0';
120		213
		214	if (expect_false (dig >= 10))
		215	{
		216	if (dig == (U8)((U8)'.' - (U8)'0'))
		217	{
		218	++s;
		219	json_atof_scan1 (s, accum, expo, 1, maxdepth);
		220	}
		221	else if ((dig \| ' ') == 'e' - '0')
		222	{
		223	int exp2 = 0;
		224	int neg = 0;
		225
		226	++s;
		227
		228	if (*s == '-')
		229	{
		230	++s;
		231	neg = 1;
		232	}
		233	else if (*s == '+')
		234	++s;
		235
		236	while ((dig = (U8)*s - '0') < 10)
		237	exp2 = exp2 * 10 + *s++ - '0';
		238
		239	*expo += neg ? -exp2 : exp2;
		240	}
		241
		242	break;
		243	}
		244
		245	++s;
		246
		247	uaccum = uaccum * 10 + dig;
		248	++eaccum;
		249
		250	// if we have too many digits, then recurse for more
		251	// we actually do this for rather few digits
		252	if (uaccum >= (UV_MAX - 9) / 10)
		253	{
		254	if (postdp) *expo -= eaccum;
		255	json_atof_scan1 (s, accum, expo, postdp, maxdepth);
		256	if (postdp) *expo += eaccum;
		257
		258	break;
		259	}
		260	}
		261
		262	// this relies greatly on the quality of the pow ()
		263	// implementation of the platform, but a good
		264	// implementation is hard to beat.
		265	if (postdp) *expo -= eaccum;
		266	accum += uaccum Perl_pow (10., *expo);
		267	*expo += eaccum;
		268	}
		269
		270	static NV
		271	json_atof (const char *s)
		272	{
		273	NV accum = 0.;
		274	int expo = 0;
		275	int neg = 0;
		276
		277	if (*s == '-')
		278	{
		279	++s;
		280	neg = 1;
		281	}
		282
		283	// a recursion depth of ten gives us >>500 bits
		284	json_atof_scan1 (s, &accum, &expo, 0, 10);
		285
		286	return neg ? -accum : accum;
		287	}
121	/////////////////////////////////////////////////////////////////////////////	288	/////////////////////////////////////////////////////////////////////////////
122	// encoder	289	// encoder
123		290
124	// structure used for encoding JSON	291	// structure used for encoding JSON
125	typedef struct	292	typedef struct
…		…
127	char *cur; // SvPVX (sv) + current output position	294	char *cur; // SvPVX (sv) + current output position
128	char *end; // SvEND (sv)	295	char *end; // SvEND (sv)
129	SV *sv; // result scalar	296	SV *sv; // result scalar
130	JSON json;	297	JSON json;
131	U32 indent; // indentation level	298	U32 indent; // indentation level
132	U32 maxdepth; // max. indentation/recursion level	299	UV limit; // escape character values >= this value when encoding
133	} enc_t;	300	} enc_t;
134		301
135	inline void	302	INLINE void
136	need (enc_t *enc, STRLEN len)	303	need (enc_t *enc, STRLEN len)
137	{	304	{
138	if (expect_false (enc->cur + len >= enc->end))	305	if (expect_false (enc->cur + len >= enc->end))
139	{	306	{
140	STRLEN cur = enc->cur - SvPVX (enc->sv);	307	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
141	SvGROW (enc->sv, cur + len + 1);	308	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
142	enc->cur = SvPVX (enc->sv) + cur;	309	enc->cur = SvPVX (enc->sv) + cur;
143	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	310	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
144	}	311	}
145	}	312	}
146		313
147	inline void	314	INLINE void
148	encode_ch (enc_t *enc, char ch)	315	encode_ch (enc_t *enc, char ch)
149	{	316	{
150	need (enc, 1);	317	need (enc, 1);
151	*enc->cur++ = ch;	318	*enc->cur++ = ch;
152	}	319	}
…		…
206	{	373	{
207	uch = ch;	374	uch = ch;
208	clen = 1;	375	clen = 1;
209	}	376	}
210		377
211	if (uch > 0x10FFFFUL)	378	if (uch < 0x80/0x20/ \|\| uch >= enc->limit)
212	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
213
214	if (uch < 0x80 \|\| enc->json.flags & F_ASCII \|\| (enc->json.flags & F_LATIN1 && uch > 0xFF))
215	{	379	{
216	if (uch > 0xFFFFUL)	380	if (uch >= 0x10000UL)
217	{	381	{
		382	if (uch >= 0x110000UL)
		383	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		384
218	need (enc, len += 11);	385	need (enc, len += 11);
219	sprintf (enc->cur, "\\u%04x\\u%04x",	386	sprintf (enc->cur, "\\u%04x\\u%04x",
220	(int)((uch - 0x10000) / 0x400 + 0xD800),	387	(int)((uch - 0x10000) / 0x400 + 0xD800),
221	(int)((uch - 0x10000) % 0x400 + 0xDC00));	388	(int)((uch - 0x10000) % 0x400 + 0xDC00));
222	enc->cur += 12;	389	enc->cur += 12;
223	}	390	}
224	else	391	else
225	{	392	{
226	static char hexdigit [16] = "0123456789abcdef";
227	need (enc, len += 5);	393	need (enc, len += 5);
228	*enc->cur++ = '\\';	394	*enc->cur++ = '\\';
229	*enc->cur++ = 'u';	395	*enc->cur++ = 'u';
230	*enc->cur++ = hexdigit [ uch >> 12 ];	396	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
231	*enc->cur++ = hexdigit [(uch >> 8) & 15];	397	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
232	*enc->cur++ = hexdigit [(uch >> 4) & 15];	398	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
233	*enc->cur++ = hexdigit [(uch >> 0) & 15];	399	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
234	}	400	}
235		401
236	str += clen;	402	str += clen;
237	}	403	}
238	else if (enc->json.flags & F_LATIN1)	404	else if (enc->json.flags & F_LATIN1)
…		…
250	while (--clen);	416	while (--clen);
251	}	417	}
252	else	418	else
253	{	419	{
254	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	420	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
255	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	421	enc->cur = encode_utf8 (enc->cur, uch);
256	++str;	422	++str;
257	}	423	}
258	}	424	}
259	}	425	}
260	}	426	}
261		427
262	--len;	428	--len;
263	}	429	}
264	}	430	}
265		431
266	inline void	432	INLINE void
267	encode_indent (enc_t *enc)	433	encode_indent (enc_t *enc)
268	{	434	{
269	if (enc->json.flags & F_INDENT)	435	if (enc->json.flags & F_INDENT)
270	{	436	{
271	int spaces = enc->indent * INDENT_STEP;	437	int spaces = enc->indent * INDENT_STEP;
…		…
274	memset (enc->cur, ' ', spaces);	440	memset (enc->cur, ' ', spaces);
275	enc->cur += spaces;	441	enc->cur += spaces;
276	}	442	}
277	}	443	}
278		444
279	inline void	445	INLINE void
280	encode_space (enc_t *enc)	446	encode_space (enc_t *enc)
281	{	447	{
282	need (enc, 1);	448	need (enc, 1);
283	encode_ch (enc, ' ');	449	encode_ch (enc, ' ');
284	}	450	}
285		451
286	inline void	452	INLINE void
287	encode_nl (enc_t *enc)	453	encode_nl (enc_t *enc)
288	{	454	{
289	if (enc->json.flags & F_INDENT)	455	if (enc->json.flags & F_INDENT)
290	{	456	{
291	need (enc, 1);	457	need (enc, 1);
292	encode_ch (enc, '\n');	458	encode_ch (enc, '\n');
293	}	459	}
294	}	460	}
295		461
296	inline void	462	INLINE void
297	encode_comma (enc_t *enc)	463	encode_comma (enc_t *enc)
298	{	464	{
299	encode_ch (enc, ',');	465	encode_ch (enc, ',');
300		466
301	if (enc->json.flags & F_INDENT)	467	if (enc->json.flags & F_INDENT)
…		…
309	static void	475	static void
310	encode_av (enc_t enc, AV av)	476	encode_av (enc_t enc, AV av)
311	{	477	{
312	int i, len = av_len (av);	478	int i, len = av_len (av);
313		479
314	if (enc->indent >= enc->maxdepth)	480	if (enc->indent >= enc->json.max_depth)
315	croak ("data structure too deep (hit recursion limit)");	481	croak (ERR_NESTING_EXCEEDED);
316		482
317	encode_ch (enc, '['); encode_nl (enc);	483	encode_ch (enc, '[');
318	++enc->indent;	484
		485	if (len >= 0)
		486	{
		487	encode_nl (enc); ++enc->indent;
319		488
320	for (i = 0; i <= len; ++i)	489	for (i = 0; i <= len; ++i)
321	{	490	{
322	SV **svp = av_fetch (av, i, 0);	491	SV **svp = av_fetch (av, i, 0);
323		492
324	encode_indent (enc);	493	encode_indent (enc);
325		494
326	if (svp)	495	if (svp)
327	encode_sv (enc, *svp);	496	encode_sv (enc, *svp);
328	else	497	else
329	encode_str (enc, "null", 4, 0);	498	encode_str (enc, "null", 4, 0);
330		499
331	if (i < len)	500	if (i < len)
332	encode_comma (enc);	501	encode_comma (enc);
333	}	502	}
334		503
		504	encode_nl (enc); --enc->indent; encode_indent (enc);
		505	}
		506
335	encode_nl (enc);	507	encode_ch (enc, ']');
336
337	--enc->indent;
338	encode_indent (enc); encode_ch (enc, ']');
339	}	508	}
340		509
341	static void	510	static void
342	encode_hk (enc_t enc, HE he)	511	encode_hk (enc_t enc, HE he)
343	{	512	{
…		…
391		560
392	static void	561	static void
393	encode_hv (enc_t enc, HV hv)	562	encode_hv (enc_t enc, HV hv)
394	{	563	{
395	HE *he;	564	HE *he;
396	int count;
397		565
398	if (enc->indent >= enc->maxdepth)	566	if (enc->indent >= enc->json.max_depth)
399	croak ("data structure too deep (hit recursion limit)");	567	croak (ERR_NESTING_EXCEEDED);
400		568
401	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	569	encode_ch (enc, '{');
402		570
403	// for canonical output we have to sort by keys first	571	// for canonical output we have to sort by keys first
404	// actually, this is mostly due to the stupid so-called	572	// actually, this is mostly due to the stupid so-called
405	// security workaround added somewhere in 5.8.x.	573	// security workaround added somewhere in 5.8.x
406	// that randomises hash orderings	574	// that randomises hash orderings
407	if (enc->json.flags & F_CANONICAL)	575	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
408	{	576	{
409	int count = hv_iterinit (hv);	577	int count = hv_iterinit (hv);
410		578
411	if (SvMAGICAL (hv))	579	if (SvMAGICAL (hv))
412	{	580	{
…		…
459		627
460	FREETMPS;	628	FREETMPS;
461	LEAVE;	629	LEAVE;
462	}	630	}
463		631
		632	encode_nl (enc); ++enc->indent;
		633
464	while (count--)	634	while (count--)
465	{	635	{
466	encode_indent (enc);	636	encode_indent (enc);
467	he = hes [count];	637	he = hes [count];
468	encode_hk (enc, he);	638	encode_hk (enc, he);
469	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));	639	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
470		640
471	if (count)	641	if (count)
472	encode_comma (enc);	642	encode_comma (enc);
473	}	643	}
		644
		645	encode_nl (enc); --enc->indent; encode_indent (enc);
474	}	646	}
475	}	647	}
476	else	648	else
477	{	649	{
478	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))	650	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))
479	if ((he = hv_iternext (hv)))	651	if ((he = hv_iternext (hv)))
		652	{
		653	encode_nl (enc); ++enc->indent;
		654
480	for (;;)	655	for (;;)
481	{	656	{
482	encode_indent (enc);	657	encode_indent (enc);
483	encode_hk (enc, he);	658	encode_hk (enc, he);
484	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));	659	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
485		660
486	if (!(he = hv_iternext (hv)))	661	if (!(he = hv_iternext (hv)))
487	break;	662	break;
488		663
489	encode_comma (enc);	664	encode_comma (enc);
490	}	665	}
491	}
492		666
		667	encode_nl (enc); --enc->indent; encode_indent (enc);
		668	}
		669	}
		670
493	encode_nl (enc);	671	encode_ch (enc, '}');
494
495	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
496	}	672	}
497		673
498	// encode objects, arrays and special \0=false and \1=true values.	674	// encode objects, arrays and special \0=false and \1=true values.
499	static void	675	static void
500	encode_rv (enc_t enc, SV sv)	676	encode_rv (enc_t enc, SV sv)
…		…
577		753
578	if (len == 1 && *pv == '1')	754	if (len == 1 && *pv == '1')
579	encode_str (enc, "true", 4, 0);	755	encode_str (enc, "true", 4, 0);
580	else if (len == 1 && *pv == '0')	756	else if (len == 1 && *pv == '0')
581	encode_str (enc, "false", 5, 0);	757	encode_str (enc, "false", 5, 0);
		758	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		759	encode_str (enc, "null", 4, 0);
582	else	760	else
583	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	761	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
584	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	762	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
585	}	763	}
		764	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		765	encode_str (enc, "null", 4, 0);
586	else	766	else
587	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	767	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
588	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	768	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
589	}	769	}
590		770
…		…
608	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	788	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
609	enc->cur += strlen (enc->cur);	789	enc->cur += strlen (enc->cur);
610	}	790	}
611	else if (SvIOKp (sv))	791	else if (SvIOKp (sv))
612	{	792	{
613	// we assume we can always read an IV as a UV	793	// we assume we can always read an IV as a UV and vice versa
614	if (SvUV (sv) & ~(UV)0x7fff)	794	// we assume two's complement
615	{	795	// we assume no aliasing issues in the union
616	// large integer, use the (rather slow) snprintf way.	796	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
617	need (enc, sizeof (UV) * 3);	797	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
618	enc->cur +=
619	SvIsUV(sv)
620	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
621	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
622	}
623	else
624	{	798	{
625	// optimise the "small number case"	799	// optimise the "small number case"
626	// code will likely be branchless and use only a single multiplication	800	// code will likely be branchless and use only a single multiplication
		801	// works for numbers up to 59074
627	I32 i = SvIV (sv);	802	I32 i = SvIVX (sv);
628	U32 u;	803	U32 u;
629	char digit, nz = 0;	804	char digit, nz = 0;
630		805
631	need (enc, 6);	806	need (enc, 6);
632		807
…		…
638		813
639	// now output digit by digit, each time masking out the integer part	814	// now output digit by digit, each time masking out the integer part
640	// and multiplying by 5 while moving the decimal point one to the right,	815	// and multiplying by 5 while moving the decimal point one to the right,
641	// resulting in a net multiplication by 10.	816	// resulting in a net multiplication by 10.
642	// we always write the digit to memory but conditionally increment	817	// we always write the digit to memory but conditionally increment
643	// the pointer, to ease the usage of conditional move instructions.	818	// the pointer, to enable the use of conditional move instructions.
644	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffff) 5;	819	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffffUL) 5;
645	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffff) 5;	820	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffffUL) 5;
646	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffff) 5;	821	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffffUL) 5;
647	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffff) 5;	822	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffffUL) 5;
648	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'	823	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
649	}	824	}
		825	else
		826	{
		827	// large integer, use the (rather slow) snprintf way.
		828	need (enc, IVUV_MAXCHARS);
		829	enc->cur +=
		830	SvIsUV(sv)
		831	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
		832	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		833	}
650	}	834	}
651	else if (SvROK (sv))	835	else if (SvROK (sv))
652	encode_rv (enc, SvRV (sv));	836	encode_rv (enc, SvRV (sv));
653	else if (!SvOK (sv))	837	else if (!SvOK (sv) \|\| enc->json.flags & F_ALLOW_UNKNOWN)
654	encode_str (enc, "null", 4, 0);	838	encode_str (enc, "null", 4, 0);
655	else	839	else
656	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	840	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
657	SvPV_nolen (sv), SvFLAGS (sv));	841	SvPV_nolen (sv), SvFLAGS (sv));
658	}	842	}
…		…
668	enc.json = *json;	852	enc.json = *json;
669	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	853	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
670	enc.cur = SvPVX (enc.sv);	854	enc.cur = SvPVX (enc.sv);
671	enc.end = SvEND (enc.sv);	855	enc.end = SvEND (enc.sv);
672	enc.indent = 0;	856	enc.indent = 0;
673	enc.maxdepth = DEC_DEPTH (enc.json.flags);	857	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		858	: enc.json.flags & F_LATIN1 ? 0x000100UL
		859	: 0x110000UL;
674		860
675	SvPOK_only (enc.sv);	861	SvPOK_only (enc.sv);
676	encode_sv (&enc, scalar);	862	encode_sv (&enc, scalar);
		863	encode_nl (&enc);
677		864
678	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	865	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
679	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	866	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
680		867
681	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))	868	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))
…		…
699	JSON json;	886	JSON json;
700	U32 depth; // recursion depth	887	U32 depth; // recursion depth
701	U32 maxdepth; // recursion depth limit	888	U32 maxdepth; // recursion depth limit
702	} dec_t;	889	} dec_t;
703		890
704	inline void	891	INLINE void
		892	decode_comment (dec_t *dec)
		893	{
		894	// only '#'-style comments allowed a.t.m.
		895
		896	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)
		897	++dec->cur;
		898	}
		899
		900	INLINE void
705	decode_ws (dec_t *dec)	901	decode_ws (dec_t *dec)
706	{	902	{
707	for (;;)	903	for (;;)
708	{	904	{
709	char ch = *dec->cur;	905	char ch = *dec->cur;
710		906
711	if (ch > 0x20	907	if (ch > 0x20)
		908	{
		909	if (expect_false (ch == '#'))
		910	{
		911	if (dec->json.flags & F_RELAXED)
		912	decode_comment (dec);
		913	else
		914	break;
		915	}
		916	else
		917	break;
		918	}
712	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	919	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
713	break;	920	break; // parse error, but let higher level handle it, gives better error messages
714
715	if (ch == '#' && dec->json.flags & F_RELAXED)
716	++dec->cur;
717		921
718	++dec->cur;	922	++dec->cur;
719	}	923	}
720	}	924	}
721		925
…		…
725	if (*dec->cur != ch) \	929	if (*dec->cur != ch) \
726	ERR (# ch " expected"); \	930	ERR (# ch " expected"); \
727	++dec->cur; \	931	++dec->cur; \
728	SE	932	SE
729		933
730	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	934	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
731	#define DEC_DEC_DEPTH --dec->depth	935	#define DEC_DEC_DEPTH --dec->depth
732		936
733	static SV decode_sv (dec_t dec);	937	static SV decode_sv (dec_t dec);
734		938
735	static signed char decode_hexdigit[256];	939	static signed char decode_hexdigit[256];
…		…
827		1031
828	if (hi >= 0x80)	1032	if (hi >= 0x80)
829	{	1033	{
830	utf8 = 1;	1034	utf8 = 1;
831		1035
832	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	1036	cur = encode_utf8 (cur, hi);
833	}	1037	}
834	else	1038	else
835	*cur++ = hi;	1039	*cur++ = hi;
836	}	1040	}
837	break;	1041	break;
…		…
839	default:	1043	default:
840	--dec_cur;	1044	--dec_cur;
841	ERR ("illegal backslash escape sequence in string");	1045	ERR ("illegal backslash escape sequence in string");
842	}	1046	}
843	}	1047	}
844	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	1048	else if (expect_true (ch >= 0x20 && ch < 0x80))
845	*cur++ = ch;	1049	*cur++ = ch;
846	else if (ch >= 0x80)	1050	else if (ch >= 0x80)
847	{	1051	{
848	STRLEN clen;	1052	STRLEN clen;
849	UV uch;
850		1053
851	--dec_cur;	1054	--dec_cur;
852		1055
853	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);	1056	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
854	if (clen == (STRLEN)-1)	1057	if (clen == (STRLEN)-1)
855	ERR ("malformed UTF-8 character in JSON string");	1058	ERR ("malformed UTF-8 character in JSON string");
856		1059
857	do	1060	do
858	cur++ = dec_cur++;	1061	cur++ = dec_cur++;
…		…
875	{	1078	{
876	STRLEN len = cur - buf;	1079	STRLEN len = cur - buf;
877		1080
878	if (sv)	1081	if (sv)
879	{	1082	{
880	SvGROW (sv, SvCUR (sv) + len + 1);	1083	STRLEN cur = SvCUR (sv);
		1084
		1085	if (SvLEN (sv) <= cur + len)
		1086	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1087
881	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	1088	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
882	SvCUR_set (sv, SvCUR (sv) + len);	1089	SvCUR_set (sv, SvCUR (sv) + len);
883	}	1090	}
884	else	1091	else
885	sv = newSVpvn (buf, len);	1092	sv = newSVpvn (buf, len);
…		…
972		1179
973	if (!is_nv)	1180	if (!is_nv)
974	{	1181	{
975	int len = dec->cur - start;	1182	int len = dec->cur - start;
976		1183
977	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1184	// special case the rather common 1..5-digit-int case
978	if (*start == '-')	1185	if (*start == '-')
979	switch (len)	1186	switch (len)
980	{	1187	{
981	case 2: return newSViv (-( start [1] - '0' * 1));	1188	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
982	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1189	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
983	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));	1190	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
984	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));	1191	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1192	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
985	}	1193	}
986	else	1194	else
987	switch (len)	1195	switch (len)
988	{	1196	{
989	case 1: return newSViv ( start [0] - '0' * 1);	1197	case 1: return newSViv ( start [0] - '0' * 1);
990	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1198	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
991	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);	1199	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
992	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);	1200	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1201	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
993	}	1202	}
994		1203
995	{	1204	{
996	UV uv;	1205	UV uv;
997	int numtype = grok_number (start, len, &uv);	1206	int numtype = grok_number (start, len, &uv);
…		…
1006	}	1215	}
1007		1216
1008	len -= *start == '-' ? 1 : 0;	1217	len -= *start == '-' ? 1 : 0;
1009		1218
1010	// does not fit into IV or UV, try NV	1219	// does not fit into IV or UV, try NV
1011	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)	1220	if (len <= NV_DIG)
1012	#if defined (LDBL_DIG)
1013	\|\| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1014	#endif
1015	)
1016	// fits into NV without loss of precision	1221	// fits into NV without loss of precision
1017	return newSVnv (Atof (start));	1222	return newSVnv (json_atof (start));
1018		1223
1019	// everything else fails, convert it to a string	1224	// everything else fails, convert it to a string
1020	return newSVpvn (start, dec->cur - start);	1225	return newSVpvn (start, dec->cur - start);
1021	}	1226	}
1022		1227
1023	// loss of precision here	1228	// loss of precision here
1024	return newSVnv (Atof (start));	1229	return newSVnv (json_atof (start));
1025		1230
1026	fail:	1231	fail:
1027	return 0;	1232	return 0;
1028	}	1233	}
1029		1234
…		…
1105	char *p = dec->cur;	1310	char *p = dec->cur;
1106	char *e = p + 24; // only try up to 24 bytes	1311	char *e = p + 24; // only try up to 24 bytes
1107		1312
1108	for (;;)	1313	for (;;)
1109	{	1314	{
1110	// the >= 0x80 is true on most architectures	1315	// the >= 0x80 is false on most architectures
1111	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')	1316	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
1112	{	1317	{
1113	// slow path, back up and use decode_str	1318	// slow path, back up and use decode_str
1114	SV *key = decode_str (dec);	1319	SV *key = decode_str (dec);
1115	if (!key)	1320	if (!key)
…		…
1198	dSP;	1403	dSP;
1199	int count;	1404	int count;
1200		1405
1201	ENTER; SAVETMPS; PUSHMARK (SP);	1406	ENTER; SAVETMPS; PUSHMARK (SP);
1202	XPUSHs (HeVAL (he));	1407	XPUSHs (HeVAL (he));
		1408	sv_2mortal (sv);
1203		1409
1204	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;	1410	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1205		1411
1206	if (count == 1)	1412	if (count == 1)
1207	{	1413	{
1208	sv = newSVsv (POPs);	1414	sv = newSVsv (POPs);
1209	FREETMPS; LEAVE;	1415	FREETMPS; LEAVE;
1210	return sv;	1416	return sv;
1211	}	1417	}
1212		1418
		1419	SvREFCNT_inc (sv);
1213	FREETMPS; LEAVE;	1420	FREETMPS; LEAVE;
1214	}	1421	}
1215	}	1422	}
1216		1423
1217	if (dec->json.cb_object)	1424	if (dec->json.cb_object)
…		…
1246		1453
1247	static SV *	1454	static SV *
1248	decode_sv (dec_t *dec)	1455	decode_sv (dec_t *dec)
1249	{	1456	{
1250	// the beauty of JSON: you need exactly one character lookahead	1457	// the beauty of JSON: you need exactly one character lookahead
1251	// to parse anything.	1458	// to parse everything.
1252	switch (*dec->cur)	1459	switch (*dec->cur)
1253	{	1460	{
1254	case '"': ++dec->cur; return decode_str (dec);	1461	case '"': ++dec->cur; return decode_str (dec);
1255	case '[': ++dec->cur; return decode_av (dec);	1462	case '[': ++dec->cur; return decode_av (dec);
1256	case '{': ++dec->cur; return decode_hv (dec);	1463	case '{': ++dec->cur; return decode_hv (dec);
1257		1464
1258	case '-':	1465	case '-':
1259	case '0': case '1': case '2': case '3': case '4':	1466	case '0': case '1': case '2': case '3': case '4':
1260	case '5': case '6': case '7': case '8': case '9':	1467	case '5': case '6': case '7': case '8': case '9':
1261	return decode_num (dec);	1468	return decode_num (dec);
…		…
1263	case 't':	1470	case 't':
1264	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1471	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1265	{	1472	{
1266	dec->cur += 4;	1473	dec->cur += 4;
1267	#if JSON_SLOW	1474	#if JSON_SLOW
1268	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);	1475	json_true = get_bool ("JSON::XS::true");
1269	#endif	1476	#endif
1270	return SvREFCNT_inc (json_true);	1477	return newSVsv (json_true);
1271	}	1478	}
1272	else	1479	else
1273	ERR ("'true' expected");	1480	ERR ("'true' expected");
1274		1481
1275	break;	1482	break;
…		…
1277	case 'f':	1484	case 'f':
1278	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1485	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1279	{	1486	{
1280	dec->cur += 5;	1487	dec->cur += 5;
1281	#if JSON_SLOW	1488	#if JSON_SLOW
1282	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1489	json_false = get_bool ("JSON::XS::false");
1283	#endif	1490	#endif
1284	return SvREFCNT_inc (json_false);	1491	return newSVsv (json_false);
1285	}	1492	}
1286	else	1493	else
1287	ERR ("'false' expected");	1494	ERR ("'false' expected");
1288		1495
1289	break;	1496	break;
…		…
1307	fail:	1514	fail:
1308	return 0;	1515	return 0;
1309	}	1516	}
1310		1517
1311	static SV *	1518	static SV *
1312	decode_json (SV string, JSON json, UV *offset_return)	1519	decode_json (SV string, JSON json, char **offset_return)
1313	{	1520	{
1314	dec_t dec;	1521	dec_t dec;
1315	UV offset;
1316	SV *sv;	1522	SV *sv;
1317		1523
		1524	/* work around bugs in 5.10 where manipulating magic values
		1525	* will perl ignore the magic in subsequent accesses
		1526	*/
1318	SvGETMAGIC (string);	1527	/SvGETMAGIC (string);/
		1528	if (SvMAGICAL (string))
		1529	string = sv_2mortal (newSVsv (string));
		1530
1319	SvUPGRADE (string, SVt_PV);	1531	SvUPGRADE (string, SVt_PV);
1320		1532
1321	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))	1533	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1534	* assertion with -DDEBUGGING, although SvCUR is documented to
		1535	* return the xpv_cur field which certainly exists after upgrading.
		1536	* according to nicholas clark, calling SvPOK fixes this.
		1537	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1538	* and hope for the best.
		1539	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1540	* assertion business is seriously broken, try yet another workaround
		1541	* for the broken -DDEBUGGING.
		1542	*/
		1543	{
		1544	#ifdef DEBUGGING
		1545	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1546	#else
		1547	STRLEN offset = SvCUR (string);
		1548	#endif
		1549
		1550	if (offset > json->max_size && json->max_size)
1322	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",	1551	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1323	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));	1552	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1553	}
1324		1554
1325	if (json->flags & F_UTF8)	1555	if (DECODE_WANTS_OCTETS (json))
1326	sv_utf8_downgrade (string, 0);	1556	sv_utf8_downgrade (string, 0);
1327	else	1557	else
1328	sv_utf8_upgrade (string);	1558	sv_utf8_upgrade (string);
1329		1559
1330	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1560	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1331		1561
1332	dec.json = *json;	1562	dec.json = *json;
1333	dec.cur = SvPVX (string);	1563	dec.cur = SvPVX (string);
1334	dec.end = SvEND (string);	1564	dec.end = SvEND (string);
1335	dec.err = 0;	1565	dec.err = 0;
1336	dec.depth = 0;	1566	dec.depth = 0;
1337	dec.maxdepth = DEC_DEPTH (dec.json.flags);
1338		1567
1339	if (dec.json.cb_object \|\| dec.json.cb_sk_object)	1568	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
1340	dec.json.flags \|= F_HOOK;	1569	dec.json.flags \|= F_HOOK;
1341		1570
1342	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1571	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
1343		1572
1344	decode_ws (&dec);	1573	decode_ws (&dec);
1345	sv = decode_sv (&dec);	1574	sv = decode_sv (&dec);
		1575
		1576	if (offset_return)
		1577	*offset_return = dec.cur;
1346		1578
1347	if (!(offset_return \|\| !sv))	1579	if (!(offset_return \|\| !sv))
1348	{	1580	{
1349	// check for trailing garbage	1581	// check for trailing garbage
1350	decode_ws (&dec);	1582	decode_ws (&dec);
…		…
1353	{	1585	{
1354	dec.err = "garbage after JSON object";	1586	dec.err = "garbage after JSON object";
1355	SvREFCNT_dec (sv);	1587	SvREFCNT_dec (sv);
1356	sv = 0;	1588	sv = 0;
1357	}	1589	}
1358	}
1359
1360	if (offset_return \|\| !sv)
1361	{
1362	offset = dec.json.flags & F_UTF8
1363	? dec.cur - SvPVX (string)
1364	: utf8_distance (dec.cur, SvPVX (string));
1365
1366	if (offset_return)
1367	*offset_return = offset;
1368	}	1590	}
1369		1591
1370	if (!sv)	1592	if (!sv)
1371	{	1593	{
1372	SV *uni = sv_newmortal ();	1594	SV *uni = sv_newmortal ();
…		…
1378	SAVEVPTR (PL_curcop);	1600	SAVEVPTR (PL_curcop);
1379	PL_curcop = &cop;	1601	PL_curcop = &cop;
1380	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1602	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1381	LEAVE;	1603	LEAVE;
1382		1604
1383	croak ("%s, at character offset %d [\"%s\"]",	1605	croak ("%s, at character offset %d (before \"%s\")",
1384	dec.err,	1606	dec.err,
1385	(int)offset,	1607	ptr_to_index (string, dec.cur),
1386	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1608	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1387	}	1609	}
1388		1610
1389	sv = sv_2mortal (sv);	1611	sv = sv_2mortal (sv);
1390		1612
1391	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))	1613	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1392	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1614	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1393		1615
1394	return sv;	1616	return sv;
		1617	}
		1618
		1619	/////////////////////////////////////////////////////////////////////////////
		1620	// incremental parser
		1621
		1622	static void
		1623	incr_parse (JSON *self)
		1624	{
		1625	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1626
		1627	// the state machine here is a bit convoluted and could be simplified a lot
		1628	// but this would make it slower, so...
		1629
		1630	for (;;)
		1631	{
		1632	//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
		1633	switch (self->incr_mode)
		1634	{
		1635	// only used for initial whitespace skipping
		1636	case INCR_M_WS:
		1637	for (;;)
		1638	{
		1639	if (*p > 0x20)
		1640	{
		1641	if (*p == '#')
		1642	{
		1643	self->incr_mode = INCR_M_C0;
		1644	goto incr_m_c;
		1645	}
		1646	else
		1647	{
		1648	self->incr_mode = INCR_M_JSON;
		1649	goto incr_m_json;
		1650	}
		1651	}
		1652	else if (!*p)
		1653	goto interrupt;
		1654
		1655	++p;
		1656	}
		1657
		1658	// skip a single char inside a string (for \\-processing)
		1659	case INCR_M_BS:
		1660	if (!*p)
		1661	goto interrupt;
		1662
		1663	++p;
		1664	self->incr_mode = INCR_M_STR;
		1665	goto incr_m_str;
		1666
		1667	// inside #-style comments
		1668	case INCR_M_C0:
		1669	case INCR_M_C1:
		1670	incr_m_c:
		1671	for (;;)
		1672	{
		1673	if (*p == '\n')
		1674	{
		1675	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1676	break;
		1677	}
		1678	else if (!*p)
		1679	goto interrupt;
		1680
		1681	++p;
		1682	}
		1683
		1684	break;
		1685
		1686	// inside a string
		1687	case INCR_M_STR:
		1688	incr_m_str:
		1689	for (;;)
		1690	{
		1691	if (*p == '"')
		1692	{
		1693	++p;
		1694	self->incr_mode = INCR_M_JSON;
		1695
		1696	if (!self->incr_nest)
		1697	goto interrupt;
		1698
		1699	goto incr_m_json;
		1700	}
		1701	else if (*p == '\\')
		1702	{
		1703	++p; // "virtually" consumes character after \
		1704
		1705	if (!*p) // if at end of string we have to switch modes
		1706	{
		1707	self->incr_mode = INCR_M_BS;
		1708	goto interrupt;
		1709	}
		1710	}
		1711	else if (!*p)
		1712	goto interrupt;
		1713
		1714	++p;
		1715	}
		1716
		1717	// after initial ws, outside string
		1718	case INCR_M_JSON:
		1719	incr_m_json:
		1720	for (;;)
		1721	{
		1722	switch (*p++)
		1723	{
		1724	case 0:
		1725	--p;
		1726	goto interrupt;
		1727
		1728	case 0x09:
		1729	case 0x0a:
		1730	case 0x0d:
		1731	case 0x20:
		1732	if (!self->incr_nest)
		1733	{
		1734	--p; // do not eat the whitespace, let the next round do it
		1735	goto interrupt;
		1736	}
		1737	break;
		1738
		1739	case '"':
		1740	self->incr_mode = INCR_M_STR;
		1741	goto incr_m_str;
		1742
		1743	case '[':
		1744	case '{':
		1745	if (++self->incr_nest > self->max_depth)
		1746	croak (ERR_NESTING_EXCEEDED);
		1747	break;
		1748
		1749	case ']':
		1750	case '}':
		1751	if (--self->incr_nest <= 0)
		1752	goto interrupt;
		1753	break;
		1754
		1755	case '#':
		1756	self->incr_mode = INCR_M_C1;
		1757	goto incr_m_c;
		1758	}
		1759	}
		1760	}
		1761
		1762	modechange:
		1763	;
		1764	}
		1765
		1766	interrupt:
		1767	self->incr_pos = p - SvPVX (self->incr_text);
		1768	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1769	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1395	}	1770	}
1396		1771
1397	/////////////////////////////////////////////////////////////////////////////	1772	/////////////////////////////////////////////////////////////////////////////
1398	// XS interface functions	1773	// XS interface functions
1399		1774
…		…
1411	: -1;	1786	: -1;
1412		1787
1413	json_stash = gv_stashpv ("JSON::XS" , 1);	1788	json_stash = gv_stashpv ("JSON::XS" , 1);
1414	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);	1789	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1415		1790
1416	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );	1791	json_true = get_bool ("JSON::XS::true");
1417	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1792	json_false = get_bool ("JSON::XS::false");
		1793
		1794	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1418	}	1795	}
1419		1796
1420	PROTOTYPES: DISABLE	1797	PROTOTYPES: DISABLE
1421		1798
1422	void CLONE (...)	1799	void CLONE (...)
…		…
1427	void new (char *klass)	1804	void new (char *klass)
1428	PPCODE:	1805	PPCODE:
1429	{	1806	{
1430	SV *pv = NEWSV (0, sizeof (JSON));	1807	SV *pv = NEWSV (0, sizeof (JSON));
1431	SvPOK_only (pv);	1808	SvPOK_only (pv);
1432	Zero (SvPVX (pv), 1, JSON);	1809	json_init ((JSON *)SvPVX (pv));
1433	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	1810	XPUSHs (sv_2mortal (sv_bless (
1434	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH)));	1811	newRV_noinc (pv),
		1812	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1813	)));
1435	}	1814	}
1436		1815
1437	void ascii (JSON *self, int enable = 1)	1816	void ascii (JSON *self, int enable = 1)
1438	ALIAS:	1817	ALIAS:
1439	ascii = F_ASCII	1818	ascii = F_ASCII
…		…
1447	allow_nonref = F_ALLOW_NONREF	1826	allow_nonref = F_ALLOW_NONREF
1448	shrink = F_SHRINK	1827	shrink = F_SHRINK
1449	allow_blessed = F_ALLOW_BLESSED	1828	allow_blessed = F_ALLOW_BLESSED
1450	convert_blessed = F_CONV_BLESSED	1829	convert_blessed = F_CONV_BLESSED
1451	relaxed = F_RELAXED	1830	relaxed = F_RELAXED
		1831	allow_unknown = F_ALLOW_UNKNOWN
1452	PPCODE:	1832	PPCODE:
1453	{	1833	{
1454	if (enable)	1834	if (enable)
1455	self->flags \|= ix;	1835	self->flags \|= ix;
1456	else	1836	else
1457	self->flags &= ~ix;	1837	self->flags &= ~ix;
1458		1838
1459	XPUSHs (ST (0));	1839	XPUSHs (ST (0));
1460	}	1840	}
1461		1841
1462	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1842	void get_ascii (JSON *self)
		1843	ALIAS:
		1844	get_ascii = F_ASCII
		1845	get_latin1 = F_LATIN1
		1846	get_utf8 = F_UTF8
		1847	get_indent = F_INDENT
		1848	get_canonical = F_CANONICAL
		1849	get_space_before = F_SPACE_BEFORE
		1850	get_space_after = F_SPACE_AFTER
		1851	get_allow_nonref = F_ALLOW_NONREF
		1852	get_shrink = F_SHRINK
		1853	get_allow_blessed = F_ALLOW_BLESSED
		1854	get_convert_blessed = F_CONV_BLESSED
		1855	get_relaxed = F_RELAXED
		1856	get_allow_unknown = F_ALLOW_UNKNOWN
1463	PPCODE:	1857	PPCODE:
1464	{	1858	XPUSHs (boolSV (self->flags & ix));
1465	UV log2 = 0;
1466		1859
1467	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	1860	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1468		1861	PPCODE:
1469	while ((1UL << log2) < max_depth)	1862	self->max_depth = max_depth;
1470	++log2;
1471
1472	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1473
1474	XPUSHs (ST (0));	1863	XPUSHs (ST (0));
1475	}
1476		1864
		1865	U32 get_max_depth (JSON *self)
		1866	CODE:
		1867	RETVAL = self->max_depth;
		1868	OUTPUT:
		1869	RETVAL
		1870
1477	void max_size (JSON *self, UV max_size = 0)	1871	void max_size (JSON *self, U32 max_size = 0)
1478	PPCODE:	1872	PPCODE:
1479	{	1873	self->max_size = max_size;
1480	UV log2 = 0;
1481
1482	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1483	if (max_size == 1) max_size = 2;
1484
1485	while ((1UL << log2) < max_size)
1486	++log2;
1487
1488	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
1489
1490	XPUSHs (ST (0));	1874	XPUSHs (ST (0));
1491	}	1875
		1876	int get_max_size (JSON *self)
		1877	CODE:
		1878	RETVAL = self->max_size;
		1879	OUTPUT:
		1880	RETVAL
1492		1881
1493	void filter_json_object (JSON self, SV cb = &PL_sv_undef)	1882	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1494	PPCODE:	1883	PPCODE:
1495	{	1884	{
1496	SvREFCNT_dec (self->cb_object);	1885	SvREFCNT_dec (self->cb_object);
…		…
1530	XPUSHs (decode_json (jsonstr, self, 0));	1919	XPUSHs (decode_json (jsonstr, self, 0));
1531		1920
1532	void decode_prefix (JSON self, SV jsonstr)	1921	void decode_prefix (JSON self, SV jsonstr)
1533	PPCODE:	1922	PPCODE:
1534	{	1923	{
1535	UV offset;	1924	char *offset;
1536	EXTEND (SP, 2);	1925	EXTEND (SP, 2);
1537	PUSHs (decode_json (jsonstr, self, &offset));	1926	PUSHs (decode_json (jsonstr, self, &offset));
1538	PUSHs (sv_2mortal (newSVuv (offset)));	1927	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		1928	}
		1929
		1930	void incr_parse (JSON self, SV jsonstr = 0)
		1931	PPCODE:
		1932	{
		1933	if (!self->incr_text)
		1934	self->incr_text = newSVpvn ("", 0);
		1935
		1936	/* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
		1937	if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
		1938	if (DECODE_WANTS_OCTETS (self))
		1939	{
		1940	if (self->incr_pos)
		1941	self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
		1942	(U8 *)SvPVX (self->incr_text) + self->incr_pos);
		1943
		1944	sv_utf8_downgrade (self->incr_text, 0);
		1945	}
		1946	else
		1947	{
		1948	sv_utf8_upgrade (self->incr_text);
		1949
		1950	if (self->incr_pos)
		1951	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1952	- (U8 *)SvPVX (self->incr_text);
		1953	}
		1954
		1955	// append data, if any
		1956	if (jsonstr)
		1957	{
		1958	/* make sure both strings have same encoding */
		1959	if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
		1960	if (SvUTF8 (jsonstr))
		1961	sv_utf8_downgrade (jsonstr, 0);
		1962	else
		1963	sv_utf8_upgrade (jsonstr);
		1964
		1965	/* and then just blindly append */
		1966	{
		1967	STRLEN len;
		1968	const char *str = SvPV (jsonstr, len);
		1969	STRLEN cur = SvCUR (self->incr_text);
		1970
		1971	if (SvLEN (self->incr_text) <= cur + len)
		1972	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1973
		1974	Move (str, SvEND (self->incr_text), len, char);
		1975	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1976	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1977	}
		1978	}
		1979
		1980	if (GIMME_V != G_VOID)
		1981	do
		1982	{
		1983	char *offset;
		1984
		1985	if (!INCR_DONE (self))
		1986	{
		1987	incr_parse (self);
		1988
		1989	if (self->incr_pos > self->max_size && self->max_size)
		1990	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1991	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1992
		1993	if (!INCR_DONE (self))
		1994	break;
		1995	}
		1996
		1997	XPUSHs (decode_json (self->incr_text, self, &offset));
		1998
		1999	self->incr_pos -= offset - SvPVX (self->incr_text);
		2000	self->incr_nest = 0;
		2001	self->incr_mode = 0;
		2002
		2003	sv_chop (self->incr_text, offset);
		2004	}
		2005	while (GIMME_V == G_ARRAY);
		2006	}
		2007
		2008	SV incr_text (JSON self)
		2009	ATTRS: lvalue
		2010	CODE:
		2011	{
		2012	if (self->incr_pos)
		2013	croak ("incr_text can not be called when the incremental parser already started parsing");
		2014
		2015	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		2016	}
		2017	OUTPUT:
		2018	RETVAL
		2019
		2020	void incr_skip (JSON *self)
		2021	CODE:
		2022	{
		2023	if (self->incr_pos)
		2024	{
		2025	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2026	self->incr_pos = 0;
		2027	self->incr_nest = 0;
		2028	self->incr_mode = 0;
		2029	}
		2030	}
		2031
		2032	void incr_reset (JSON *self)
		2033	CODE:
		2034	{
		2035	SvREFCNT_dec (self->incr_text);
		2036	self->incr_text = 0;
		2037	self->incr_pos = 0;
		2038	self->incr_nest = 0;
		2039	self->incr_mode = 0;
1539	}	2040	}
1540		2041
1541	void DESTROY (JSON *self)	2042	void DESTROY (JSON *self)
1542	CODE:	2043	CODE:
1543	SvREFCNT_dec (self->cb_sk_object);	2044	SvREFCNT_dec (self->cb_sk_object);
1544	SvREFCNT_dec (self->cb_object);	2045	SvREFCNT_dec (self->cb_object);
		2046	SvREFCNT_dec (self->incr_text);
1545		2047
1546	PROTOTYPES: ENABLE	2048	PROTOTYPES: ENABLE
1547		2049
1548	void to_json (SV *scalar)	2050	void encode_json (SV *scalar)
		2051	ALIAS:
		2052	to_json_ = 0
		2053	encode_json = F_UTF8
1549	PPCODE:	2054	PPCODE:
1550	{	2055	{
1551	JSON json = { F_DEFAULT \| F_UTF8 };	2056	JSON json;
		2057	json_init (&json);
		2058	json.flags \|= ix;
1552	XPUSHs (encode_json (scalar, &json));	2059	XPUSHs (encode_json (scalar, &json));
1553	}	2060	}
1554		2061
1555	void from_json (SV *jsonstr)	2062	void decode_json (SV *jsonstr)
		2063	ALIAS:
		2064	from_json_ = 0
		2065	decode_json = F_UTF8
1556	PPCODE:	2066	PPCODE:
1557	{	2067	{
1558	JSON json = { F_DEFAULT \| F_UTF8 };	2068	JSON json;
		2069	json_init (&json);
		2070	json.flags \|= ix;
1559	XPUSHs (decode_json (jsonstr, &json, 0));	2071	XPUSHs (decode_json (jsonstr, &json, 0));
1560	}	2072	}
1561		2073

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.63 by root, Mon Aug 27 01:49:01 2007 UTC vs. Revision 1.108 by root, Tue Aug 17 23:27:36 2010 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.63 by root, Mon Aug 27 01:49:01 2007 UTC vs.
Revision 1.108 by root, Tue Aug 17 23:27:36 2010 UTC