[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 1.107 by root, Wed Mar 17 01:45:43 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, int maxdepth)
		198	{
		199	UV uaccum = 0;
		200	int eaccum = 0;
		201
		202	// if we recurse too deep, skip all remaining digits
		203	// to avoid a stack overflow attack
		204	if (expect_false (--maxdepth <= 0))
		205	while (((U8)*s - '0') < 10)
		206	++s;
		207
		208	for (;;)
118	}	209	{
119	}	210	U8 dig = (U8)*s - '0';
120		211
		212	if (expect_false (dig >= 10))
		213	{
		214	if (dig == (U8)((U8)'.' - (U8)'0'))
		215	{
		216	++s;
		217	json_atof_scan1 (s, accum, expo, 1, maxdepth);
		218	}
		219	else if ((dig \| ' ') == 'e' - '0')
		220	{
		221	int exp2 = 0;
		222	int neg = 0;
		223
		224	++s;
		225
		226	if (*s == '-')
		227	{
		228	++s;
		229	neg = 1;
		230	}
		231	else if (*s == '+')
		232	++s;
		233
		234	while ((dig = (U8)*s - '0') < 10)
		235	exp2 = exp2 * 10 + *s++ - '0';
		236
		237	*expo += neg ? -exp2 : exp2;
		238	}
		239
		240	break;
		241	}
		242
		243	++s;
		244
		245	uaccum = uaccum * 10 + dig;
		246	++eaccum;
		247
		248	// if we have too many digits, then recurse for more
		249	// we actually do this for rather few digits
		250	if (uaccum >= (UV_MAX - 9) / 10)
		251	{
		252	if (postdp) *expo -= eaccum;
		253	json_atof_scan1 (s, accum, expo, postdp, maxdepth);
		254	if (postdp) *expo += eaccum;
		255
		256	break;
		257	}
		258	}
		259
		260	// this relies greatly on the quality of the pow ()
		261	// implementation of the platform, but a good
		262	// implementation is hard to beat.
		263	if (postdp) *expo -= eaccum;
		264	accum += uaccum Perl_pow (10., *expo);
		265	*expo += eaccum;
		266	}
		267
		268	static NV
		269	json_atof (const char *s)
		270	{
		271	NV accum = 0.;
		272	int expo = 0;
		273	int neg = 0;
		274
		275	if (*s == '-')
		276	{
		277	++s;
		278	neg = 1;
		279	}
		280
		281	// a recursion depth of ten gives us >>500 bits
		282	json_atof_scan1 (s, &accum, &expo, 0, 10);
		283
		284	return neg ? -accum : accum;
		285	}
121	/////////////////////////////////////////////////////////////////////////////	286	/////////////////////////////////////////////////////////////////////////////
122	// encoder	287	// encoder
123		288
124	// structure used for encoding JSON	289	// structure used for encoding JSON
125	typedef struct	290	typedef struct
…		…
127	char *cur; // SvPVX (sv) + current output position	292	char *cur; // SvPVX (sv) + current output position
128	char *end; // SvEND (sv)	293	char *end; // SvEND (sv)
129	SV *sv; // result scalar	294	SV *sv; // result scalar
130	JSON json;	295	JSON json;
131	U32 indent; // indentation level	296	U32 indent; // indentation level
132	U32 maxdepth; // max. indentation/recursion level	297	UV limit; // escape character values >= this value when encoding
133	} enc_t;	298	} enc_t;
134		299
135	inline void	300	INLINE void
136	need (enc_t *enc, STRLEN len)	301	need (enc_t *enc, STRLEN len)
137	{	302	{
138	if (expect_false (enc->cur + len >= enc->end))	303	if (expect_false (enc->cur + len >= enc->end))
139	{	304	{
140	STRLEN cur = enc->cur - SvPVX (enc->sv);	305	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
141	SvGROW (enc->sv, cur + len + 1);	306	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
142	enc->cur = SvPVX (enc->sv) + cur;	307	enc->cur = SvPVX (enc->sv) + cur;
143	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	308	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
144	}	309	}
145	}	310	}
146		311
147	inline void	312	INLINE void
148	encode_ch (enc_t *enc, char ch)	313	encode_ch (enc_t *enc, char ch)
149	{	314	{
150	need (enc, 1);	315	need (enc, 1);
151	*enc->cur++ = ch;	316	*enc->cur++ = ch;
152	}	317	}
…		…
206	{	371	{
207	uch = ch;	372	uch = ch;
208	clen = 1;	373	clen = 1;
209	}	374	}
210		375
211	if (uch > 0x10FFFFUL)	376	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	{	377	{
216	if (uch > 0xFFFFUL)	378	if (uch >= 0x10000UL)
217	{	379	{
		380	if (uch >= 0x110000UL)
		381	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		382
218	need (enc, len += 11);	383	need (enc, len += 11);
219	sprintf (enc->cur, "\\u%04x\\u%04x",	384	sprintf (enc->cur, "\\u%04x\\u%04x",
220	(int)((uch - 0x10000) / 0x400 + 0xD800),	385	(int)((uch - 0x10000) / 0x400 + 0xD800),
221	(int)((uch - 0x10000) % 0x400 + 0xDC00));	386	(int)((uch - 0x10000) % 0x400 + 0xDC00));
222	enc->cur += 12;	387	enc->cur += 12;
223	}	388	}
224	else	389	else
225	{	390	{
226	static char hexdigit [16] = "0123456789abcdef";
227	need (enc, len += 5);	391	need (enc, len += 5);
228	*enc->cur++ = '\\';	392	*enc->cur++ = '\\';
229	*enc->cur++ = 'u';	393	*enc->cur++ = 'u';
230	*enc->cur++ = hexdigit [ uch >> 12 ];	394	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
231	*enc->cur++ = hexdigit [(uch >> 8) & 15];	395	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
232	*enc->cur++ = hexdigit [(uch >> 4) & 15];	396	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
233	*enc->cur++ = hexdigit [(uch >> 0) & 15];	397	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
234	}	398	}
235		399
236	str += clen;	400	str += clen;
237	}	401	}
238	else if (enc->json.flags & F_LATIN1)	402	else if (enc->json.flags & F_LATIN1)
…		…
250	while (--clen);	414	while (--clen);
251	}	415	}
252	else	416	else
253	{	417	{
254	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	418	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
255	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	419	enc->cur = encode_utf8 (enc->cur, uch);
256	++str;	420	++str;
257	}	421	}
258	}	422	}
259	}	423	}
260	}	424	}
261		425
262	--len;	426	--len;
263	}	427	}
264	}	428	}
265		429
266	inline void	430	INLINE void
267	encode_indent (enc_t *enc)	431	encode_indent (enc_t *enc)
268	{	432	{
269	if (enc->json.flags & F_INDENT)	433	if (enc->json.flags & F_INDENT)
270	{	434	{
271	int spaces = enc->indent * INDENT_STEP;	435	int spaces = enc->indent * INDENT_STEP;
…		…
274	memset (enc->cur, ' ', spaces);	438	memset (enc->cur, ' ', spaces);
275	enc->cur += spaces;	439	enc->cur += spaces;
276	}	440	}
277	}	441	}
278		442
279	inline void	443	INLINE void
280	encode_space (enc_t *enc)	444	encode_space (enc_t *enc)
281	{	445	{
282	need (enc, 1);	446	need (enc, 1);
283	encode_ch (enc, ' ');	447	encode_ch (enc, ' ');
284	}	448	}
285		449
286	inline void	450	INLINE void
287	encode_nl (enc_t *enc)	451	encode_nl (enc_t *enc)
288	{	452	{
289	if (enc->json.flags & F_INDENT)	453	if (enc->json.flags & F_INDENT)
290	{	454	{
291	need (enc, 1);	455	need (enc, 1);
292	encode_ch (enc, '\n');	456	encode_ch (enc, '\n');
293	}	457	}
294	}	458	}
295		459
296	inline void	460	INLINE void
297	encode_comma (enc_t *enc)	461	encode_comma (enc_t *enc)
298	{	462	{
299	encode_ch (enc, ',');	463	encode_ch (enc, ',');
300		464
301	if (enc->json.flags & F_INDENT)	465	if (enc->json.flags & F_INDENT)
…		…
309	static void	473	static void
310	encode_av (enc_t enc, AV av)	474	encode_av (enc_t enc, AV av)
311	{	475	{
312	int i, len = av_len (av);	476	int i, len = av_len (av);
313		477
314	if (enc->indent >= enc->maxdepth)	478	if (enc->indent >= enc->json.max_depth)
315	croak ("data structure too deep (hit recursion limit)");	479	croak (ERR_NESTING_EXCEEDED);
316		480
317	encode_ch (enc, '['); encode_nl (enc);	481	encode_ch (enc, '[');
318	++enc->indent;	482
		483	if (len >= 0)
		484	{
		485	encode_nl (enc); ++enc->indent;
319		486
320	for (i = 0; i <= len; ++i)	487	for (i = 0; i <= len; ++i)
321	{	488	{
322	SV **svp = av_fetch (av, i, 0);	489	SV **svp = av_fetch (av, i, 0);
323		490
324	encode_indent (enc);	491	encode_indent (enc);
325		492
326	if (svp)	493	if (svp)
327	encode_sv (enc, *svp);	494	encode_sv (enc, *svp);
328	else	495	else
329	encode_str (enc, "null", 4, 0);	496	encode_str (enc, "null", 4, 0);
330		497
331	if (i < len)	498	if (i < len)
332	encode_comma (enc);	499	encode_comma (enc);
333	}	500	}
334		501
		502	encode_nl (enc); --enc->indent; encode_indent (enc);
		503	}
		504
335	encode_nl (enc);	505	encode_ch (enc, ']');
336
337	--enc->indent;
338	encode_indent (enc); encode_ch (enc, ']');
339	}	506	}
340		507
341	static void	508	static void
342	encode_hk (enc_t enc, HE he)	509	encode_hk (enc_t enc, HE he)
343	{	510	{
…		…
391		558
392	static void	559	static void
393	encode_hv (enc_t enc, HV hv)	560	encode_hv (enc_t enc, HV hv)
394	{	561	{
395	HE *he;	562	HE *he;
396	int count;
397		563
398	if (enc->indent >= enc->maxdepth)	564	if (enc->indent >= enc->json.max_depth)
399	croak ("data structure too deep (hit recursion limit)");	565	croak (ERR_NESTING_EXCEEDED);
400		566
401	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	567	encode_ch (enc, '{');
402		568
403	// for canonical output we have to sort by keys first	569	// for canonical output we have to sort by keys first
404	// actually, this is mostly due to the stupid so-called	570	// actually, this is mostly due to the stupid so-called
405	// security workaround added somewhere in 5.8.x.	571	// security workaround added somewhere in 5.8.x
406	// that randomises hash orderings	572	// that randomises hash orderings
407	if (enc->json.flags & F_CANONICAL)	573	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
408	{	574	{
409	int count = hv_iterinit (hv);	575	int count = hv_iterinit (hv);
410		576
411	if (SvMAGICAL (hv))	577	if (SvMAGICAL (hv))
412	{	578	{
…		…
459		625
460	FREETMPS;	626	FREETMPS;
461	LEAVE;	627	LEAVE;
462	}	628	}
463		629
		630	encode_nl (enc); ++enc->indent;
		631
464	while (count--)	632	while (count--)
465	{	633	{
466	encode_indent (enc);	634	encode_indent (enc);
467	he = hes [count];	635	he = hes [count];
468	encode_hk (enc, he);	636	encode_hk (enc, he);
469	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));	637	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
470		638
471	if (count)	639	if (count)
472	encode_comma (enc);	640	encode_comma (enc);
473	}	641	}
		642
		643	encode_nl (enc); --enc->indent; encode_indent (enc);
474	}	644	}
475	}	645	}
476	else	646	else
477	{	647	{
478	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))	648	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))
479	if ((he = hv_iternext (hv)))	649	if ((he = hv_iternext (hv)))
		650	{
		651	encode_nl (enc); ++enc->indent;
		652
480	for (;;)	653	for (;;)
481	{	654	{
482	encode_indent (enc);	655	encode_indent (enc);
483	encode_hk (enc, he);	656	encode_hk (enc, he);
484	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));	657	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
485		658
486	if (!(he = hv_iternext (hv)))	659	if (!(he = hv_iternext (hv)))
487	break;	660	break;
488		661
489	encode_comma (enc);	662	encode_comma (enc);
490	}	663	}
491	}
492		664
		665	encode_nl (enc); --enc->indent; encode_indent (enc);
		666	}
		667	}
		668
493	encode_nl (enc);	669	encode_ch (enc, '}');
494
495	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
496	}	670	}
497		671
498	// encode objects, arrays and special \0=false and \1=true values.	672	// encode objects, arrays and special \0=false and \1=true values.
499	static void	673	static void
500	encode_rv (enc_t enc, SV sv)	674	encode_rv (enc_t enc, SV sv)
…		…
577		751
578	if (len == 1 && *pv == '1')	752	if (len == 1 && *pv == '1')
579	encode_str (enc, "true", 4, 0);	753	encode_str (enc, "true", 4, 0);
580	else if (len == 1 && *pv == '0')	754	else if (len == 1 && *pv == '0')
581	encode_str (enc, "false", 5, 0);	755	encode_str (enc, "false", 5, 0);
		756	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		757	encode_str (enc, "null", 4, 0);
582	else	758	else
583	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	759	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
584	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	760	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
585	}	761	}
		762	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		763	encode_str (enc, "null", 4, 0);
586	else	764	else
587	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	765	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
588	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	766	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
589	}	767	}
590		768
…		…
608	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	786	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
609	enc->cur += strlen (enc->cur);	787	enc->cur += strlen (enc->cur);
610	}	788	}
611	else if (SvIOKp (sv))	789	else if (SvIOKp (sv))
612	{	790	{
613	// we assume we can always read an IV as a UV	791	// we assume we can always read an IV as a UV and vice versa
614	if (SvUV (sv) & ~(UV)0x7fff)	792	// we assume two's complement
615	{	793	// we assume no aliasing issues in the union
616	// large integer, use the (rather slow) snprintf way.	794	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
617	need (enc, sizeof (UV) * 3);	795	: 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	{	796	{
625	// optimise the "small number case"	797	// optimise the "small number case"
626	// code will likely be branchless and use only a single multiplication	798	// code will likely be branchless and use only a single multiplication
		799	// works for numbers up to 59074
627	I32 i = SvIV (sv);	800	I32 i = SvIVX (sv);
628	U32 u;	801	U32 u;
629	char digit, nz = 0;	802	char digit, nz = 0;
630		803
631	need (enc, 6);	804	need (enc, 6);
632		805
…		…
638		811
639	// now output digit by digit, each time masking out the integer part	812	// 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,	813	// and multiplying by 5 while moving the decimal point one to the right,
641	// resulting in a net multiplication by 10.	814	// resulting in a net multiplication by 10.
642	// we always write the digit to memory but conditionally increment	815	// we always write the digit to memory but conditionally increment
643	// the pointer, to ease the usage of conditional move instructions.	816	// 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;	817	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;	818	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;	819	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;	820	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'	821	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
649	}	822	}
		823	else
		824	{
		825	// large integer, use the (rather slow) snprintf way.
		826	need (enc, IVUV_MAXCHARS);
		827	enc->cur +=
		828	SvIsUV(sv)
		829	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
		830	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		831	}
650	}	832	}
651	else if (SvROK (sv))	833	else if (SvROK (sv))
652	encode_rv (enc, SvRV (sv));	834	encode_rv (enc, SvRV (sv));
653	else if (!SvOK (sv))	835	else if (!SvOK (sv) \|\| enc->json.flags & F_ALLOW_UNKNOWN)
654	encode_str (enc, "null", 4, 0);	836	encode_str (enc, "null", 4, 0);
655	else	837	else
656	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	838	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
657	SvPV_nolen (sv), SvFLAGS (sv));	839	SvPV_nolen (sv), SvFLAGS (sv));
658	}	840	}
…		…
668	enc.json = *json;	850	enc.json = *json;
669	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	851	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
670	enc.cur = SvPVX (enc.sv);	852	enc.cur = SvPVX (enc.sv);
671	enc.end = SvEND (enc.sv);	853	enc.end = SvEND (enc.sv);
672	enc.indent = 0;	854	enc.indent = 0;
673	enc.maxdepth = DEC_DEPTH (enc.json.flags);	855	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		856	: enc.json.flags & F_LATIN1 ? 0x000100UL
		857	: 0x110000UL;
674		858
675	SvPOK_only (enc.sv);	859	SvPOK_only (enc.sv);
676	encode_sv (&enc, scalar);	860	encode_sv (&enc, scalar);
		861	encode_nl (&enc);
677		862
678	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	863	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
679	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	864	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
680		865
681	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))	866	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))
…		…
699	JSON json;	884	JSON json;
700	U32 depth; // recursion depth	885	U32 depth; // recursion depth
701	U32 maxdepth; // recursion depth limit	886	U32 maxdepth; // recursion depth limit
702	} dec_t;	887	} dec_t;
703		888
704	inline void	889	INLINE void
705	decode_comment (dec_t *dec)	890	decode_comment (dec_t *dec)
706	{	891	{
707	// only '#'-style comments allowed a.t.m.	892	// only '#'-style comments allowed a.t.m.
708		893
709	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)	894	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)
710	++dec->cur;	895	++dec->cur;
711	}	896	}
712		897
713	inline void	898	INLINE void
714	decode_ws (dec_t *dec)	899	decode_ws (dec_t *dec)
715	{	900	{
716	for (;;)	901	for (;;)
717	{	902	{
718	char ch = *dec->cur;	903	char ch = *dec->cur;
…		…
742	if (*dec->cur != ch) \	927	if (*dec->cur != ch) \
743	ERR (# ch " expected"); \	928	ERR (# ch " expected"); \
744	++dec->cur; \	929	++dec->cur; \
745	SE	930	SE
746		931
747	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	932	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
748	#define DEC_DEC_DEPTH --dec->depth	933	#define DEC_DEC_DEPTH --dec->depth
749		934
750	static SV decode_sv (dec_t dec);	935	static SV decode_sv (dec_t dec);
751		936
752	static signed char decode_hexdigit[256];	937	static signed char decode_hexdigit[256];
…		…
844		1029
845	if (hi >= 0x80)	1030	if (hi >= 0x80)
846	{	1031	{
847	utf8 = 1;	1032	utf8 = 1;
848		1033
849	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	1034	cur = encode_utf8 (cur, hi);
850	}	1035	}
851	else	1036	else
852	*cur++ = hi;	1037	*cur++ = hi;
853	}	1038	}
854	break;	1039	break;
…		…
856	default:	1041	default:
857	--dec_cur;	1042	--dec_cur;
858	ERR ("illegal backslash escape sequence in string");	1043	ERR ("illegal backslash escape sequence in string");
859	}	1044	}
860	}	1045	}
861	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	1046	else if (expect_true (ch >= 0x20 && ch < 0x80))
862	*cur++ = ch;	1047	*cur++ = ch;
863	else if (ch >= 0x80)	1048	else if (ch >= 0x80)
864	{	1049	{
865	STRLEN clen;	1050	STRLEN clen;
866	UV uch;
867		1051
868	--dec_cur;	1052	--dec_cur;
869		1053
870	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);	1054	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
871	if (clen == (STRLEN)-1)	1055	if (clen == (STRLEN)-1)
872	ERR ("malformed UTF-8 character in JSON string");	1056	ERR ("malformed UTF-8 character in JSON string");
873		1057
874	do	1058	do
875	cur++ = dec_cur++;	1059	cur++ = dec_cur++;
…		…
892	{	1076	{
893	STRLEN len = cur - buf;	1077	STRLEN len = cur - buf;
894		1078
895	if (sv)	1079	if (sv)
896	{	1080	{
897	SvGROW (sv, SvCUR (sv) + len + 1);	1081	STRLEN cur = SvCUR (sv);
		1082
		1083	if (SvLEN (sv) <= cur + len)
		1084	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1085
898	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	1086	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
899	SvCUR_set (sv, SvCUR (sv) + len);	1087	SvCUR_set (sv, SvCUR (sv) + len);
900	}	1088	}
901	else	1089	else
902	sv = newSVpvn (buf, len);	1090	sv = newSVpvn (buf, len);
…		…
989		1177
990	if (!is_nv)	1178	if (!is_nv)
991	{	1179	{
992	int len = dec->cur - start;	1180	int len = dec->cur - start;
993		1181
994	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1182	// special case the rather common 1..5-digit-int case
995	if (*start == '-')	1183	if (*start == '-')
996	switch (len)	1184	switch (len)
997	{	1185	{
998	case 2: return newSViv (-( start [1] - '0' * 1));	1186	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
999	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1187	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));	1188	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));	1189	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1190	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1002	}	1191	}
1003	else	1192	else
1004	switch (len)	1193	switch (len)
1005	{	1194	{
1006	case 1: return newSViv ( start [0] - '0' * 1);	1195	case 1: return newSViv ( start [0] - '0' * 1);
1007	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1196	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);	1197	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);	1198	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1199	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1010	}	1200	}
1011		1201
1012	{	1202	{
1013	UV uv;	1203	UV uv;
1014	int numtype = grok_number (start, len, &uv);	1204	int numtype = grok_number (start, len, &uv);
…		…
1023	}	1213	}
1024		1214
1025	len -= *start == '-' ? 1 : 0;	1215	len -= *start == '-' ? 1 : 0;
1026		1216
1027	// does not fit into IV or UV, try NV	1217	// does not fit into IV or UV, try NV
1028	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)	1218	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	1219	// fits into NV without loss of precision
1034	return newSVnv (Atof (start));	1220	return newSVnv (json_atof (start));
1035		1221
1036	// everything else fails, convert it to a string	1222	// everything else fails, convert it to a string
1037	return newSVpvn (start, dec->cur - start);	1223	return newSVpvn (start, dec->cur - start);
1038	}	1224	}
1039		1225
1040	// loss of precision here	1226	// loss of precision here
1041	return newSVnv (Atof (start));	1227	return newSVnv (json_atof (start));
1042		1228
1043	fail:	1229	fail:
1044	return 0;	1230	return 0;
1045	}	1231	}
1046		1232
…		…
1122	char *p = dec->cur;	1308	char *p = dec->cur;
1123	char *e = p + 24; // only try up to 24 bytes	1309	char *e = p + 24; // only try up to 24 bytes
1124		1310
1125	for (;;)	1311	for (;;)
1126	{	1312	{
1127	// the >= 0x80 is true on most architectures	1313	// the >= 0x80 is false on most architectures
1128	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')	1314	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
1129	{	1315	{
1130	// slow path, back up and use decode_str	1316	// slow path, back up and use decode_str
1131	SV *key = decode_str (dec);	1317	SV *key = decode_str (dec);
1132	if (!key)	1318	if (!key)
…		…
1215	dSP;	1401	dSP;
1216	int count;	1402	int count;
1217		1403
1218	ENTER; SAVETMPS; PUSHMARK (SP);	1404	ENTER; SAVETMPS; PUSHMARK (SP);
1219	XPUSHs (HeVAL (he));	1405	XPUSHs (HeVAL (he));
		1406	sv_2mortal (sv);
1220		1407
1221	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;	1408	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1222		1409
1223	if (count == 1)	1410	if (count == 1)
1224	{	1411	{
1225	sv = newSVsv (POPs);	1412	sv = newSVsv (POPs);
1226	FREETMPS; LEAVE;	1413	FREETMPS; LEAVE;
1227	return sv;	1414	return sv;
1228	}	1415	}
1229		1416
		1417	SvREFCNT_inc (sv);
1230	FREETMPS; LEAVE;	1418	FREETMPS; LEAVE;
1231	}	1419	}
1232	}	1420	}
1233		1421
1234	if (dec->json.cb_object)	1422	if (dec->json.cb_object)
…		…
1263		1451
1264	static SV *	1452	static SV *
1265	decode_sv (dec_t *dec)	1453	decode_sv (dec_t *dec)
1266	{	1454	{
1267	// the beauty of JSON: you need exactly one character lookahead	1455	// the beauty of JSON: you need exactly one character lookahead
1268	// to parse anything.	1456	// to parse everything.
1269	switch (*dec->cur)	1457	switch (*dec->cur)
1270	{	1458	{
1271	case '"': ++dec->cur; return decode_str (dec);	1459	case '"': ++dec->cur; return decode_str (dec);
1272	case '[': ++dec->cur; return decode_av (dec);	1460	case '[': ++dec->cur; return decode_av (dec);
1273	case '{': ++dec->cur; return decode_hv (dec);	1461	case '{': ++dec->cur; return decode_hv (dec);
1274		1462
1275	case '-':	1463	case '-':
1276	case '0': case '1': case '2': case '3': case '4':	1464	case '0': case '1': case '2': case '3': case '4':
1277	case '5': case '6': case '7': case '8': case '9':	1465	case '5': case '6': case '7': case '8': case '9':
1278	return decode_num (dec);	1466	return decode_num (dec);
…		…
1280	case 't':	1468	case 't':
1281	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1469	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1282	{	1470	{
1283	dec->cur += 4;	1471	dec->cur += 4;
1284	#if JSON_SLOW	1472	#if JSON_SLOW
1285	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);	1473	json_true = get_bool ("JSON::XS::true");
1286	#endif	1474	#endif
1287	return SvREFCNT_inc (json_true);	1475	return newSVsv (json_true);
1288	}	1476	}
1289	else	1477	else
1290	ERR ("'true' expected");	1478	ERR ("'true' expected");
1291		1479
1292	break;	1480	break;
…		…
1294	case 'f':	1482	case 'f':
1295	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1483	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1296	{	1484	{
1297	dec->cur += 5;	1485	dec->cur += 5;
1298	#if JSON_SLOW	1486	#if JSON_SLOW
1299	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1487	json_false = get_bool ("JSON::XS::false");
1300	#endif	1488	#endif
1301	return SvREFCNT_inc (json_false);	1489	return newSVsv (json_false);
1302	}	1490	}
1303	else	1491	else
1304	ERR ("'false' expected");	1492	ERR ("'false' expected");
1305		1493
1306	break;	1494	break;
…		…
1324	fail:	1512	fail:
1325	return 0;	1513	return 0;
1326	}	1514	}
1327		1515
1328	static SV *	1516	static SV *
1329	decode_json (SV string, JSON json, UV *offset_return)	1517	decode_json (SV string, JSON json, char **offset_return)
1330	{	1518	{
1331	dec_t dec;	1519	dec_t dec;
1332	UV offset;
1333	SV *sv;	1520	SV *sv;
1334		1521
		1522	/* work around bugs in 5.10 where manipulating magic values
		1523	* will perl ignore the magic in subsequent accesses
		1524	*/
1335	SvGETMAGIC (string);	1525	/SvGETMAGIC (string);/
		1526	if (SvMAGICAL (string))
		1527	string = sv_2mortal (newSVsv (string));
		1528
1336	SvUPGRADE (string, SVt_PV);	1529	SvUPGRADE (string, SVt_PV);
1337		1530
1338	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))	1531	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1532	* assertion with -DDEBUGGING, although SvCUR is documented to
		1533	* return the xpv_cur field which certainly exists after upgrading.
		1534	* according to nicholas clark, calling SvPOK fixes this.
		1535	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1536	* and hope for the best.
		1537	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1538	* assertion business is seriously broken, try yet another workaround
		1539	* for the broken -DDEBUGGING.
		1540	*/
		1541	{
		1542	#ifdef DEBUGGING
		1543	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1544	#else
		1545	STRLEN offset = SvCUR (string);
		1546	#endif
		1547
		1548	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",	1549	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));	1550	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1551	}
1341		1552
1342	if (json->flags & F_UTF8)	1553	if (json->flags & F_UTF8)
1343	sv_utf8_downgrade (string, 0);	1554	sv_utf8_downgrade (string, 0);
1344	else	1555	else
1345	sv_utf8_upgrade (string);	1556	sv_utf8_upgrade (string);
1346		1557
1347	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1558	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1348		1559
1349	dec.json = *json;	1560	dec.json = *json;
1350	dec.cur = SvPVX (string);	1561	dec.cur = SvPVX (string);
1351	dec.end = SvEND (string);	1562	dec.end = SvEND (string);
1352	dec.err = 0;	1563	dec.err = 0;
1353	dec.depth = 0;	1564	dec.depth = 0;
1354	dec.maxdepth = DEC_DEPTH (dec.json.flags);
1355		1565
1356	if (dec.json.cb_object \|\| dec.json.cb_sk_object)	1566	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
1357	dec.json.flags \|= F_HOOK;	1567	dec.json.flags \|= F_HOOK;
1358		1568
1359	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1569	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
1360		1570
1361	decode_ws (&dec);	1571	decode_ws (&dec);
1362	sv = decode_sv (&dec);	1572	sv = decode_sv (&dec);
		1573
		1574	if (offset_return)
		1575	*offset_return = dec.cur;
1363		1576
1364	if (!(offset_return \|\| !sv))	1577	if (!(offset_return \|\| !sv))
1365	{	1578	{
1366	// check for trailing garbage	1579	// check for trailing garbage
1367	decode_ws (&dec);	1580	decode_ws (&dec);
…		…
1370	{	1583	{
1371	dec.err = "garbage after JSON object";	1584	dec.err = "garbage after JSON object";
1372	SvREFCNT_dec (sv);	1585	SvREFCNT_dec (sv);
1373	sv = 0;	1586	sv = 0;
1374	}	1587	}
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	}	1588	}
1386		1589
1387	if (!sv)	1590	if (!sv)
1388	{	1591	{
1389	SV *uni = sv_newmortal ();	1592	SV *uni = sv_newmortal ();
…		…
1395	SAVEVPTR (PL_curcop);	1598	SAVEVPTR (PL_curcop);
1396	PL_curcop = &cop;	1599	PL_curcop = &cop;
1397	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1600	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1398	LEAVE;	1601	LEAVE;
1399		1602
1400	croak ("%s, at character offset %d [\"%s\"]",	1603	croak ("%s, at character offset %d (before \"%s\")",
1401	dec.err,	1604	dec.err,
1402	(int)offset,	1605	ptr_to_index (string, dec.cur),
1403	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1606	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1404	}	1607	}
1405		1608
1406	sv = sv_2mortal (sv);	1609	sv = sv_2mortal (sv);
1407		1610
1408	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))	1611	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)");	1612	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1410		1613
1411	return sv;	1614	return sv;
		1615	}
		1616
		1617	/////////////////////////////////////////////////////////////////////////////
		1618	// incremental parser
		1619
		1620	static void
		1621	incr_parse (JSON *self)
		1622	{
		1623	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1624
		1625	// the state machine here is a bit convoluted and could be simplified a lot
		1626	// but this would make it slower, so...
		1627
		1628	for (;;)
		1629	{
		1630	//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
		1631	switch (self->incr_mode)
		1632	{
		1633	// only used for initial whitespace skipping
		1634	case INCR_M_WS:
		1635	for (;;)
		1636	{
		1637	if (*p > 0x20)
		1638	{
		1639	if (*p == '#')
		1640	{
		1641	self->incr_mode = INCR_M_C0;
		1642	goto incr_m_c;
		1643	}
		1644	else
		1645	{
		1646	self->incr_mode = INCR_M_JSON;
		1647	goto incr_m_json;
		1648	}
		1649	}
		1650	else if (!*p)
		1651	goto interrupt;
		1652
		1653	++p;
		1654	}
		1655
		1656	// skip a single char inside a string (for \\-processing)
		1657	case INCR_M_BS:
		1658	if (!*p)
		1659	goto interrupt;
		1660
		1661	++p;
		1662	self->incr_mode = INCR_M_STR;
		1663	goto incr_m_str;
		1664
		1665	// inside #-style comments
		1666	case INCR_M_C0:
		1667	case INCR_M_C1:
		1668	incr_m_c:
		1669	for (;;)
		1670	{
		1671	if (*p == '\n')
		1672	{
		1673	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1674	break;
		1675	}
		1676	else if (!*p)
		1677	goto interrupt;
		1678
		1679	++p;
		1680	}
		1681
		1682	break;
		1683
		1684	// inside a string
		1685	case INCR_M_STR:
		1686	incr_m_str:
		1687	for (;;)
		1688	{
		1689	if (*p == '"')
		1690	{
		1691	++p;
		1692	self->incr_mode = INCR_M_JSON;
		1693
		1694	if (!self->incr_nest)
		1695	goto interrupt;
		1696
		1697	goto incr_m_json;
		1698	}
		1699	else if (*p == '\\')
		1700	{
		1701	++p; // "virtually" consumes character after \
		1702
		1703	if (!*p) // if at end of string we have to switch modes
		1704	{
		1705	self->incr_mode = INCR_M_BS;
		1706	goto interrupt;
		1707	}
		1708	}
		1709	else if (!*p)
		1710	goto interrupt;
		1711
		1712	++p;
		1713	}
		1714
		1715	// after initial ws, outside string
		1716	case INCR_M_JSON:
		1717	incr_m_json:
		1718	for (;;)
		1719	{
		1720	switch (*p++)
		1721	{
		1722	case 0:
		1723	--p;
		1724	goto interrupt;
		1725
		1726	case 0x09:
		1727	case 0x0a:
		1728	case 0x0d:
		1729	case 0x20:
		1730	if (!self->incr_nest)
		1731	{
		1732	--p; // do not eat the whitespace, let the next round do it
		1733	goto interrupt;
		1734	}
		1735	break;
		1736
		1737	case '"':
		1738	self->incr_mode = INCR_M_STR;
		1739	goto incr_m_str;
		1740
		1741	case '[':
		1742	case '{':
		1743	if (++self->incr_nest > self->max_depth)
		1744	croak (ERR_NESTING_EXCEEDED);
		1745	break;
		1746
		1747	case ']':
		1748	case '}':
		1749	if (--self->incr_nest <= 0)
		1750	goto interrupt;
		1751	break;
		1752
		1753	case '#':
		1754	self->incr_mode = INCR_M_C1;
		1755	goto incr_m_c;
		1756	}
		1757	}
		1758	}
		1759
		1760	modechange:
		1761	;
		1762	}
		1763
		1764	interrupt:
		1765	self->incr_pos = p - SvPVX (self->incr_text);
		1766	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1767	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1412	}	1768	}
1413		1769
1414	/////////////////////////////////////////////////////////////////////////////	1770	/////////////////////////////////////////////////////////////////////////////
1415	// XS interface functions	1771	// XS interface functions
1416		1772
…		…
1428	: -1;	1784	: -1;
1429		1785
1430	json_stash = gv_stashpv ("JSON::XS" , 1);	1786	json_stash = gv_stashpv ("JSON::XS" , 1);
1431	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);	1787	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1432		1788
1433	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );	1789	json_true = get_bool ("JSON::XS::true");
1434	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1790	json_false = get_bool ("JSON::XS::false");
		1791
		1792	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1435	}	1793	}
1436		1794
1437	PROTOTYPES: DISABLE	1795	PROTOTYPES: DISABLE
1438		1796
1439	void CLONE (...)	1797	void CLONE (...)
…		…
1444	void new (char *klass)	1802	void new (char *klass)
1445	PPCODE:	1803	PPCODE:
1446	{	1804	{
1447	SV *pv = NEWSV (0, sizeof (JSON));	1805	SV *pv = NEWSV (0, sizeof (JSON));
1448	SvPOK_only (pv);	1806	SvPOK_only (pv);
1449	Zero (SvPVX (pv), 1, JSON);	1807	json_init ((JSON *)SvPVX (pv));
1450	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	1808	XPUSHs (sv_2mortal (sv_bless (
1451	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH)));	1809	newRV_noinc (pv),
		1810	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1811	)));
1452	}	1812	}
1453		1813
1454	void ascii (JSON *self, int enable = 1)	1814	void ascii (JSON *self, int enable = 1)
1455	ALIAS:	1815	ALIAS:
1456	ascii = F_ASCII	1816	ascii = F_ASCII
…		…
1464	allow_nonref = F_ALLOW_NONREF	1824	allow_nonref = F_ALLOW_NONREF
1465	shrink = F_SHRINK	1825	shrink = F_SHRINK
1466	allow_blessed = F_ALLOW_BLESSED	1826	allow_blessed = F_ALLOW_BLESSED
1467	convert_blessed = F_CONV_BLESSED	1827	convert_blessed = F_CONV_BLESSED
1468	relaxed = F_RELAXED	1828	relaxed = F_RELAXED
		1829	allow_unknown = F_ALLOW_UNKNOWN
1469	PPCODE:	1830	PPCODE:
1470	{	1831	{
1471	if (enable)	1832	if (enable)
1472	self->flags \|= ix;	1833	self->flags \|= ix;
1473	else	1834	else
1474	self->flags &= ~ix;	1835	self->flags &= ~ix;
1475		1836
1476	XPUSHs (ST (0));	1837	XPUSHs (ST (0));
1477	}	1838	}
1478		1839
1479	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1840	void get_ascii (JSON *self)
		1841	ALIAS:
		1842	get_ascii = F_ASCII
		1843	get_latin1 = F_LATIN1
		1844	get_utf8 = F_UTF8
		1845	get_indent = F_INDENT
		1846	get_canonical = F_CANONICAL
		1847	get_space_before = F_SPACE_BEFORE
		1848	get_space_after = F_SPACE_AFTER
		1849	get_allow_nonref = F_ALLOW_NONREF
		1850	get_shrink = F_SHRINK
		1851	get_allow_blessed = F_ALLOW_BLESSED
		1852	get_convert_blessed = F_CONV_BLESSED
		1853	get_relaxed = F_RELAXED
		1854	get_allow_unknown = F_ALLOW_UNKNOWN
1480	PPCODE:	1855	PPCODE:
1481	{	1856	XPUSHs (boolSV (self->flags & ix));
1482	UV log2 = 0;
1483		1857
1484	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	1858	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1485		1859	PPCODE:
1486	while ((1UL << log2) < max_depth)	1860	self->max_depth = max_depth;
1487	++log2;
1488
1489	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1490
1491	XPUSHs (ST (0));	1861	XPUSHs (ST (0));
1492	}
1493		1862
		1863	U32 get_max_depth (JSON *self)
		1864	CODE:
		1865	RETVAL = self->max_depth;
		1866	OUTPUT:
		1867	RETVAL
		1868
1494	void max_size (JSON *self, UV max_size = 0)	1869	void max_size (JSON *self, U32 max_size = 0)
1495	PPCODE:	1870	PPCODE:
1496	{	1871	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));	1872	XPUSHs (ST (0));
1508	}	1873
		1874	int get_max_size (JSON *self)
		1875	CODE:
		1876	RETVAL = self->max_size;
		1877	OUTPUT:
		1878	RETVAL
1509		1879
1510	void filter_json_object (JSON self, SV cb = &PL_sv_undef)	1880	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1511	PPCODE:	1881	PPCODE:
1512	{	1882	{
1513	SvREFCNT_dec (self->cb_object);	1883	SvREFCNT_dec (self->cb_object);
…		…
1547	XPUSHs (decode_json (jsonstr, self, 0));	1917	XPUSHs (decode_json (jsonstr, self, 0));
1548		1918
1549	void decode_prefix (JSON self, SV jsonstr)	1919	void decode_prefix (JSON self, SV jsonstr)
1550	PPCODE:	1920	PPCODE:
1551	{	1921	{
1552	UV offset;	1922	char *offset;
1553	EXTEND (SP, 2);	1923	EXTEND (SP, 2);
1554	PUSHs (decode_json (jsonstr, self, &offset));	1924	PUSHs (decode_json (jsonstr, self, &offset));
1555	PUSHs (sv_2mortal (newSVuv (offset)));	1925	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		1926	}
		1927
		1928	void incr_parse (JSON self, SV jsonstr = 0)
		1929	PPCODE:
		1930	{
		1931	if (!self->incr_text)
		1932	self->incr_text = newSVpvn ("", 0);
		1933
		1934	// append data, if any
		1935	if (jsonstr)
		1936	{
		1937	if (SvUTF8 (jsonstr))
		1938	{
		1939	if (!SvUTF8 (self->incr_text))
		1940	{
		1941	/* utf-8-ness differs, need to upgrade */
		1942	sv_utf8_upgrade (self->incr_text);
		1943
		1944	if (self->incr_pos)
		1945	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1946	- (U8 *)SvPVX (self->incr_text);
		1947	}
		1948	}
		1949	else if (SvUTF8 (self->incr_text))
		1950	sv_utf8_upgrade (jsonstr);
		1951
		1952	{
		1953	STRLEN len;
		1954	const char *str = SvPV (jsonstr, len);
		1955	STRLEN cur = SvCUR (self->incr_text);
		1956
		1957	if (SvLEN (self->incr_text) <= cur + len)
		1958	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1959
		1960	Move (str, SvEND (self->incr_text), len, char);
		1961	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1962	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1963	}
		1964	}
		1965
		1966	if (GIMME_V != G_VOID)
		1967	do
		1968	{
		1969	char *offset;
		1970
		1971	if (!INCR_DONE (self))
		1972	{
		1973	incr_parse (self);
		1974
		1975	if (self->incr_pos > self->max_size && self->max_size)
		1976	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1977	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1978
		1979	if (!INCR_DONE (self))
		1980	break;
		1981	}
		1982
		1983	XPUSHs (decode_json (self->incr_text, self, &offset));
		1984
		1985	self->incr_pos -= offset - SvPVX (self->incr_text);
		1986	self->incr_nest = 0;
		1987	self->incr_mode = 0;
		1988
		1989	sv_chop (self->incr_text, offset);
		1990	}
		1991	while (GIMME_V == G_ARRAY);
		1992	}
		1993
		1994	SV incr_text (JSON self)
		1995	ATTRS: lvalue
		1996	CODE:
		1997	{
		1998	if (self->incr_pos)
		1999	croak ("incr_text can not be called when the incremental parser already started parsing");
		2000
		2001	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		2002	}
		2003	OUTPUT:
		2004	RETVAL
		2005
		2006	void incr_skip (JSON *self)
		2007	CODE:
		2008	{
		2009	if (self->incr_pos)
		2010	{
		2011	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2012	self->incr_pos = 0;
		2013	self->incr_nest = 0;
		2014	self->incr_mode = 0;
		2015	}
		2016	}
		2017
		2018	void incr_reset (JSON *self)
		2019	CODE:
		2020	{
		2021	SvREFCNT_dec (self->incr_text);
		2022	self->incr_text = 0;
		2023	self->incr_pos = 0;
		2024	self->incr_nest = 0;
		2025	self->incr_mode = 0;
1556	}	2026	}
1557		2027
1558	void DESTROY (JSON *self)	2028	void DESTROY (JSON *self)
1559	CODE:	2029	CODE:
1560	SvREFCNT_dec (self->cb_sk_object);	2030	SvREFCNT_dec (self->cb_sk_object);
1561	SvREFCNT_dec (self->cb_object);	2031	SvREFCNT_dec (self->cb_object);
		2032	SvREFCNT_dec (self->incr_text);
1562		2033
1563	PROTOTYPES: ENABLE	2034	PROTOTYPES: ENABLE
1564		2035
1565	void to_json (SV *scalar)	2036	void encode_json (SV *scalar)
		2037	ALIAS:
		2038	to_json_ = 0
		2039	encode_json = F_UTF8
1566	PPCODE:	2040	PPCODE:
1567	{	2041	{
1568	JSON json = { F_DEFAULT \| F_UTF8 };	2042	JSON json;
		2043	json_init (&json);
		2044	json.flags \|= ix;
1569	XPUSHs (encode_json (scalar, &json));	2045	XPUSHs (encode_json (scalar, &json));
1570	}	2046	}
1571		2047
1572	void from_json (SV *jsonstr)	2048	void decode_json (SV *jsonstr)
		2049	ALIAS:
		2050	from_json_ = 0
		2051	decode_json = F_UTF8
1573	PPCODE:	2052	PPCODE:
1574	{	2053	{
1575	JSON json = { F_DEFAULT \| F_UTF8 };	2054	JSON json;
		2055	json_init (&json);
		2056	json.flags \|= ix;
1576	XPUSHs (decode_json (jsonstr, &json, 0));	2057	XPUSHs (decode_json (jsonstr, &json, 0));
1577	}	2058	}
1578		2059

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 1.107 by root, Wed Mar 17 01:45:43 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 1.107 by root, Wed Mar 17 01:45:43 2010 UTC