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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.67 by root, Wed Nov 28 14:01:01 2007 UTC vs.
Revision 1.106 by root, Tue Jan 19 01:36:34 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, '[');	481	encode_ch (enc, '[');
318		482
319	if (len >= 0)	483	if (len >= 0)
320	{	484	{
…		…
394		558
395	static void	559	static void
396	encode_hv (enc_t enc, HV hv)	560	encode_hv (enc_t enc, HV hv)
397	{	561	{
398	HE *he;	562	HE *he;
399	int count;
400		563
401	if (enc->indent >= enc->maxdepth)	564	if (enc->indent >= enc->json.max_depth)
402	croak ("data structure too deep (hit recursion limit)");	565	croak (ERR_NESTING_EXCEEDED);
403		566
404	encode_ch (enc, '{');	567	encode_ch (enc, '{');
405		568
406	// for canonical output we have to sort by keys first	569	// for canonical output we have to sort by keys first
407	// actually, this is mostly due to the stupid so-called	570	// actually, this is mostly due to the stupid so-called
408	// security workaround added somewhere in 5.8.x.	571	// security workaround added somewhere in 5.8.x
409	// that randomises hash orderings	572	// that randomises hash orderings
410	if (enc->json.flags & F_CANONICAL)	573	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
411	{	574	{
412	int count = hv_iterinit (hv);	575	int count = hv_iterinit (hv);
413		576
414	if (SvMAGICAL (hv))	577	if (SvMAGICAL (hv))
415	{	578	{
…		…
588		751
589	if (len == 1 && *pv == '1')	752	if (len == 1 && *pv == '1')
590	encode_str (enc, "true", 4, 0);	753	encode_str (enc, "true", 4, 0);
591	else if (len == 1 && *pv == '0')	754	else if (len == 1 && *pv == '0')
592	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);
593	else	758	else
594	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",
595	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	760	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
596	}	761	}
		762	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		763	encode_str (enc, "null", 4, 0);
597	else	764	else
598	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",
599	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	766	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
600	}	767	}
601		768
…		…
619	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	786	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
620	enc->cur += strlen (enc->cur);	787	enc->cur += strlen (enc->cur);
621	}	788	}
622	else if (SvIOKp (sv))	789	else if (SvIOKp (sv))
623	{	790	{
624	// 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
625	if (SvUV (sv) & ~(UV)0x7fff)	792	// we assume two's complement
626	{	793	// we assume no aliasing issues in the union
627	// large integer, use the (rather slow) snprintf way.	794	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
628	need (enc, sizeof (UV) * 3);	795	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
629	enc->cur +=
630	SvIsUV(sv)
631	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
632	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
633	}
634	else
635	{	796	{
636	// optimise the "small number case"	797	// optimise the "small number case"
637	// 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
638	I32 i = SvIV (sv);	800	I32 i = SvIVX (sv);
639	U32 u;	801	U32 u;
640	char digit, nz = 0;	802	char digit, nz = 0;
641		803
642	need (enc, 6);	804	need (enc, 6);
643		805
…		…
649		811
650	// 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
651	// 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,
652	// resulting in a net multiplication by 10.	814	// resulting in a net multiplication by 10.
653	// we always write the digit to memory but conditionally increment	815	// we always write the digit to memory but conditionally increment
654	// the pointer, to ease the usage of conditional move instructions.	816	// the pointer, to enable the use of conditional move instructions.
655	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;
656	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;
657	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;
658	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;
659	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'
660	}	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	}
661	}	832	}
662	else if (SvROK (sv))	833	else if (SvROK (sv))
663	encode_rv (enc, SvRV (sv));	834	encode_rv (enc, SvRV (sv));
664	else if (!SvOK (sv))	835	else if (!SvOK (sv) \|\| enc->json.flags & F_ALLOW_UNKNOWN)
665	encode_str (enc, "null", 4, 0);	836	encode_str (enc, "null", 4, 0);
666	else	837	else
667	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",
668	SvPV_nolen (sv), SvFLAGS (sv));	839	SvPV_nolen (sv), SvFLAGS (sv));
669	}	840	}
…		…
679	enc.json = *json;	850	enc.json = *json;
680	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	851	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
681	enc.cur = SvPVX (enc.sv);	852	enc.cur = SvPVX (enc.sv);
682	enc.end = SvEND (enc.sv);	853	enc.end = SvEND (enc.sv);
683	enc.indent = 0;	854	enc.indent = 0;
684	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;
685		858
686	SvPOK_only (enc.sv);	859	SvPOK_only (enc.sv);
687	encode_sv (&enc, scalar);	860	encode_sv (&enc, scalar);
		861	encode_nl (&enc);
688		862
689	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	863	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
690	*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
691		865
692	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))	866	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))
…		…
710	JSON json;	884	JSON json;
711	U32 depth; // recursion depth	885	U32 depth; // recursion depth
712	U32 maxdepth; // recursion depth limit	886	U32 maxdepth; // recursion depth limit
713	} dec_t;	887	} dec_t;
714		888
715	inline void	889	INLINE void
716	decode_comment (dec_t *dec)	890	decode_comment (dec_t *dec)
717	{	891	{
718	// only '#'-style comments allowed a.t.m.	892	// only '#'-style comments allowed a.t.m.
719		893
720	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)	894	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)
721	++dec->cur;	895	++dec->cur;
722	}	896	}
723		897
724	inline void	898	INLINE void
725	decode_ws (dec_t *dec)	899	decode_ws (dec_t *dec)
726	{	900	{
727	for (;;)	901	for (;;)
728	{	902	{
729	char ch = *dec->cur;	903	char ch = *dec->cur;
…		…
753	if (*dec->cur != ch) \	927	if (*dec->cur != ch) \
754	ERR (# ch " expected"); \	928	ERR (# ch " expected"); \
755	++dec->cur; \	929	++dec->cur; \
756	SE	930	SE
757		931
758	#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)
759	#define DEC_DEC_DEPTH --dec->depth	933	#define DEC_DEC_DEPTH --dec->depth
760		934
761	static SV decode_sv (dec_t dec);	935	static SV decode_sv (dec_t dec);
762		936
763	static signed char decode_hexdigit[256];	937	static signed char decode_hexdigit[256];
…		…
855		1029
856	if (hi >= 0x80)	1030	if (hi >= 0x80)
857	{	1031	{
858	utf8 = 1;	1032	utf8 = 1;
859		1033
860	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	1034	cur = encode_utf8 (cur, hi);
861	}	1035	}
862	else	1036	else
863	*cur++ = hi;	1037	*cur++ = hi;
864	}	1038	}
865	break;	1039	break;
…		…
867	default:	1041	default:
868	--dec_cur;	1042	--dec_cur;
869	ERR ("illegal backslash escape sequence in string");	1043	ERR ("illegal backslash escape sequence in string");
870	}	1044	}
871	}	1045	}
872	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	1046	else if (expect_true (ch >= 0x20 && ch < 0x80))
873	*cur++ = ch;	1047	*cur++ = ch;
874	else if (ch >= 0x80)	1048	else if (ch >= 0x80)
875	{	1049	{
876	STRLEN clen;	1050	STRLEN clen;
877	UV uch;
878		1051
879	--dec_cur;	1052	--dec_cur;
880		1053
881	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);	1054	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
882	if (clen == (STRLEN)-1)	1055	if (clen == (STRLEN)-1)
883	ERR ("malformed UTF-8 character in JSON string");	1056	ERR ("malformed UTF-8 character in JSON string");
884		1057
885	do	1058	do
886	cur++ = dec_cur++;	1059	cur++ = dec_cur++;
…		…
903	{	1076	{
904	STRLEN len = cur - buf;	1077	STRLEN len = cur - buf;
905		1078
906	if (sv)	1079	if (sv)
907	{	1080	{
908	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
909	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	1086	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
910	SvCUR_set (sv, SvCUR (sv) + len);	1087	SvCUR_set (sv, SvCUR (sv) + len);
911	}	1088	}
912	else	1089	else
913	sv = newSVpvn (buf, len);	1090	sv = newSVpvn (buf, len);
…		…
1000		1177
1001	if (!is_nv)	1178	if (!is_nv)
1002	{	1179	{
1003	int len = dec->cur - start;	1180	int len = dec->cur - start;
1004		1181
1005	// 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
1006	if (*start == '-')	1183	if (*start == '-')
1007	switch (len)	1184	switch (len)
1008	{	1185	{
1009	case 2: return newSViv (-( start [1] - '0' * 1));	1186	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1010	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1187	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1011	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));
1012	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));
1013	}	1191	}
1014	else	1192	else
1015	switch (len)	1193	switch (len)
1016	{	1194	{
1017	case 1: return newSViv ( start [0] - '0' * 1);	1195	case 1: return newSViv ( start [0] - '0' * 1);
1018	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1196	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1019	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);
1020	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);
1021	}	1200	}
1022		1201
1023	{	1202	{
1024	UV uv;	1203	UV uv;
1025	int numtype = grok_number (start, len, &uv);	1204	int numtype = grok_number (start, len, &uv);
…		…
1034	}	1213	}
1035		1214
1036	len -= *start == '-' ? 1 : 0;	1215	len -= *start == '-' ? 1 : 0;
1037		1216
1038	// does not fit into IV or UV, try NV	1217	// does not fit into IV or UV, try NV
1039	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)	1218	if (len <= NV_DIG)
1040	#if defined (LDBL_DIG)
1041	\|\| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1042	#endif
1043	)
1044	// fits into NV without loss of precision	1219	// fits into NV without loss of precision
1045	return newSVnv (Atof (start));	1220	return newSVnv (json_atof (start));
1046		1221
1047	// everything else fails, convert it to a string	1222	// everything else fails, convert it to a string
1048	return newSVpvn (start, dec->cur - start);	1223	return newSVpvn (start, dec->cur - start);
1049	}	1224	}
1050		1225
1051	// loss of precision here	1226	// loss of precision here
1052	return newSVnv (Atof (start));	1227	return newSVnv (json_atof (start));
1053		1228
1054	fail:	1229	fail:
1055	return 0;	1230	return 0;
1056	}	1231	}
1057		1232
…		…
1133	char *p = dec->cur;	1308	char *p = dec->cur;
1134	char *e = p + 24; // only try up to 24 bytes	1309	char *e = p + 24; // only try up to 24 bytes
1135		1310
1136	for (;;)	1311	for (;;)
1137	{	1312	{
1138	// the >= 0x80 is true on most architectures	1313	// the >= 0x80 is false on most architectures
1139	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')	1314	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
1140	{	1315	{
1141	// slow path, back up and use decode_str	1316	// slow path, back up and use decode_str
1142	SV *key = decode_str (dec);	1317	SV *key = decode_str (dec);
1143	if (!key)	1318	if (!key)
…		…
1274		1449
1275	static SV *	1450	static SV *
1276	decode_sv (dec_t *dec)	1451	decode_sv (dec_t *dec)
1277	{	1452	{
1278	// the beauty of JSON: you need exactly one character lookahead	1453	// the beauty of JSON: you need exactly one character lookahead
1279	// to parse anything.	1454	// to parse everything.
1280	switch (*dec->cur)	1455	switch (*dec->cur)
1281	{	1456	{
1282	case '"': ++dec->cur; return decode_str (dec);	1457	case '"': ++dec->cur; return decode_str (dec);
1283	case '[': ++dec->cur; return decode_av (dec);	1458	case '[': ++dec->cur; return decode_av (dec);
1284	case '{': ++dec->cur; return decode_hv (dec);	1459	case '{': ++dec->cur; return decode_hv (dec);
1285		1460
1286	case '-':	1461	case '-':
1287	case '0': case '1': case '2': case '3': case '4':	1462	case '0': case '1': case '2': case '3': case '4':
1288	case '5': case '6': case '7': case '8': case '9':	1463	case '5': case '6': case '7': case '8': case '9':
1289	return decode_num (dec);	1464	return decode_num (dec);
…		…
1291	case 't':	1466	case 't':
1292	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1467	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1293	{	1468	{
1294	dec->cur += 4;	1469	dec->cur += 4;
1295	#if JSON_SLOW	1470	#if JSON_SLOW
1296	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);	1471	json_true = get_bool ("JSON::XS::true");
1297	#endif	1472	#endif
1298	return SvREFCNT_inc (json_true);	1473	return newSVsv (json_true);
1299	}	1474	}
1300	else	1475	else
1301	ERR ("'true' expected");	1476	ERR ("'true' expected");
1302		1477
1303	break;	1478	break;
…		…
1305	case 'f':	1480	case 'f':
1306	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1481	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1307	{	1482	{
1308	dec->cur += 5;	1483	dec->cur += 5;
1309	#if JSON_SLOW	1484	#if JSON_SLOW
1310	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1485	json_false = get_bool ("JSON::XS::false");
1311	#endif	1486	#endif
1312	return SvREFCNT_inc (json_false);	1487	return newSVsv (json_false);
1313	}	1488	}
1314	else	1489	else
1315	ERR ("'false' expected");	1490	ERR ("'false' expected");
1316		1491
1317	break;	1492	break;
…		…
1335	fail:	1510	fail:
1336	return 0;	1511	return 0;
1337	}	1512	}
1338		1513
1339	static SV *	1514	static SV *
1340	decode_json (SV string, JSON json, UV *offset_return)	1515	decode_json (SV string, JSON json, char **offset_return)
1341	{	1516	{
1342	dec_t dec;	1517	dec_t dec;
1343	UV offset;
1344	SV *sv;	1518	SV *sv;
1345		1519
		1520	/* work around bugs in 5.10 where manipulating magic values
		1521	* will perl ignore the magic in subsequent accesses
		1522	*/
1346	SvGETMAGIC (string);	1523	/SvGETMAGIC (string);/
		1524	if (SvMAGICAL (string))
		1525	string = sv_2mortal (newSVsv (string));
		1526
1347	SvUPGRADE (string, SVt_PV);	1527	SvUPGRADE (string, SVt_PV);
1348		1528
1349	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))	1529	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1530	* assertion with -DDEBUGGING, although SvCUR is documented to
		1531	* return the xpv_cur field which certainly exists after upgrading.
		1532	* according to nicholas clark, calling SvPOK fixes this.
		1533	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1534	* and hope for the best.
		1535	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1536	* assertion business is seriously broken, try yet another workaround
		1537	* for the broken -DDEBUGGING.
		1538	*/
		1539	{
		1540	#ifdef DEBUGGING
		1541	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1542	#else
		1543	STRLEN offset = SvCUR (string);
		1544	#endif
		1545
		1546	if (offset > json->max_size && json->max_size)
1350	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",	1547	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1351	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));	1548	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1549	}
1352		1550
1353	if (json->flags & F_UTF8)	1551	if (json->flags & F_UTF8)
1354	sv_utf8_downgrade (string, 0);	1552	sv_utf8_downgrade (string, 0);
1355	else	1553	else
1356	sv_utf8_upgrade (string);	1554	sv_utf8_upgrade (string);
1357		1555
1358	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1556	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1359		1557
1360	dec.json = *json;	1558	dec.json = *json;
1361	dec.cur = SvPVX (string);	1559	dec.cur = SvPVX (string);
1362	dec.end = SvEND (string);	1560	dec.end = SvEND (string);
1363	dec.err = 0;	1561	dec.err = 0;
1364	dec.depth = 0;	1562	dec.depth = 0;
1365	dec.maxdepth = DEC_DEPTH (dec.json.flags);
1366		1563
1367	if (dec.json.cb_object \|\| dec.json.cb_sk_object)	1564	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
1368	dec.json.flags \|= F_HOOK;	1565	dec.json.flags \|= F_HOOK;
1369		1566
1370	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1567	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
1371		1568
1372	decode_ws (&dec);	1569	decode_ws (&dec);
1373	sv = decode_sv (&dec);	1570	sv = decode_sv (&dec);
		1571
		1572	if (offset_return)
		1573	*offset_return = dec.cur;
1374		1574
1375	if (!(offset_return \|\| !sv))	1575	if (!(offset_return \|\| !sv))
1376	{	1576	{
1377	// check for trailing garbage	1577	// check for trailing garbage
1378	decode_ws (&dec);	1578	decode_ws (&dec);
…		…
1381	{	1581	{
1382	dec.err = "garbage after JSON object";	1582	dec.err = "garbage after JSON object";
1383	SvREFCNT_dec (sv);	1583	SvREFCNT_dec (sv);
1384	sv = 0;	1584	sv = 0;
1385	}	1585	}
1386	}
1387
1388	if (offset_return \|\| !sv)
1389	{
1390	offset = dec.json.flags & F_UTF8
1391	? dec.cur - SvPVX (string)
1392	: utf8_distance (dec.cur, SvPVX (string));
1393
1394	if (offset_return)
1395	*offset_return = offset;
1396	}	1586	}
1397		1587
1398	if (!sv)	1588	if (!sv)
1399	{	1589	{
1400	SV *uni = sv_newmortal ();	1590	SV *uni = sv_newmortal ();
…		…
1406	SAVEVPTR (PL_curcop);	1596	SAVEVPTR (PL_curcop);
1407	PL_curcop = &cop;	1597	PL_curcop = &cop;
1408	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1598	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1409	LEAVE;	1599	LEAVE;
1410		1600
1411	croak ("%s, at character offset %d [\"%s\"]",	1601	croak ("%s, at character offset %d (before \"%s\")",
1412	dec.err,	1602	dec.err,
1413	(int)offset,	1603	ptr_to_index (string, dec.cur),
1414	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1604	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1415	}	1605	}
1416		1606
1417	sv = sv_2mortal (sv);	1607	sv = sv_2mortal (sv);
1418		1608
1419	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))	1609	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1420	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1610	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1421		1611
1422	return sv;	1612	return sv;
		1613	}
		1614
		1615	/////////////////////////////////////////////////////////////////////////////
		1616	// incremental parser
		1617
		1618	static void
		1619	incr_parse (JSON *self)
		1620	{
		1621	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1622
		1623	// the state machine here is a bit convoluted and could be simplified a lot
		1624	// but this would make it slower, so...
		1625
		1626	for (;;)
		1627	{
		1628	//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
		1629	switch (self->incr_mode)
		1630	{
		1631	// only used for initial whitespace skipping
		1632	case INCR_M_WS:
		1633	for (;;)
		1634	{
		1635	if (*p > 0x20)
		1636	{
		1637	if (*p == '#')
		1638	{
		1639	self->incr_mode = INCR_M_C0;
		1640	goto incr_m_c;
		1641	}
		1642	else
		1643	{
		1644	self->incr_mode = INCR_M_JSON;
		1645	goto incr_m_json;
		1646	}
		1647	}
		1648	else if (!*p)
		1649	goto interrupt;
		1650
		1651	++p;
		1652	}
		1653
		1654	// skip a single char inside a string (for \\-processing)
		1655	case INCR_M_BS:
		1656	if (!*p)
		1657	goto interrupt;
		1658
		1659	++p;
		1660	self->incr_mode = INCR_M_STR;
		1661	goto incr_m_str;
		1662
		1663	// inside #-style comments
		1664	case INCR_M_C0:
		1665	case INCR_M_C1:
		1666	incr_m_c:
		1667	for (;;)
		1668	{
		1669	if (*p == '\n')
		1670	{
		1671	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1672	break;
		1673	}
		1674	else if (!*p)
		1675	goto interrupt;
		1676
		1677	++p;
		1678	}
		1679
		1680	break;
		1681
		1682	// inside a string
		1683	case INCR_M_STR:
		1684	incr_m_str:
		1685	for (;;)
		1686	{
		1687	if (*p == '"')
		1688	{
		1689	++p;
		1690	self->incr_mode = INCR_M_JSON;
		1691
		1692	if (!self->incr_nest)
		1693	goto interrupt;
		1694
		1695	goto incr_m_json;
		1696	}
		1697	else if (*p == '\\')
		1698	{
		1699	++p; // "virtually" consumes character after \
		1700
		1701	if (!*p) // if at end of string we have to switch modes
		1702	{
		1703	self->incr_mode = INCR_M_BS;
		1704	goto interrupt;
		1705	}
		1706	}
		1707	else if (!*p)
		1708	goto interrupt;
		1709
		1710	++p;
		1711	}
		1712
		1713	// after initial ws, outside string
		1714	case INCR_M_JSON:
		1715	incr_m_json:
		1716	for (;;)
		1717	{
		1718	switch (*p++)
		1719	{
		1720	case 0:
		1721	--p;
		1722	goto interrupt;
		1723
		1724	case 0x09:
		1725	case 0x0a:
		1726	case 0x0d:
		1727	case 0x20:
		1728	if (!self->incr_nest)
		1729	{
		1730	--p; // do not eat the whitespace, let the next round do it
		1731	goto interrupt;
		1732	}
		1733	break;
		1734
		1735	case '"':
		1736	self->incr_mode = INCR_M_STR;
		1737	goto incr_m_str;
		1738
		1739	case '[':
		1740	case '{':
		1741	if (++self->incr_nest > self->max_depth)
		1742	croak (ERR_NESTING_EXCEEDED);
		1743	break;
		1744
		1745	case ']':
		1746	case '}':
		1747	if (--self->incr_nest <= 0)
		1748	goto interrupt;
		1749	break;
		1750
		1751	case '#':
		1752	self->incr_mode = INCR_M_C1;
		1753	goto incr_m_c;
		1754	}
		1755	}
		1756	}
		1757
		1758	modechange:
		1759	;
		1760	}
		1761
		1762	interrupt:
		1763	self->incr_pos = p - SvPVX (self->incr_text);
		1764	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1765	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1423	}	1766	}
1424		1767
1425	/////////////////////////////////////////////////////////////////////////////	1768	/////////////////////////////////////////////////////////////////////////////
1426	// XS interface functions	1769	// XS interface functions
1427		1770
…		…
1439	: -1;	1782	: -1;
1440		1783
1441	json_stash = gv_stashpv ("JSON::XS" , 1);	1784	json_stash = gv_stashpv ("JSON::XS" , 1);
1442	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);	1785	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1443		1786
1444	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );	1787	json_true = get_bool ("JSON::XS::true");
1445	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1788	json_false = get_bool ("JSON::XS::false");
		1789
		1790	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1446	}	1791	}
1447		1792
1448	PROTOTYPES: DISABLE	1793	PROTOTYPES: DISABLE
1449		1794
1450	void CLONE (...)	1795	void CLONE (...)
…		…
1455	void new (char *klass)	1800	void new (char *klass)
1456	PPCODE:	1801	PPCODE:
1457	{	1802	{
1458	SV *pv = NEWSV (0, sizeof (JSON));	1803	SV *pv = NEWSV (0, sizeof (JSON));
1459	SvPOK_only (pv);	1804	SvPOK_only (pv);
1460	Zero (SvPVX (pv), 1, JSON);	1805	json_init ((JSON *)SvPVX (pv));
1461	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	1806	XPUSHs (sv_2mortal (sv_bless (
1462	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH)));	1807	newRV_noinc (pv),
		1808	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1809	)));
1463	}	1810	}
1464		1811
1465	void ascii (JSON *self, int enable = 1)	1812	void ascii (JSON *self, int enable = 1)
1466	ALIAS:	1813	ALIAS:
1467	ascii = F_ASCII	1814	ascii = F_ASCII
…		…
1475	allow_nonref = F_ALLOW_NONREF	1822	allow_nonref = F_ALLOW_NONREF
1476	shrink = F_SHRINK	1823	shrink = F_SHRINK
1477	allow_blessed = F_ALLOW_BLESSED	1824	allow_blessed = F_ALLOW_BLESSED
1478	convert_blessed = F_CONV_BLESSED	1825	convert_blessed = F_CONV_BLESSED
1479	relaxed = F_RELAXED	1826	relaxed = F_RELAXED
		1827	allow_unknown = F_ALLOW_UNKNOWN
1480	PPCODE:	1828	PPCODE:
1481	{	1829	{
1482	if (enable)	1830	if (enable)
1483	self->flags \|= ix;	1831	self->flags \|= ix;
1484	else	1832	else
…		…
1494	get_utf8 = F_UTF8	1842	get_utf8 = F_UTF8
1495	get_indent = F_INDENT	1843	get_indent = F_INDENT
1496	get_canonical = F_CANONICAL	1844	get_canonical = F_CANONICAL
1497	get_space_before = F_SPACE_BEFORE	1845	get_space_before = F_SPACE_BEFORE
1498	get_space_after = F_SPACE_AFTER	1846	get_space_after = F_SPACE_AFTER
1499	get_pretty = F_PRETTY
1500	get_allow_nonref = F_ALLOW_NONREF	1847	get_allow_nonref = F_ALLOW_NONREF
1501	get_shrink = F_SHRINK	1848	get_shrink = F_SHRINK
1502	get_allow_blessed = F_ALLOW_BLESSED	1849	get_allow_blessed = F_ALLOW_BLESSED
1503	get_convert_blessed = F_CONV_BLESSED	1850	get_convert_blessed = F_CONV_BLESSED
1504	get_relaxed = F_RELAXED	1851	get_relaxed = F_RELAXED
		1852	get_allow_unknown = F_ALLOW_UNKNOWN
1505	PPCODE:	1853	PPCODE:
1506	XPUSHs (boolSV (self->flags & ix));	1854	XPUSHs (boolSV (self->flags & ix));
1507		1855
1508	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1856	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1509	PPCODE:	1857	PPCODE:
1510	{	1858	self->max_depth = max_depth;
1511	UV log2 = 0;
1512
1513	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1514
1515	while ((1UL << log2) < max_depth)
1516	++log2;
1517
1518	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1519
1520	XPUSHs (ST (0));	1859	XPUSHs (ST (0));
1521	}
1522		1860
1523	U32 get_max_depth (JSON *self)	1861	U32 get_max_depth (JSON *self)
1524	CODE:	1862	CODE:
1525	RETVAL = DEC_DEPTH (self->flags);	1863	RETVAL = self->max_depth;
1526	OUTPUT:	1864	OUTPUT:
1527	RETVAL	1865	RETVAL
1528		1866
1529	void max_size (JSON *self, UV max_size = 0)	1867	void max_size (JSON *self, U32 max_size = 0)
1530	PPCODE:	1868	PPCODE:
1531	{	1869	self->max_size = max_size;
1532	UV log2 = 0;
1533
1534	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1535	if (max_size == 1) max_size = 2;
1536
1537	while ((1UL << log2) < max_size)
1538	++log2;
1539
1540	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
1541
1542	XPUSHs (ST (0));	1870	XPUSHs (ST (0));
1543	}
1544		1871
1545	int get_max_size (JSON *self)	1872	int get_max_size (JSON *self)
1546	CODE:	1873	CODE:
1547	RETVAL = DEC_SIZE (self->flags);	1874	RETVAL = self->max_size;
1548	OUTPUT:	1875	OUTPUT:
1549	RETVAL	1876	RETVAL
1550		1877
1551	void filter_json_object (JSON self, SV cb = &PL_sv_undef)	1878	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1552	PPCODE:	1879	PPCODE:
…		…
1588	XPUSHs (decode_json (jsonstr, self, 0));	1915	XPUSHs (decode_json (jsonstr, self, 0));
1589		1916
1590	void decode_prefix (JSON self, SV jsonstr)	1917	void decode_prefix (JSON self, SV jsonstr)
1591	PPCODE:	1918	PPCODE:
1592	{	1919	{
1593	UV offset;	1920	char *offset;
1594	EXTEND (SP, 2);	1921	EXTEND (SP, 2);
1595	PUSHs (decode_json (jsonstr, self, &offset));	1922	PUSHs (decode_json (jsonstr, self, &offset));
1596	PUSHs (sv_2mortal (newSVuv (offset)));	1923	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		1924	}
		1925
		1926	void incr_parse (JSON self, SV jsonstr = 0)
		1927	PPCODE:
		1928	{
		1929	if (!self->incr_text)
		1930	self->incr_text = newSVpvn ("", 0);
		1931
		1932	// append data, if any
		1933	if (jsonstr)
		1934	{
		1935	if (SvUTF8 (jsonstr))
		1936	{
		1937	if (!SvUTF8 (self->incr_text))
		1938	{
		1939	/* utf-8-ness differs, need to upgrade */
		1940	sv_utf8_upgrade (self->incr_text);
		1941
		1942	if (self->incr_pos)
		1943	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1944	- (U8 *)SvPVX (self->incr_text);
		1945	}
		1946	}
		1947	else if (SvUTF8 (self->incr_text))
		1948	sv_utf8_upgrade (jsonstr);
		1949
		1950	{
		1951	STRLEN len;
		1952	const char *str = SvPV (jsonstr, len);
		1953	STRLEN cur = SvCUR (self->incr_text);
		1954
		1955	if (SvLEN (self->incr_text) <= cur + len)
		1956	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1957
		1958	Move (str, SvEND (self->incr_text), len, char);
		1959	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1960	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1961	}
		1962	}
		1963
		1964	if (GIMME_V != G_VOID)
		1965	do
		1966	{
		1967	char *offset;
		1968
		1969	if (!INCR_DONE (self))
		1970	{
		1971	incr_parse (self);
		1972
		1973	if (self->incr_pos > self->max_size && self->max_size)
		1974	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1975	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1976
		1977	if (!INCR_DONE (self))
		1978	break;
		1979	}
		1980
		1981	XPUSHs (decode_json (self->incr_text, self, &offset));
		1982
		1983	self->incr_pos -= offset - SvPVX (self->incr_text);
		1984	self->incr_nest = 0;
		1985	self->incr_mode = 0;
		1986
		1987	sv_chop (self->incr_text, offset);
		1988	}
		1989	while (GIMME_V == G_ARRAY);
		1990	}
		1991
		1992	SV incr_text (JSON self)
		1993	ATTRS: lvalue
		1994	CODE:
		1995	{
		1996	if (self->incr_pos)
		1997	croak ("incr_text can not be called when the incremental parser already started parsing");
		1998
		1999	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		2000	}
		2001	OUTPUT:
		2002	RETVAL
		2003
		2004	void incr_skip (JSON *self)
		2005	CODE:
		2006	{
		2007	if (self->incr_pos)
		2008	{
		2009	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2010	self->incr_pos = 0;
		2011	self->incr_nest = 0;
		2012	self->incr_mode = 0;
		2013	}
		2014	}
		2015
		2016	void incr_reset (JSON *self)
		2017	CODE:
		2018	{
		2019	SvREFCNT_dec (self->incr_text);
		2020	self->incr_text = 0;
		2021	self->incr_pos = 0;
		2022	self->incr_nest = 0;
		2023	self->incr_mode = 0;
1597	}	2024	}
1598		2025
1599	void DESTROY (JSON *self)	2026	void DESTROY (JSON *self)
1600	CODE:	2027	CODE:
1601	SvREFCNT_dec (self->cb_sk_object);	2028	SvREFCNT_dec (self->cb_sk_object);
1602	SvREFCNT_dec (self->cb_object);	2029	SvREFCNT_dec (self->cb_object);
		2030	SvREFCNT_dec (self->incr_text);
1603		2031
1604	PROTOTYPES: ENABLE	2032	PROTOTYPES: ENABLE
1605		2033
1606	void to_json (SV *scalar)	2034	void encode_json (SV *scalar)
		2035	ALIAS:
		2036	to_json_ = 0
		2037	encode_json = F_UTF8
1607	PPCODE:	2038	PPCODE:
1608	{	2039	{
1609	JSON json = { F_DEFAULT \| F_UTF8 };	2040	JSON json;
		2041	json_init (&json);
		2042	json.flags \|= ix;
1610	XPUSHs (encode_json (scalar, &json));	2043	XPUSHs (encode_json (scalar, &json));
1611	}	2044	}
1612		2045
1613	void from_json (SV *jsonstr)	2046	void decode_json (SV *jsonstr)
		2047	ALIAS:
		2048	from_json_ = 0
		2049	decode_json = F_UTF8
1614	PPCODE:	2050	PPCODE:
1615	{	2051	{
1616	JSON json = { F_DEFAULT \| F_UTF8 };	2052	JSON json;
		2053	json_init (&json);
		2054	json.flags \|= ix;
1617	XPUSHs (decode_json (jsonstr, &json, 0));	2055	XPUSHs (decode_json (jsonstr, &json, 0));
1618	}	2056	}
1619		2057

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.67 by root, Wed Nov 28 14:01:01 2007 UTC vs. Revision 1.106 by root, Tue Jan 19 01:36:34 2010 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.67 by root, Wed Nov 28 14:01:01 2007 UTC vs.
Revision 1.106 by root, Tue Jan 19 01:36:34 2010 UTC