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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.64 by root, Mon Aug 27 02:03:23 2007 UTC vs.
Revision 1.105 by root, Tue Jan 19 01:07:27 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
51	#define SB do {	47	#define SB do {
52	#define SE } while (0)	48	#define SE } while (0)
53		49
54	#if __GNUC__ >= 3	50	#if __GNUC__ >= 3
55	# define expect(expr,value) __builtin_expect ((expr),(value))	51	# define expect(expr,value) __builtin_expect ((expr), (value))
56	# define inline inline	52	# define INLINE static inline
57	#else	53	#else
58	# define expect(expr,value) (expr)	54	# define expect(expr,value) (expr)
59	# define inline static	55	# define INLINE static
60	#endif	56	#endif
61		57
62	#define expect_false(expr) expect ((expr) != 0, 0)	58	#define expect_false(expr) expect ((expr) != 0, 0)
63	#define expect_true(expr) expect ((expr) != 0, 1)	59	#define expect_true(expr) expect ((expr) != 0, 1)
		60
		61	#define IN_RANGE_INC(type,val,beg,end) \
		62	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		63	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		64
		65	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
64		66
65	#ifdef USE_ITHREADS	67	#ifdef USE_ITHREADS
66	# define JSON_SLOW 1	68	# define JSON_SLOW 1
67	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))	69	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
68	#else	70	#else
…		…
71	#endif	73	#endif
72		74
73	static HV json_stash, json_boolean_stash; // JSON::XS::	75	static HV json_stash, json_boolean_stash; // JSON::XS::
74	static SV json_true, json_false;	76	static SV json_true, json_false;
75		77
		78	enum {
		79	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		80	INCR_M_STR, // inside string
		81	INCR_M_BS, // inside backslash
		82	INCR_M_C0, // inside comment in initial whitespace sequence
		83	INCR_M_C1, // inside comment in other places
		84	INCR_M_JSON // outside anything, count nesting
		85	};
		86
		87	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		88
76	typedef struct {	89	typedef struct {
77	U32 flags;	90	U32 flags;
		91	U32 max_depth;
		92	STRLEN max_size;
		93
78	SV *cb_object;	94	SV *cb_object;
79	HV *cb_sk_object;	95	HV *cb_sk_object;
		96
		97	// for the incremental parser
		98	SV *incr_text; // the source text so far
		99	STRLEN incr_pos; // the current offset into the text
		100	int incr_nest; // {[]}-nesting level
		101	unsigned char incr_mode;
80	} JSON;	102	} JSON;
		103
		104	INLINE void
		105	json_init (JSON *json)
		106	{
		107	Zero (json, 1, JSON);
		108	json->max_depth = 512;
		109	}
81		110
82	/////////////////////////////////////////////////////////////////////////////	111	/////////////////////////////////////////////////////////////////////////////
83	// utility functions	112	// utility functions
84		113
85	inline void	114	INLINE SV *
		115	get_bool (const char *name)
		116	{
		117	SV *sv = get_sv (name, 1);
		118
		119	SvREADONLY_on (sv);
		120	SvREADONLY_on (SvRV (sv));
		121
		122	return sv;
		123	}
		124
		125	INLINE void
86	shrink (SV *sv)	126	shrink (SV *sv)
87	{	127	{
88	sv_utf8_downgrade (sv, 1);	128	sv_utf8_downgrade (sv, 1);
		129
89	if (SvLEN (sv) > SvCUR (sv) + 1)	130	if (SvLEN (sv) > SvCUR (sv) + 1)
90	{	131	{
91	#ifdef SvPV_shrink_to_cur	132	#ifdef SvPV_shrink_to_cur
92	SvPV_shrink_to_cur (sv);	133	SvPV_shrink_to_cur (sv);
93	#elif defined (SvPV_renew)	134	#elif defined (SvPV_renew)
…		…
99	// decode an utf-8 character and return it, or (UV)-1 in	140	// decode an utf-8 character and return it, or (UV)-1 in
100	// case of an error.	141	// case of an error.
101	// we special-case "safe" characters from U+80 .. U+7FF,	142	// we special-case "safe" characters from U+80 .. U+7FF,
102	// but use the very good perl function to parse anything else.	143	// but use the very good perl function to parse anything else.
103	// note that we never call this function for a ascii codepoints	144	// note that we never call this function for a ascii codepoints
104	inline UV	145	INLINE UV
105	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	146	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
106	{	147	{
107	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))	148	if (expect_true (len >= 2
108	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	149	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
109	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	150	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
110	{	151	{
111	*clen = 2;	152	*clen = 2;
112	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);	153	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
113	}	154	}
114	else	155	else
115	{	156	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
116	*clen = (STRLEN)-1;	157	}
117	return (UV)-1;	158
		159	// likewise for encoding, also never called for ascii codepoints
		160	// this function takes advantage of this fact, although current gccs
		161	// seem to optimise the check for >= 0x80 away anyways
		162	INLINE unsigned char *
		163	encode_utf8 (unsigned char *s, UV ch)
		164	{
		165	if (expect_false (ch < 0x000080))
		166	*s++ = ch;
		167	else if (expect_true (ch < 0x000800))
		168	*s++ = 0xc0 \| ( ch >> 6),
		169	*s++ = 0x80 \| ( ch & 0x3f);
		170	else if ( ch < 0x010000)
		171	*s++ = 0xe0 \| ( ch >> 12),
		172	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		173	*s++ = 0x80 \| ( ch & 0x3f);
		174	else if ( ch < 0x110000)
		175	*s++ = 0xf0 \| ( ch >> 18),
		176	*s++ = 0x80 \| ((ch >> 12) & 0x3f),
		177	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		178	*s++ = 0x80 \| ( ch & 0x3f);
		179
		180	return s;
		181	}
		182
		183	// convert offset pointer to character index, sv must be string
		184	static STRLEN
		185	ptr_to_index (SV sv, char offset)
		186	{
		187	return SvUTF8 (sv)
		188	? utf8_distance (offset, SvPVX (sv))
		189	: offset - SvPVX (sv);
		190	}
		191
		192	/////////////////////////////////////////////////////////////////////////////
		193	// fp hell
		194
		195	// scan a group of digits, and a trailing exponent
		196	static void
		197	json_atof_scan1 (const char s, NV accum, int *expo, int postdp)
		198	{
		199	UV uaccum = 0;
		200	int eaccum = 0;
		201
		202	for (;;)
118	}	203	{
119	}	204	U8 dig = (U8)*s - '0';
120		205
		206	if (expect_false (dig >= 10))
		207	{
		208	if (dig == (U8)((U8)'.' - (U8)'0'))
		209	{
		210	++s;
		211	json_atof_scan1 (s, accum, expo, 1);
		212	}
		213	else if ((dig \| ' ') == 'e' - '0')
		214	{
		215	int exp2 = 0;
		216	int neg = 0;
		217
		218	++s;
		219
		220	if (*s == '-')
		221	{
		222	++s;
		223	neg = 1;
		224	}
		225	else if (*s == '+')
		226	++s;
		227
		228	while ((dig = (U8)*s - '0') < 10)
		229	exp2 = exp2 * 10 + *s++ - '0';
		230
		231	*expo += neg ? -exp2 : exp2;
		232	}
		233
		234	break;
		235	}
		236
		237	++s;
		238
		239	uaccum = uaccum * 10 + dig;
		240	++eaccum;
		241
		242	// if we have too many digits, then recurse for more
		243	// we actually do this for rather few digits
		244	if (uaccum >= (UV_MAX - 9) / 10)
		245	{
		246	if (postdp) *expo -= eaccum;
		247	json_atof_scan1 (s, accum, expo, postdp);
		248	if (postdp) *expo += eaccum;
		249
		250	break;
		251	}
		252	}
		253
		254	// this relies greatly on the quality of the pow ()
		255	// implementation of the platform, but a good
		256	// implementation is hard to beat.
		257	if (postdp) *expo -= eaccum;
		258	accum += uaccum Perl_pow (10., *expo);
		259	*expo += eaccum;
		260	}
		261
		262	static NV
		263	json_atof (const char *s)
		264	{
		265	NV accum = 0.;
		266	int expo = 0;
		267	int neg = 0;
		268
		269	if (*s == '-')
		270	{
		271	++s;
		272	neg = 1;
		273	}
		274
		275	json_atof_scan1 (s, &accum, &expo, 0);
		276
		277	return neg ? -accum : accum;
		278	}
121	/////////////////////////////////////////////////////////////////////////////	279	/////////////////////////////////////////////////////////////////////////////
122	// encoder	280	// encoder
123		281
124	// structure used for encoding JSON	282	// structure used for encoding JSON
125	typedef struct	283	typedef struct
…		…
127	char *cur; // SvPVX (sv) + current output position	285	char *cur; // SvPVX (sv) + current output position
128	char *end; // SvEND (sv)	286	char *end; // SvEND (sv)
129	SV *sv; // result scalar	287	SV *sv; // result scalar
130	JSON json;	288	JSON json;
131	U32 indent; // indentation level	289	U32 indent; // indentation level
132	U32 maxdepth; // max. indentation/recursion level	290	UV limit; // escape character values >= this value when encoding
133	} enc_t;	291	} enc_t;
134		292
135	inline void	293	INLINE void
136	need (enc_t *enc, STRLEN len)	294	need (enc_t *enc, STRLEN len)
137	{	295	{
138	if (expect_false (enc->cur + len >= enc->end))	296	if (expect_false (enc->cur + len >= enc->end))
139	{	297	{
140	STRLEN cur = enc->cur - SvPVX (enc->sv);	298	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
141	SvGROW (enc->sv, cur + len + 1);	299	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
142	enc->cur = SvPVX (enc->sv) + cur;	300	enc->cur = SvPVX (enc->sv) + cur;
143	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	301	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
144	}	302	}
145	}	303	}
146		304
147	inline void	305	INLINE void
148	encode_ch (enc_t *enc, char ch)	306	encode_ch (enc_t *enc, char ch)
149	{	307	{
150	need (enc, 1);	308	need (enc, 1);
151	*enc->cur++ = ch;	309	*enc->cur++ = ch;
152	}	310	}
…		…
206	{	364	{
207	uch = ch;	365	uch = ch;
208	clen = 1;	366	clen = 1;
209	}	367	}
210		368
211	if (uch > 0x10FFFFUL)	369	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	{	370	{
216	if (uch > 0xFFFFUL)	371	if (uch >= 0x10000UL)
217	{	372	{
		373	if (uch >= 0x110000UL)
		374	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		375
218	need (enc, len += 11);	376	need (enc, len += 11);
219	sprintf (enc->cur, "\\u%04x\\u%04x",	377	sprintf (enc->cur, "\\u%04x\\u%04x",
220	(int)((uch - 0x10000) / 0x400 + 0xD800),	378	(int)((uch - 0x10000) / 0x400 + 0xD800),
221	(int)((uch - 0x10000) % 0x400 + 0xDC00));	379	(int)((uch - 0x10000) % 0x400 + 0xDC00));
222	enc->cur += 12;	380	enc->cur += 12;
223	}	381	}
224	else	382	else
225	{	383	{
226	static char hexdigit [16] = "0123456789abcdef";
227	need (enc, len += 5);	384	need (enc, len += 5);
228	*enc->cur++ = '\\';	385	*enc->cur++ = '\\';
229	*enc->cur++ = 'u';	386	*enc->cur++ = 'u';
230	*enc->cur++ = hexdigit [ uch >> 12 ];	387	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
231	*enc->cur++ = hexdigit [(uch >> 8) & 15];	388	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
232	*enc->cur++ = hexdigit [(uch >> 4) & 15];	389	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
233	*enc->cur++ = hexdigit [(uch >> 0) & 15];	390	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
234	}	391	}
235		392
236	str += clen;	393	str += clen;
237	}	394	}
238	else if (enc->json.flags & F_LATIN1)	395	else if (enc->json.flags & F_LATIN1)
…		…
250	while (--clen);	407	while (--clen);
251	}	408	}
252	else	409	else
253	{	410	{
254	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	411	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
255	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	412	enc->cur = encode_utf8 (enc->cur, uch);
256	++str;	413	++str;
257	}	414	}
258	}	415	}
259	}	416	}
260	}	417	}
261		418
262	--len;	419	--len;
263	}	420	}
264	}	421	}
265		422
266	inline void	423	INLINE void
267	encode_indent (enc_t *enc)	424	encode_indent (enc_t *enc)
268	{	425	{
269	if (enc->json.flags & F_INDENT)	426	if (enc->json.flags & F_INDENT)
270	{	427	{
271	int spaces = enc->indent * INDENT_STEP;	428	int spaces = enc->indent * INDENT_STEP;
…		…
274	memset (enc->cur, ' ', spaces);	431	memset (enc->cur, ' ', spaces);
275	enc->cur += spaces;	432	enc->cur += spaces;
276	}	433	}
277	}	434	}
278		435
279	inline void	436	INLINE void
280	encode_space (enc_t *enc)	437	encode_space (enc_t *enc)
281	{	438	{
282	need (enc, 1);	439	need (enc, 1);
283	encode_ch (enc, ' ');	440	encode_ch (enc, ' ');
284	}	441	}
285		442
286	inline void	443	INLINE void
287	encode_nl (enc_t *enc)	444	encode_nl (enc_t *enc)
288	{	445	{
289	if (enc->json.flags & F_INDENT)	446	if (enc->json.flags & F_INDENT)
290	{	447	{
291	need (enc, 1);	448	need (enc, 1);
292	encode_ch (enc, '\n');	449	encode_ch (enc, '\n');
293	}	450	}
294	}	451	}
295		452
296	inline void	453	INLINE void
297	encode_comma (enc_t *enc)	454	encode_comma (enc_t *enc)
298	{	455	{
299	encode_ch (enc, ',');	456	encode_ch (enc, ',');
300		457
301	if (enc->json.flags & F_INDENT)	458	if (enc->json.flags & F_INDENT)
…		…
309	static void	466	static void
310	encode_av (enc_t enc, AV av)	467	encode_av (enc_t enc, AV av)
311	{	468	{
312	int i, len = av_len (av);	469	int i, len = av_len (av);
313		470
314	if (enc->indent >= enc->maxdepth)	471	if (enc->indent >= enc->json.max_depth)
315	croak ("data structure too deep (hit recursion limit)");	472	croak (ERR_NESTING_EXCEEDED);
316		473
317	encode_ch (enc, '['); encode_nl (enc);	474	encode_ch (enc, '[');
318	++enc->indent;	475
		476	if (len >= 0)
		477	{
		478	encode_nl (enc); ++enc->indent;
319		479
320	for (i = 0; i <= len; ++i)	480	for (i = 0; i <= len; ++i)
321	{	481	{
322	SV **svp = av_fetch (av, i, 0);	482	SV **svp = av_fetch (av, i, 0);
323		483
324	encode_indent (enc);	484	encode_indent (enc);
325		485
326	if (svp)	486	if (svp)
327	encode_sv (enc, *svp);	487	encode_sv (enc, *svp);
328	else	488	else
329	encode_str (enc, "null", 4, 0);	489	encode_str (enc, "null", 4, 0);
330		490
331	if (i < len)	491	if (i < len)
332	encode_comma (enc);	492	encode_comma (enc);
333	}	493	}
334		494
		495	encode_nl (enc); --enc->indent; encode_indent (enc);
		496	}
		497
335	encode_nl (enc);	498	encode_ch (enc, ']');
336
337	--enc->indent;
338	encode_indent (enc); encode_ch (enc, ']');
339	}	499	}
340		500
341	static void	501	static void
342	encode_hk (enc_t enc, HE he)	502	encode_hk (enc_t enc, HE he)
343	{	503	{
…		…
391		551
392	static void	552	static void
393	encode_hv (enc_t enc, HV hv)	553	encode_hv (enc_t enc, HV hv)
394	{	554	{
395	HE *he;	555	HE *he;
396	int count;
397		556
398	if (enc->indent >= enc->maxdepth)	557	if (enc->indent >= enc->json.max_depth)
399	croak ("data structure too deep (hit recursion limit)");	558	croak (ERR_NESTING_EXCEEDED);
400		559
401	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	560	encode_ch (enc, '{');
402		561
403	// for canonical output we have to sort by keys first	562	// for canonical output we have to sort by keys first
404	// actually, this is mostly due to the stupid so-called	563	// actually, this is mostly due to the stupid so-called
405	// security workaround added somewhere in 5.8.x.	564	// security workaround added somewhere in 5.8.x
406	// that randomises hash orderings	565	// that randomises hash orderings
407	if (enc->json.flags & F_CANONICAL)	566	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
408	{	567	{
409	int count = hv_iterinit (hv);	568	int count = hv_iterinit (hv);
410		569
411	if (SvMAGICAL (hv))	570	if (SvMAGICAL (hv))
412	{	571	{
…		…
459		618
460	FREETMPS;	619	FREETMPS;
461	LEAVE;	620	LEAVE;
462	}	621	}
463		622
		623	encode_nl (enc); ++enc->indent;
		624
464	while (count--)	625	while (count--)
465	{	626	{
466	encode_indent (enc);	627	encode_indent (enc);
467	he = hes [count];	628	he = hes [count];
468	encode_hk (enc, he);	629	encode_hk (enc, he);
469	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));	630	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
470		631
471	if (count)	632	if (count)
472	encode_comma (enc);	633	encode_comma (enc);
473	}	634	}
		635
		636	encode_nl (enc); --enc->indent; encode_indent (enc);
474	}	637	}
475	}	638	}
476	else	639	else
477	{	640	{
478	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))	641	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))
479	if ((he = hv_iternext (hv)))	642	if ((he = hv_iternext (hv)))
		643	{
		644	encode_nl (enc); ++enc->indent;
		645
480	for (;;)	646	for (;;)
481	{	647	{
482	encode_indent (enc);	648	encode_indent (enc);
483	encode_hk (enc, he);	649	encode_hk (enc, he);
484	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));	650	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
485		651
486	if (!(he = hv_iternext (hv)))	652	if (!(he = hv_iternext (hv)))
487	break;	653	break;
488		654
489	encode_comma (enc);	655	encode_comma (enc);
490	}	656	}
491	}
492		657
		658	encode_nl (enc); --enc->indent; encode_indent (enc);
		659	}
		660	}
		661
493	encode_nl (enc);	662	encode_ch (enc, '}');
494
495	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
496	}	663	}
497		664
498	// encode objects, arrays and special \0=false and \1=true values.	665	// encode objects, arrays and special \0=false and \1=true values.
499	static void	666	static void
500	encode_rv (enc_t enc, SV sv)	667	encode_rv (enc_t enc, SV sv)
…		…
577		744
578	if (len == 1 && *pv == '1')	745	if (len == 1 && *pv == '1')
579	encode_str (enc, "true", 4, 0);	746	encode_str (enc, "true", 4, 0);
580	else if (len == 1 && *pv == '0')	747	else if (len == 1 && *pv == '0')
581	encode_str (enc, "false", 5, 0);	748	encode_str (enc, "false", 5, 0);
		749	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		750	encode_str (enc, "null", 4, 0);
582	else	751	else
583	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	752	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
584	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	753	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
585	}	754	}
		755	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		756	encode_str (enc, "null", 4, 0);
586	else	757	else
587	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	758	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
588	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	759	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
589	}	760	}
590		761
…		…
608	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	779	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
609	enc->cur += strlen (enc->cur);	780	enc->cur += strlen (enc->cur);
610	}	781	}
611	else if (SvIOKp (sv))	782	else if (SvIOKp (sv))
612	{	783	{
613	// we assume we can always read an IV as a UV	784	// we assume we can always read an IV as a UV and vice versa
614	if (SvUV (sv) & ~(UV)0x7fff)	785	// we assume two's complement
615	{	786	// we assume no aliasing issues in the union
616	// large integer, use the (rather slow) snprintf way.	787	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
617	need (enc, sizeof (UV) * 3);	788	: 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	{	789	{
625	// optimise the "small number case"	790	// optimise the "small number case"
626	// code will likely be branchless and use only a single multiplication	791	// code will likely be branchless and use only a single multiplication
		792	// works for numbers up to 59074
627	I32 i = SvIV (sv);	793	I32 i = SvIVX (sv);
628	U32 u;	794	U32 u;
629	char digit, nz = 0;	795	char digit, nz = 0;
630		796
631	need (enc, 6);	797	need (enc, 6);
632		798
…		…
638		804
639	// now output digit by digit, each time masking out the integer part	805	// 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,	806	// and multiplying by 5 while moving the decimal point one to the right,
641	// resulting in a net multiplication by 10.	807	// resulting in a net multiplication by 10.
642	// we always write the digit to memory but conditionally increment	808	// we always write the digit to memory but conditionally increment
643	// the pointer, to ease the usage of conditional move instructions.	809	// 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;	810	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;	811	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;	812	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;	813	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'	814	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
649	}	815	}
		816	else
		817	{
		818	// large integer, use the (rather slow) snprintf way.
		819	need (enc, IVUV_MAXCHARS);
		820	enc->cur +=
		821	SvIsUV(sv)
		822	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
		823	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		824	}
650	}	825	}
651	else if (SvROK (sv))	826	else if (SvROK (sv))
652	encode_rv (enc, SvRV (sv));	827	encode_rv (enc, SvRV (sv));
653	else if (!SvOK (sv))	828	else if (!SvOK (sv) \|\| enc->json.flags & F_ALLOW_UNKNOWN)
654	encode_str (enc, "null", 4, 0);	829	encode_str (enc, "null", 4, 0);
655	else	830	else
656	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	831	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
657	SvPV_nolen (sv), SvFLAGS (sv));	832	SvPV_nolen (sv), SvFLAGS (sv));
658	}	833	}
…		…
668	enc.json = *json;	843	enc.json = *json;
669	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	844	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
670	enc.cur = SvPVX (enc.sv);	845	enc.cur = SvPVX (enc.sv);
671	enc.end = SvEND (enc.sv);	846	enc.end = SvEND (enc.sv);
672	enc.indent = 0;	847	enc.indent = 0;
673	enc.maxdepth = DEC_DEPTH (enc.json.flags);	848	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		849	: enc.json.flags & F_LATIN1 ? 0x000100UL
		850	: 0x110000UL;
674		851
675	SvPOK_only (enc.sv);	852	SvPOK_only (enc.sv);
676	encode_sv (&enc, scalar);	853	encode_sv (&enc, scalar);
		854	encode_nl (&enc);
677		855
678	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	856	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
679	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	857	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
680		858
681	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))	859	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))
…		…
699	JSON json;	877	JSON json;
700	U32 depth; // recursion depth	878	U32 depth; // recursion depth
701	U32 maxdepth; // recursion depth limit	879	U32 maxdepth; // recursion depth limit
702	} dec_t;	880	} dec_t;
703		881
704	inline void	882	INLINE void
705	decode_comment (dec_t *dec)	883	decode_comment (dec_t *dec)
706	{	884	{
707	// only '#'-style comments allowed a.t.m.	885	// only '#'-style comments allowed a.t.m.
708		886
709	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)	887	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)
710	++dec->cur;	888	++dec->cur;
711	}	889	}
712		890
713	inline void	891	INLINE void
714	decode_ws (dec_t *dec)	892	decode_ws (dec_t *dec)
715	{	893	{
716	for (;;)	894	for (;;)
717	{	895	{
718	char ch = *dec->cur;	896	char ch = *dec->cur;
…		…
742	if (*dec->cur != ch) \	920	if (*dec->cur != ch) \
743	ERR (# ch " expected"); \	921	ERR (# ch " expected"); \
744	++dec->cur; \	922	++dec->cur; \
745	SE	923	SE
746		924
747	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	925	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
748	#define DEC_DEC_DEPTH --dec->depth	926	#define DEC_DEC_DEPTH --dec->depth
749		927
750	static SV decode_sv (dec_t dec);	928	static SV decode_sv (dec_t dec);
751		929
752	static signed char decode_hexdigit[256];	930	static signed char decode_hexdigit[256];
…		…
844		1022
845	if (hi >= 0x80)	1023	if (hi >= 0x80)
846	{	1024	{
847	utf8 = 1;	1025	utf8 = 1;
848		1026
849	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	1027	cur = encode_utf8 (cur, hi);
850	}	1028	}
851	else	1029	else
852	*cur++ = hi;	1030	*cur++ = hi;
853	}	1031	}
854	break;	1032	break;
…		…
856	default:	1034	default:
857	--dec_cur;	1035	--dec_cur;
858	ERR ("illegal backslash escape sequence in string");	1036	ERR ("illegal backslash escape sequence in string");
859	}	1037	}
860	}	1038	}
861	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	1039	else if (expect_true (ch >= 0x20 && ch < 0x80))
862	*cur++ = ch;	1040	*cur++ = ch;
863	else if (ch >= 0x80)	1041	else if (ch >= 0x80)
864	{	1042	{
865	STRLEN clen;	1043	STRLEN clen;
866	UV uch;
867		1044
868	--dec_cur;	1045	--dec_cur;
869		1046
870	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);	1047	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
871	if (clen == (STRLEN)-1)	1048	if (clen == (STRLEN)-1)
872	ERR ("malformed UTF-8 character in JSON string");	1049	ERR ("malformed UTF-8 character in JSON string");
873		1050
874	do	1051	do
875	cur++ = dec_cur++;	1052	cur++ = dec_cur++;
…		…
892	{	1069	{
893	STRLEN len = cur - buf;	1070	STRLEN len = cur - buf;
894		1071
895	if (sv)	1072	if (sv)
896	{	1073	{
897	SvGROW (sv, SvCUR (sv) + len + 1);	1074	STRLEN cur = SvCUR (sv);
		1075
		1076	if (SvLEN (sv) <= cur + len)
		1077	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1078
898	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	1079	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
899	SvCUR_set (sv, SvCUR (sv) + len);	1080	SvCUR_set (sv, SvCUR (sv) + len);
900	}	1081	}
901	else	1082	else
902	sv = newSVpvn (buf, len);	1083	sv = newSVpvn (buf, len);
…		…
989		1170
990	if (!is_nv)	1171	if (!is_nv)
991	{	1172	{
992	int len = dec->cur - start;	1173	int len = dec->cur - start;
993		1174
994	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1175	// special case the rather common 1..5-digit-int case
995	if (*start == '-')	1176	if (*start == '-')
996	switch (len)	1177	switch (len)
997	{	1178	{
998	case 2: return newSViv (-( start [1] - '0' * 1));	1179	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
999	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1180	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1000	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));	1181	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1001	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));	1182	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1183	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1002	}	1184	}
1003	else	1185	else
1004	switch (len)	1186	switch (len)
1005	{	1187	{
1006	case 1: return newSViv ( start [0] - '0' * 1);	1188	case 1: return newSViv ( start [0] - '0' * 1);
1007	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1189	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1008	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);	1190	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1009	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);	1191	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1192	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1010	}	1193	}
1011		1194
1012	{	1195	{
1013	UV uv;	1196	UV uv;
1014	int numtype = grok_number (start, len, &uv);	1197	int numtype = grok_number (start, len, &uv);
…		…
1023	}	1206	}
1024		1207
1025	len -= *start == '-' ? 1 : 0;	1208	len -= *start == '-' ? 1 : 0;
1026		1209
1027	// does not fit into IV or UV, try NV	1210	// does not fit into IV or UV, try NV
1028	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)	1211	if (len <= NV_DIG)
1029	#if defined (LDBL_DIG)
1030	\|\| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1031	#endif
1032	)
1033	// fits into NV without loss of precision	1212	// fits into NV without loss of precision
1034	return newSVnv (Atof (start));	1213	return newSVnv (json_atof (start));
1035		1214
1036	// everything else fails, convert it to a string	1215	// everything else fails, convert it to a string
1037	return newSVpvn (start, dec->cur - start);	1216	return newSVpvn (start, dec->cur - start);
1038	}	1217	}
1039		1218
1040	// loss of precision here	1219	// loss of precision here
1041	return newSVnv (Atof (start));	1220	return newSVnv (json_atof (start));
1042		1221
1043	fail:	1222	fail:
1044	return 0;	1223	return 0;
1045	}	1224	}
1046		1225
…		…
1122	char *p = dec->cur;	1301	char *p = dec->cur;
1123	char *e = p + 24; // only try up to 24 bytes	1302	char *e = p + 24; // only try up to 24 bytes
1124		1303
1125	for (;;)	1304	for (;;)
1126	{	1305	{
1127	// the >= 0x80 is true on most architectures	1306	// the >= 0x80 is false on most architectures
1128	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')	1307	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
1129	{	1308	{
1130	// slow path, back up and use decode_str	1309	// slow path, back up and use decode_str
1131	SV *key = decode_str (dec);	1310	SV *key = decode_str (dec);
1132	if (!key)	1311	if (!key)
…		…
1263		1442
1264	static SV *	1443	static SV *
1265	decode_sv (dec_t *dec)	1444	decode_sv (dec_t *dec)
1266	{	1445	{
1267	// the beauty of JSON: you need exactly one character lookahead	1446	// the beauty of JSON: you need exactly one character lookahead
1268	// to parse anything.	1447	// to parse everything.
1269	switch (*dec->cur)	1448	switch (*dec->cur)
1270	{	1449	{
1271	case '"': ++dec->cur; return decode_str (dec);	1450	case '"': ++dec->cur; return decode_str (dec);
1272	case '[': ++dec->cur; return decode_av (dec);	1451	case '[': ++dec->cur; return decode_av (dec);
1273	case '{': ++dec->cur; return decode_hv (dec);	1452	case '{': ++dec->cur; return decode_hv (dec);
1274		1453
1275	case '-':	1454	case '-':
1276	case '0': case '1': case '2': case '3': case '4':	1455	case '0': case '1': case '2': case '3': case '4':
1277	case '5': case '6': case '7': case '8': case '9':	1456	case '5': case '6': case '7': case '8': case '9':
1278	return decode_num (dec);	1457	return decode_num (dec);
…		…
1280	case 't':	1459	case 't':
1281	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1460	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1282	{	1461	{
1283	dec->cur += 4;	1462	dec->cur += 4;
1284	#if JSON_SLOW	1463	#if JSON_SLOW
1285	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);	1464	json_true = get_bool ("JSON::XS::true");
1286	#endif	1465	#endif
1287	return SvREFCNT_inc (json_true);	1466	return newSVsv (json_true);
1288	}	1467	}
1289	else	1468	else
1290	ERR ("'true' expected");	1469	ERR ("'true' expected");
1291		1470
1292	break;	1471	break;
…		…
1294	case 'f':	1473	case 'f':
1295	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1474	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1296	{	1475	{
1297	dec->cur += 5;	1476	dec->cur += 5;
1298	#if JSON_SLOW	1477	#if JSON_SLOW
1299	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1478	json_false = get_bool ("JSON::XS::false");
1300	#endif	1479	#endif
1301	return SvREFCNT_inc (json_false);	1480	return newSVsv (json_false);
1302	}	1481	}
1303	else	1482	else
1304	ERR ("'false' expected");	1483	ERR ("'false' expected");
1305		1484
1306	break;	1485	break;
…		…
1324	fail:	1503	fail:
1325	return 0;	1504	return 0;
1326	}	1505	}
1327		1506
1328	static SV *	1507	static SV *
1329	decode_json (SV string, JSON json, UV *offset_return)	1508	decode_json (SV string, JSON json, char **offset_return)
1330	{	1509	{
1331	dec_t dec;	1510	dec_t dec;
1332	UV offset;
1333	SV *sv;	1511	SV *sv;
1334		1512
		1513	/* work around bugs in 5.10 where manipulating magic values
		1514	* will perl ignore the magic in subsequent accesses
		1515	*/
1335	SvGETMAGIC (string);	1516	/SvGETMAGIC (string);/
		1517	if (SvMAGICAL (string))
		1518	string = sv_2mortal (newSVsv (string));
		1519
1336	SvUPGRADE (string, SVt_PV);	1520	SvUPGRADE (string, SVt_PV);
1337		1521
1338	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))	1522	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1523	* assertion with -DDEBUGGING, although SvCUR is documented to
		1524	* return the xpv_cur field which certainly exists after upgrading.
		1525	* according to nicholas clark, calling SvPOK fixes this.
		1526	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1527	* and hope for the best.
		1528	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1529	* assertion business is seriously broken, try yet another workaround
		1530	* for the broken -DDEBUGGING.
		1531	*/
		1532	{
		1533	#ifdef DEBUGGING
		1534	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1535	#else
		1536	STRLEN offset = SvCUR (string);
		1537	#endif
		1538
		1539	if (offset > json->max_size && json->max_size)
1339	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",	1540	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1340	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));	1541	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1542	}
1341		1543
1342	if (json->flags & F_UTF8)	1544	if (json->flags & F_UTF8)
1343	sv_utf8_downgrade (string, 0);	1545	sv_utf8_downgrade (string, 0);
1344	else	1546	else
1345	sv_utf8_upgrade (string);	1547	sv_utf8_upgrade (string);
1346		1548
1347	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1549	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1348		1550
1349	dec.json = *json;	1551	dec.json = *json;
1350	dec.cur = SvPVX (string);	1552	dec.cur = SvPVX (string);
1351	dec.end = SvEND (string);	1553	dec.end = SvEND (string);
1352	dec.err = 0;	1554	dec.err = 0;
1353	dec.depth = 0;	1555	dec.depth = 0;
1354	dec.maxdepth = DEC_DEPTH (dec.json.flags);
1355		1556
1356	if (dec.json.cb_object \|\| dec.json.cb_sk_object)	1557	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
1357	dec.json.flags \|= F_HOOK;	1558	dec.json.flags \|= F_HOOK;
1358		1559
1359	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1560	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
1360		1561
1361	decode_ws (&dec);	1562	decode_ws (&dec);
1362	sv = decode_sv (&dec);	1563	sv = decode_sv (&dec);
		1564
		1565	if (offset_return)
		1566	*offset_return = dec.cur;
1363		1567
1364	if (!(offset_return \|\| !sv))	1568	if (!(offset_return \|\| !sv))
1365	{	1569	{
1366	// check for trailing garbage	1570	// check for trailing garbage
1367	decode_ws (&dec);	1571	decode_ws (&dec);
…		…
1370	{	1574	{
1371	dec.err = "garbage after JSON object";	1575	dec.err = "garbage after JSON object";
1372	SvREFCNT_dec (sv);	1576	SvREFCNT_dec (sv);
1373	sv = 0;	1577	sv = 0;
1374	}	1578	}
1375	}
1376
1377	if (offset_return \|\| !sv)
1378	{
1379	offset = dec.json.flags & F_UTF8
1380	? dec.cur - SvPVX (string)
1381	: utf8_distance (dec.cur, SvPVX (string));
1382
1383	if (offset_return)
1384	*offset_return = offset;
1385	}	1579	}
1386		1580
1387	if (!sv)	1581	if (!sv)
1388	{	1582	{
1389	SV *uni = sv_newmortal ();	1583	SV *uni = sv_newmortal ();
…		…
1395	SAVEVPTR (PL_curcop);	1589	SAVEVPTR (PL_curcop);
1396	PL_curcop = &cop;	1590	PL_curcop = &cop;
1397	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1591	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1398	LEAVE;	1592	LEAVE;
1399		1593
1400	croak ("%s, at character offset %d [\"%s\"]",	1594	croak ("%s, at character offset %d (before \"%s\")",
1401	dec.err,	1595	dec.err,
1402	(int)offset,	1596	ptr_to_index (string, dec.cur),
1403	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1597	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1404	}	1598	}
1405		1599
1406	sv = sv_2mortal (sv);	1600	sv = sv_2mortal (sv);
1407		1601
1408	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))	1602	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1409	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1603	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1410		1604
1411	return sv;	1605	return sv;
		1606	}
		1607
		1608	/////////////////////////////////////////////////////////////////////////////
		1609	// incremental parser
		1610
		1611	static void
		1612	incr_parse (JSON *self)
		1613	{
		1614	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1615
		1616	// the state machine here is a bit convoluted and could be simplified a lot
		1617	// but this would make it slower, so...
		1618
		1619	for (;;)
		1620	{
		1621	//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
		1622	switch (self->incr_mode)
		1623	{
		1624	// only used for initial whitespace skipping
		1625	case INCR_M_WS:
		1626	for (;;)
		1627	{
		1628	if (*p > 0x20)
		1629	{
		1630	if (*p == '#')
		1631	{
		1632	self->incr_mode = INCR_M_C0;
		1633	goto incr_m_c;
		1634	}
		1635	else
		1636	{
		1637	self->incr_mode = INCR_M_JSON;
		1638	goto incr_m_json;
		1639	}
		1640	}
		1641	else if (!*p)
		1642	goto interrupt;
		1643
		1644	++p;
		1645	}
		1646
		1647	// skip a single char inside a string (for \\-processing)
		1648	case INCR_M_BS:
		1649	if (!*p)
		1650	goto interrupt;
		1651
		1652	++p;
		1653	self->incr_mode = INCR_M_STR;
		1654	goto incr_m_str;
		1655
		1656	// inside #-style comments
		1657	case INCR_M_C0:
		1658	case INCR_M_C1:
		1659	incr_m_c:
		1660	for (;;)
		1661	{
		1662	if (*p == '\n')
		1663	{
		1664	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1665	break;
		1666	}
		1667	else if (!*p)
		1668	goto interrupt;
		1669
		1670	++p;
		1671	}
		1672
		1673	break;
		1674
		1675	// inside a string
		1676	case INCR_M_STR:
		1677	incr_m_str:
		1678	for (;;)
		1679	{
		1680	if (*p == '"')
		1681	{
		1682	++p;
		1683	self->incr_mode = INCR_M_JSON;
		1684
		1685	if (!self->incr_nest)
		1686	goto interrupt;
		1687
		1688	goto incr_m_json;
		1689	}
		1690	else if (*p == '\\')
		1691	{
		1692	++p; // "virtually" consumes character after \
		1693
		1694	if (!*p) // if at end of string we have to switch modes
		1695	{
		1696	self->incr_mode = INCR_M_BS;
		1697	goto interrupt;
		1698	}
		1699	}
		1700	else if (!*p)
		1701	goto interrupt;
		1702
		1703	++p;
		1704	}
		1705
		1706	// after initial ws, outside string
		1707	case INCR_M_JSON:
		1708	incr_m_json:
		1709	for (;;)
		1710	{
		1711	switch (*p++)
		1712	{
		1713	case 0:
		1714	--p;
		1715	goto interrupt;
		1716
		1717	case 0x09:
		1718	case 0x0a:
		1719	case 0x0d:
		1720	case 0x20:
		1721	if (!self->incr_nest)
		1722	{
		1723	--p; // do not eat the whitespace, let the next round do it
		1724	goto interrupt;
		1725	}
		1726	break;
		1727
		1728	case '"':
		1729	self->incr_mode = INCR_M_STR;
		1730	goto incr_m_str;
		1731
		1732	case '[':
		1733	case '{':
		1734	if (++self->incr_nest > self->max_depth)
		1735	croak (ERR_NESTING_EXCEEDED);
		1736	break;
		1737
		1738	case ']':
		1739	case '}':
		1740	if (--self->incr_nest <= 0)
		1741	goto interrupt;
		1742	break;
		1743
		1744	case '#':
		1745	self->incr_mode = INCR_M_C1;
		1746	goto incr_m_c;
		1747	}
		1748	}
		1749	}
		1750
		1751	modechange:
		1752	;
		1753	}
		1754
		1755	interrupt:
		1756	self->incr_pos = p - SvPVX (self->incr_text);
		1757	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1758	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1412	}	1759	}
1413		1760
1414	/////////////////////////////////////////////////////////////////////////////	1761	/////////////////////////////////////////////////////////////////////////////
1415	// XS interface functions	1762	// XS interface functions
1416		1763
…		…
1428	: -1;	1775	: -1;
1429		1776
1430	json_stash = gv_stashpv ("JSON::XS" , 1);	1777	json_stash = gv_stashpv ("JSON::XS" , 1);
1431	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);	1778	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1432		1779
1433	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );	1780	json_true = get_bool ("JSON::XS::true");
1434	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1781	json_false = get_bool ("JSON::XS::false");
		1782
		1783	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1435	}	1784	}
1436		1785
1437	PROTOTYPES: DISABLE	1786	PROTOTYPES: DISABLE
1438		1787
1439	void CLONE (...)	1788	void CLONE (...)
…		…
1444	void new (char *klass)	1793	void new (char *klass)
1445	PPCODE:	1794	PPCODE:
1446	{	1795	{
1447	SV *pv = NEWSV (0, sizeof (JSON));	1796	SV *pv = NEWSV (0, sizeof (JSON));
1448	SvPOK_only (pv);	1797	SvPOK_only (pv);
1449	Zero (SvPVX (pv), 1, JSON);	1798	json_init ((JSON *)SvPVX (pv));
1450	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	1799	XPUSHs (sv_2mortal (sv_bless (
1451	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH)));	1800	newRV_noinc (pv),
		1801	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1802	)));
1452	}	1803	}
1453		1804
1454	void ascii (JSON *self, int enable = 1)	1805	void ascii (JSON *self, int enable = 1)
1455	ALIAS:	1806	ALIAS:
1456	ascii = F_ASCII	1807	ascii = F_ASCII
…		…
1464	allow_nonref = F_ALLOW_NONREF	1815	allow_nonref = F_ALLOW_NONREF
1465	shrink = F_SHRINK	1816	shrink = F_SHRINK
1466	allow_blessed = F_ALLOW_BLESSED	1817	allow_blessed = F_ALLOW_BLESSED
1467	convert_blessed = F_CONV_BLESSED	1818	convert_blessed = F_CONV_BLESSED
1468	relaxed = F_RELAXED	1819	relaxed = F_RELAXED
		1820	allow_unknown = F_ALLOW_UNKNOWN
1469	PPCODE:	1821	PPCODE:
1470	{	1822	{
1471	if (enable)	1823	if (enable)
1472	self->flags \|= ix;	1824	self->flags \|= ix;
1473	else	1825	else
1474	self->flags &= ~ix;	1826	self->flags &= ~ix;
1475		1827
1476	XPUSHs (ST (0));	1828	XPUSHs (ST (0));
1477	}	1829	}
1478		1830
1479	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1831	void get_ascii (JSON *self)
		1832	ALIAS:
		1833	get_ascii = F_ASCII
		1834	get_latin1 = F_LATIN1
		1835	get_utf8 = F_UTF8
		1836	get_indent = F_INDENT
		1837	get_canonical = F_CANONICAL
		1838	get_space_before = F_SPACE_BEFORE
		1839	get_space_after = F_SPACE_AFTER
		1840	get_allow_nonref = F_ALLOW_NONREF
		1841	get_shrink = F_SHRINK
		1842	get_allow_blessed = F_ALLOW_BLESSED
		1843	get_convert_blessed = F_CONV_BLESSED
		1844	get_relaxed = F_RELAXED
		1845	get_allow_unknown = F_ALLOW_UNKNOWN
1480	PPCODE:	1846	PPCODE:
1481	{	1847	XPUSHs (boolSV (self->flags & ix));
1482	UV log2 = 0;
1483		1848
1484	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	1849	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1485		1850	PPCODE:
1486	while ((1UL << log2) < max_depth)	1851	self->max_depth = max_depth;
1487	++log2;
1488
1489	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1490
1491	XPUSHs (ST (0));	1852	XPUSHs (ST (0));
1492	}
1493		1853
		1854	U32 get_max_depth (JSON *self)
		1855	CODE:
		1856	RETVAL = self->max_depth;
		1857	OUTPUT:
		1858	RETVAL
		1859
1494	void max_size (JSON *self, UV max_size = 0)	1860	void max_size (JSON *self, U32 max_size = 0)
1495	PPCODE:	1861	PPCODE:
1496	{	1862	self->max_size = max_size;
1497	UV log2 = 0;
1498
1499	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1500	if (max_size == 1) max_size = 2;
1501
1502	while ((1UL << log2) < max_size)
1503	++log2;
1504
1505	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
1506
1507	XPUSHs (ST (0));	1863	XPUSHs (ST (0));
1508	}	1864
		1865	int get_max_size (JSON *self)
		1866	CODE:
		1867	RETVAL = self->max_size;
		1868	OUTPUT:
		1869	RETVAL
1509		1870
1510	void filter_json_object (JSON self, SV cb = &PL_sv_undef)	1871	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1511	PPCODE:	1872	PPCODE:
1512	{	1873	{
1513	SvREFCNT_dec (self->cb_object);	1874	SvREFCNT_dec (self->cb_object);
…		…
1547	XPUSHs (decode_json (jsonstr, self, 0));	1908	XPUSHs (decode_json (jsonstr, self, 0));
1548		1909
1549	void decode_prefix (JSON self, SV jsonstr)	1910	void decode_prefix (JSON self, SV jsonstr)
1550	PPCODE:	1911	PPCODE:
1551	{	1912	{
1552	UV offset;	1913	char *offset;
1553	EXTEND (SP, 2);	1914	EXTEND (SP, 2);
1554	PUSHs (decode_json (jsonstr, self, &offset));	1915	PUSHs (decode_json (jsonstr, self, &offset));
1555	PUSHs (sv_2mortal (newSVuv (offset)));	1916	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		1917	}
		1918
		1919	void incr_parse (JSON self, SV jsonstr = 0)
		1920	PPCODE:
		1921	{
		1922	if (!self->incr_text)
		1923	self->incr_text = newSVpvn ("", 0);
		1924
		1925	// append data, if any
		1926	if (jsonstr)
		1927	{
		1928	if (SvUTF8 (jsonstr))
		1929	{
		1930	if (!SvUTF8 (self->incr_text))
		1931	{
		1932	/* utf-8-ness differs, need to upgrade */
		1933	sv_utf8_upgrade (self->incr_text);
		1934
		1935	if (self->incr_pos)
		1936	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1937	- (U8 *)SvPVX (self->incr_text);
		1938	}
		1939	}
		1940	else if (SvUTF8 (self->incr_text))
		1941	sv_utf8_upgrade (jsonstr);
		1942
		1943	{
		1944	STRLEN len;
		1945	const char *str = SvPV (jsonstr, len);
		1946	STRLEN cur = SvCUR (self->incr_text);
		1947
		1948	if (SvLEN (self->incr_text) <= cur + len)
		1949	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1950
		1951	Move (str, SvEND (self->incr_text), len, char);
		1952	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1953	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1954	}
		1955	}
		1956
		1957	if (GIMME_V != G_VOID)
		1958	do
		1959	{
		1960	char *offset;
		1961
		1962	if (!INCR_DONE (self))
		1963	{
		1964	incr_parse (self);
		1965
		1966	if (self->incr_pos > self->max_size && self->max_size)
		1967	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1968	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1969
		1970	if (!INCR_DONE (self))
		1971	break;
		1972	}
		1973
		1974	XPUSHs (decode_json (self->incr_text, self, &offset));
		1975
		1976	self->incr_pos -= offset - SvPVX (self->incr_text);
		1977	self->incr_nest = 0;
		1978	self->incr_mode = 0;
		1979
		1980	sv_chop (self->incr_text, offset);
		1981	}
		1982	while (GIMME_V == G_ARRAY);
		1983	}
		1984
		1985	SV incr_text (JSON self)
		1986	ATTRS: lvalue
		1987	CODE:
		1988	{
		1989	if (self->incr_pos)
		1990	croak ("incr_text can not be called when the incremental parser already started parsing");
		1991
		1992	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		1993	}
		1994	OUTPUT:
		1995	RETVAL
		1996
		1997	void incr_skip (JSON *self)
		1998	CODE:
		1999	{
		2000	if (self->incr_pos)
		2001	{
		2002	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2003	self->incr_pos = 0;
		2004	self->incr_nest = 0;
		2005	self->incr_mode = 0;
		2006	}
		2007	}
		2008
		2009	void incr_reset (JSON *self)
		2010	CODE:
		2011	{
		2012	SvREFCNT_dec (self->incr_text);
		2013	self->incr_text = 0;
		2014	self->incr_pos = 0;
		2015	self->incr_nest = 0;
		2016	self->incr_mode = 0;
1556	}	2017	}
1557		2018
1558	void DESTROY (JSON *self)	2019	void DESTROY (JSON *self)
1559	CODE:	2020	CODE:
1560	SvREFCNT_dec (self->cb_sk_object);	2021	SvREFCNT_dec (self->cb_sk_object);
1561	SvREFCNT_dec (self->cb_object);	2022	SvREFCNT_dec (self->cb_object);
		2023	SvREFCNT_dec (self->incr_text);
1562		2024
1563	PROTOTYPES: ENABLE	2025	PROTOTYPES: ENABLE
1564		2026
1565	void to_json (SV *scalar)	2027	void encode_json (SV *scalar)
		2028	ALIAS:
		2029	to_json_ = 0
		2030	encode_json = F_UTF8
1566	PPCODE:	2031	PPCODE:
1567	{	2032	{
1568	JSON json = { F_DEFAULT \| F_UTF8 };	2033	JSON json;
		2034	json_init (&json);
		2035	json.flags \|= ix;
1569	XPUSHs (encode_json (scalar, &json));	2036	XPUSHs (encode_json (scalar, &json));
1570	}	2037	}
1571		2038
1572	void from_json (SV *jsonstr)	2039	void decode_json (SV *jsonstr)
		2040	ALIAS:
		2041	from_json_ = 0
		2042	decode_json = F_UTF8
1573	PPCODE:	2043	PPCODE:
1574	{	2044	{
1575	JSON json = { F_DEFAULT \| F_UTF8 };	2045	JSON json;
		2046	json_init (&json);
		2047	json.flags \|= ix;
1576	XPUSHs (decode_json (jsonstr, &json, 0));	2048	XPUSHs (decode_json (jsonstr, &json, 0));
1577	}	2049	}
1578		2050

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.64 by root, Mon Aug 27 02:03:23 2007 UTC vs. Revision 1.105 by root, Tue Jan 19 01:07:27 2010 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.64 by root, Mon Aug 27 02:03:23 2007 UTC vs.
Revision 1.105 by root, Tue Jan 19 01:07:27 2010 UTC