[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 1.101 by root, Tue Sep 8 18:00:03 2009 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	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2)
20		23
21	#define F_ASCII 0x00000001UL	24	#define F_ASCII 0x00000001UL
22	#define F_LATIN1 0x00000002UL	25	#define F_LATIN1 0x00000002UL
23	#define F_UTF8 0x00000004UL	26	#define F_UTF8 0x00000004UL
24	#define F_INDENT 0x00000008UL	27	#define F_INDENT 0x00000008UL
…		…
28	#define F_ALLOW_NONREF 0x00000100UL	31	#define F_ALLOW_NONREF 0x00000100UL
29	#define F_SHRINK 0x00000200UL	32	#define F_SHRINK 0x00000200UL
30	#define F_ALLOW_BLESSED 0x00000400UL	33	#define F_ALLOW_BLESSED 0x00000400UL
31	#define F_CONV_BLESSED 0x00000800UL	34	#define F_CONV_BLESSED 0x00000800UL
32	#define F_RELAXED 0x00001000UL	35	#define F_RELAXED 0x00001000UL
33		36	#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	37	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
39		38
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	39	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER
44	#define F_DEFAULT (9UL << S_MAXDEPTH)
45		40
46	#define INIT_SIZE 32 // initial scalar size to be allocated	41	#define INIT_SIZE 32 // initial scalar size to be allocated
47	#define INDENT_STEP 3 // spaces per indentation level	42	#define INDENT_STEP 3 // spaces per indentation level
48		43
49	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size	44	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
50		45
51	#define SB do {	46	#define SB do {
52	#define SE } while (0)	47	#define SE } while (0)
53		48
54	#if __GNUC__ >= 3	49	#if __GNUC__ >= 3
55	# define expect(expr,value) __builtin_expect ((expr),(value))	50	# define expect(expr,value) __builtin_expect ((expr), (value))
56	# define inline inline	51	# define INLINE static inline
57	#else	52	#else
58	# define expect(expr,value) (expr)	53	# define expect(expr,value) (expr)
59	# define inline static	54	# define INLINE static
60	#endif	55	#endif
61		56
62	#define expect_false(expr) expect ((expr) != 0, 0)	57	#define expect_false(expr) expect ((expr) != 0, 0)
63	#define expect_true(expr) expect ((expr) != 0, 1)	58	#define expect_true(expr) expect ((expr) != 0, 1)
		59
		60	#define IN_RANGE_INC(type,val,beg,end) \
		61	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		62	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		63
		64	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
64		65
65	#ifdef USE_ITHREADS	66	#ifdef USE_ITHREADS
66	# define JSON_SLOW 1	67	# define JSON_SLOW 1
67	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))	68	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
68	#else	69	#else
…		…
71	#endif	72	#endif
72		73
73	static HV json_stash, json_boolean_stash; // JSON::XS::	74	static HV json_stash, json_boolean_stash; // JSON::XS::
74	static SV json_true, json_false;	75	static SV json_true, json_false;
75		76
		77	enum {
		78	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		79	INCR_M_STR, // inside string
		80	INCR_M_BS, // inside backslash
		81	INCR_M_JSON // outside anything, count nesting
		82	};
		83
		84	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		85
76	typedef struct {	86	typedef struct {
77	U32 flags;	87	U32 flags;
		88	U32 max_depth;
		89	STRLEN max_size;
		90
78	SV *cb_object;	91	SV *cb_object;
79	HV *cb_sk_object;	92	HV *cb_sk_object;
		93
		94	// for the incremental parser
		95	SV *incr_text; // the source text so far
		96	STRLEN incr_pos; // the current offset into the text
		97	int incr_nest; // {[]}-nesting level
		98	unsigned char incr_mode;
80	} JSON;	99	} JSON;
		100
		101	INLINE void
		102	json_init (JSON *json)
		103	{
		104	Zero (json, 1, JSON);
		105	json->max_depth = 512;
		106	}
81		107
82	/////////////////////////////////////////////////////////////////////////////	108	/////////////////////////////////////////////////////////////////////////////
83	// utility functions	109	// utility functions
84		110
85	inline void	111	INLINE SV *
		112	get_bool (const char *name)
		113	{
		114	SV *sv = get_sv (name, 1);
		115
		116	SvREADONLY_on (sv);
		117	SvREADONLY_on (SvRV (sv));
		118
		119	return sv;
		120	}
		121
		122	INLINE void
86	shrink (SV *sv)	123	shrink (SV *sv)
87	{	124	{
88	sv_utf8_downgrade (sv, 1);	125	sv_utf8_downgrade (sv, 1);
		126
89	if (SvLEN (sv) > SvCUR (sv) + 1)	127	if (SvLEN (sv) > SvCUR (sv) + 1)
90	{	128	{
91	#ifdef SvPV_shrink_to_cur	129	#ifdef SvPV_shrink_to_cur
92	SvPV_shrink_to_cur (sv);	130	SvPV_shrink_to_cur (sv);
93	#elif defined (SvPV_renew)	131	#elif defined (SvPV_renew)
…		…
99	// decode an utf-8 character and return it, or (UV)-1 in	137	// decode an utf-8 character and return it, or (UV)-1 in
100	// case of an error.	138	// case of an error.
101	// we special-case "safe" characters from U+80 .. U+7FF,	139	// we special-case "safe" characters from U+80 .. U+7FF,
102	// but use the very good perl function to parse anything else.	140	// but use the very good perl function to parse anything else.
103	// note that we never call this function for a ascii codepoints	141	// note that we never call this function for a ascii codepoints
104	inline UV	142	INLINE UV
105	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	143	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
106	{	144	{
107	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))	145	if (expect_true (len >= 2
108	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	146	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
109	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	147	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
110	{	148	{
111	*clen = 2;	149	*clen = 2;
112	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);	150	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
113	}	151	}
114	else	152	else
115	{	153	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
116	*clen = (STRLEN)-1;	154	}
117	return (UV)-1;	155
118	}	156	// likewise for encoding, also never called for ascii codepoints
		157	// this function takes advantage of this fact, although current gccs
		158	// seem to optimise the check for >= 0x80 away anyways
		159	INLINE unsigned char *
		160	encode_utf8 (unsigned char *s, UV ch)
		161	{
		162	if (expect_false (ch < 0x000080))
		163	*s++ = ch;
		164	else if (expect_true (ch < 0x000800))
		165	*s++ = 0xc0 \| ( ch >> 6),
		166	*s++ = 0x80 \| ( ch & 0x3f);
		167	else if ( ch < 0x010000)
		168	*s++ = 0xe0 \| ( ch >> 12),
		169	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		170	*s++ = 0x80 \| ( ch & 0x3f);
		171	else if ( ch < 0x110000)
		172	*s++ = 0xf0 \| ( ch >> 18),
		173	*s++ = 0x80 \| ((ch >> 12) & 0x3f),
		174	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		175	*s++ = 0x80 \| ( ch & 0x3f);
		176
		177	return s;
		178	}
		179
		180	// convert offset pointer to character index, sv must be string
		181	static STRLEN
		182	ptr_to_index (SV sv, char offset)
		183	{
		184	return SvUTF8 (sv)
		185	? utf8_distance (offset, SvPVX (sv))
		186	: offset - SvPVX (sv);
119	}	187	}
120		188
121	/////////////////////////////////////////////////////////////////////////////	189	/////////////////////////////////////////////////////////////////////////////
122	// encoder	190	// encoder
123		191
…		…
127	char *cur; // SvPVX (sv) + current output position	195	char *cur; // SvPVX (sv) + current output position
128	char *end; // SvEND (sv)	196	char *end; // SvEND (sv)
129	SV *sv; // result scalar	197	SV *sv; // result scalar
130	JSON json;	198	JSON json;
131	U32 indent; // indentation level	199	U32 indent; // indentation level
132	U32 maxdepth; // max. indentation/recursion level	200	UV limit; // escape character values >= this value when encoding
133	} enc_t;	201	} enc_t;
134		202
135	inline void	203	INLINE void
136	need (enc_t *enc, STRLEN len)	204	need (enc_t *enc, STRLEN len)
137	{	205	{
138	if (expect_false (enc->cur + len >= enc->end))	206	if (expect_false (enc->cur + len >= enc->end))
139	{	207	{
140	STRLEN cur = enc->cur - SvPVX (enc->sv);	208	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
141	SvGROW (enc->sv, cur + len + 1);	209	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
142	enc->cur = SvPVX (enc->sv) + cur;	210	enc->cur = SvPVX (enc->sv) + cur;
143	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	211	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
144	}	212	}
145	}	213	}
146		214
147	inline void	215	INLINE void
148	encode_ch (enc_t *enc, char ch)	216	encode_ch (enc_t *enc, char ch)
149	{	217	{
150	need (enc, 1);	218	need (enc, 1);
151	*enc->cur++ = ch;	219	*enc->cur++ = ch;
152	}	220	}
…		…
206	{	274	{
207	uch = ch;	275	uch = ch;
208	clen = 1;	276	clen = 1;
209	}	277	}
210		278
211	if (uch > 0x10FFFFUL)	279	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	{	280	{
216	if (uch > 0xFFFFUL)	281	if (uch >= 0x10000UL)
217	{	282	{
		283	if (uch >= 0x110000UL)
		284	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		285
218	need (enc, len += 11);	286	need (enc, len += 11);
219	sprintf (enc->cur, "\\u%04x\\u%04x",	287	sprintf (enc->cur, "\\u%04x\\u%04x",
220	(int)((uch - 0x10000) / 0x400 + 0xD800),	288	(int)((uch - 0x10000) / 0x400 + 0xD800),
221	(int)((uch - 0x10000) % 0x400 + 0xDC00));	289	(int)((uch - 0x10000) % 0x400 + 0xDC00));
222	enc->cur += 12;	290	enc->cur += 12;
223	}	291	}
224	else	292	else
225	{	293	{
226	static char hexdigit [16] = "0123456789abcdef";
227	need (enc, len += 5);	294	need (enc, len += 5);
228	*enc->cur++ = '\\';	295	*enc->cur++ = '\\';
229	*enc->cur++ = 'u';	296	*enc->cur++ = 'u';
230	*enc->cur++ = hexdigit [ uch >> 12 ];	297	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
231	*enc->cur++ = hexdigit [(uch >> 8) & 15];	298	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
232	*enc->cur++ = hexdigit [(uch >> 4) & 15];	299	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
233	*enc->cur++ = hexdigit [(uch >> 0) & 15];	300	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
234	}	301	}
235		302
236	str += clen;	303	str += clen;
237	}	304	}
238	else if (enc->json.flags & F_LATIN1)	305	else if (enc->json.flags & F_LATIN1)
…		…
250	while (--clen);	317	while (--clen);
251	}	318	}
252	else	319	else
253	{	320	{
254	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	321	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
255	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	322	enc->cur = encode_utf8 (enc->cur, uch);
256	++str;	323	++str;
257	}	324	}
258	}	325	}
259	}	326	}
260	}	327	}
261		328
262	--len;	329	--len;
263	}	330	}
264	}	331	}
265		332
266	inline void	333	INLINE void
267	encode_indent (enc_t *enc)	334	encode_indent (enc_t *enc)
268	{	335	{
269	if (enc->json.flags & F_INDENT)	336	if (enc->json.flags & F_INDENT)
270	{	337	{
271	int spaces = enc->indent * INDENT_STEP;	338	int spaces = enc->indent * INDENT_STEP;
…		…
274	memset (enc->cur, ' ', spaces);	341	memset (enc->cur, ' ', spaces);
275	enc->cur += spaces;	342	enc->cur += spaces;
276	}	343	}
277	}	344	}
278		345
279	inline void	346	INLINE void
280	encode_space (enc_t *enc)	347	encode_space (enc_t *enc)
281	{	348	{
282	need (enc, 1);	349	need (enc, 1);
283	encode_ch (enc, ' ');	350	encode_ch (enc, ' ');
284	}	351	}
285		352
286	inline void	353	INLINE void
287	encode_nl (enc_t *enc)	354	encode_nl (enc_t *enc)
288	{	355	{
289	if (enc->json.flags & F_INDENT)	356	if (enc->json.flags & F_INDENT)
290	{	357	{
291	need (enc, 1);	358	need (enc, 1);
292	encode_ch (enc, '\n');	359	encode_ch (enc, '\n');
293	}	360	}
294	}	361	}
295		362
296	inline void	363	INLINE void
297	encode_comma (enc_t *enc)	364	encode_comma (enc_t *enc)
298	{	365	{
299	encode_ch (enc, ',');	366	encode_ch (enc, ',');
300		367
301	if (enc->json.flags & F_INDENT)	368	if (enc->json.flags & F_INDENT)
…		…
309	static void	376	static void
310	encode_av (enc_t enc, AV av)	377	encode_av (enc_t enc, AV av)
311	{	378	{
312	int i, len = av_len (av);	379	int i, len = av_len (av);
313		380
314	if (enc->indent >= enc->maxdepth)	381	if (enc->indent >= enc->json.max_depth)
315	croak ("data structure too deep (hit recursion limit)");	382	croak (ERR_NESTING_EXCEEDED);
316		383
317	encode_ch (enc, '['); encode_nl (enc);	384	encode_ch (enc, '[');
318	++enc->indent;	385
		386	if (len >= 0)
		387	{
		388	encode_nl (enc); ++enc->indent;
319		389
320	for (i = 0; i <= len; ++i)	390	for (i = 0; i <= len; ++i)
321	{	391	{
322	SV **svp = av_fetch (av, i, 0);	392	SV **svp = av_fetch (av, i, 0);
323		393
324	encode_indent (enc);	394	encode_indent (enc);
325		395
326	if (svp)	396	if (svp)
327	encode_sv (enc, *svp);	397	encode_sv (enc, *svp);
328	else	398	else
329	encode_str (enc, "null", 4, 0);	399	encode_str (enc, "null", 4, 0);
330		400
331	if (i < len)	401	if (i < len)
332	encode_comma (enc);	402	encode_comma (enc);
333	}	403	}
334		404
		405	encode_nl (enc); --enc->indent; encode_indent (enc);
		406	}
		407
335	encode_nl (enc);	408	encode_ch (enc, ']');
336
337	--enc->indent;
338	encode_indent (enc); encode_ch (enc, ']');
339	}	409	}
340		410
341	static void	411	static void
342	encode_hk (enc_t enc, HE he)	412	encode_hk (enc_t enc, HE he)
343	{	413	{
…		…
391		461
392	static void	462	static void
393	encode_hv (enc_t enc, HV hv)	463	encode_hv (enc_t enc, HV hv)
394	{	464	{
395	HE *he;	465	HE *he;
396	int count;
397		466
398	if (enc->indent >= enc->maxdepth)	467	if (enc->indent >= enc->json.max_depth)
399	croak ("data structure too deep (hit recursion limit)");	468	croak (ERR_NESTING_EXCEEDED);
400		469
401	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	470	encode_ch (enc, '{');
402		471
403	// for canonical output we have to sort by keys first	472	// for canonical output we have to sort by keys first
404	// actually, this is mostly due to the stupid so-called	473	// actually, this is mostly due to the stupid so-called
405	// security workaround added somewhere in 5.8.x.	474	// security workaround added somewhere in 5.8.x
406	// that randomises hash orderings	475	// that randomises hash orderings
407	if (enc->json.flags & F_CANONICAL)	476	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
408	{	477	{
409	int count = hv_iterinit (hv);	478	int count = hv_iterinit (hv);
410		479
411	if (SvMAGICAL (hv))	480	if (SvMAGICAL (hv))
412	{	481	{
…		…
459		528
460	FREETMPS;	529	FREETMPS;
461	LEAVE;	530	LEAVE;
462	}	531	}
463		532
		533	encode_nl (enc); ++enc->indent;
		534
464	while (count--)	535	while (count--)
465	{	536	{
466	encode_indent (enc);	537	encode_indent (enc);
467	he = hes [count];	538	he = hes [count];
468	encode_hk (enc, he);	539	encode_hk (enc, he);
469	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));	540	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
470		541
471	if (count)	542	if (count)
472	encode_comma (enc);	543	encode_comma (enc);
473	}	544	}
		545
		546	encode_nl (enc); --enc->indent; encode_indent (enc);
474	}	547	}
475	}	548	}
476	else	549	else
477	{	550	{
478	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))	551	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))
479	if ((he = hv_iternext (hv)))	552	if ((he = hv_iternext (hv)))
		553	{
		554	encode_nl (enc); ++enc->indent;
		555
480	for (;;)	556	for (;;)
481	{	557	{
482	encode_indent (enc);	558	encode_indent (enc);
483	encode_hk (enc, he);	559	encode_hk (enc, he);
484	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));	560	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
485		561
486	if (!(he = hv_iternext (hv)))	562	if (!(he = hv_iternext (hv)))
487	break;	563	break;
488		564
489	encode_comma (enc);	565	encode_comma (enc);
490	}	566	}
491	}
492		567
		568	encode_nl (enc); --enc->indent; encode_indent (enc);
		569	}
		570	}
		571
493	encode_nl (enc);	572	encode_ch (enc, '}');
494
495	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
496	}	573	}
497		574
498	// encode objects, arrays and special \0=false and \1=true values.	575	// encode objects, arrays and special \0=false and \1=true values.
499	static void	576	static void
500	encode_rv (enc_t enc, SV sv)	577	encode_rv (enc_t enc, SV sv)
…		…
577		654
578	if (len == 1 && *pv == '1')	655	if (len == 1 && *pv == '1')
579	encode_str (enc, "true", 4, 0);	656	encode_str (enc, "true", 4, 0);
580	else if (len == 1 && *pv == '0')	657	else if (len == 1 && *pv == '0')
581	encode_str (enc, "false", 5, 0);	658	encode_str (enc, "false", 5, 0);
		659	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		660	encode_str (enc, "null", 4, 0);
582	else	661	else
583	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	662	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
584	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	663	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
585	}	664	}
		665	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		666	encode_str (enc, "null", 4, 0);
586	else	667	else
587	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	668	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
588	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	669	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
589	}	670	}
590		671
…		…
608	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	689	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
609	enc->cur += strlen (enc->cur);	690	enc->cur += strlen (enc->cur);
610	}	691	}
611	else if (SvIOKp (sv))	692	else if (SvIOKp (sv))
612	{	693	{
613	// we assume we can always read an IV as a UV	694	// we assume we can always read an IV as a UV and vice versa
614	if (SvUV (sv) & ~(UV)0x7fff)	695	// we assume two's complement
615	{	696	// we assume no aliasing issues in the union
616	// large integer, use the (rather slow) snprintf way.	697	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
617	need (enc, sizeof (UV) * 3);	698	: 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	{	699	{
625	// optimise the "small number case"	700	// optimise the "small number case"
626	// code will likely be branchless and use only a single multiplication	701	// code will likely be branchless and use only a single multiplication
		702	// works for numbers up to 59074
627	I32 i = SvIV (sv);	703	I32 i = SvIVX (sv);
628	U32 u;	704	U32 u;
629	char digit, nz = 0;	705	char digit, nz = 0;
630		706
631	need (enc, 6);	707	need (enc, 6);
632		708
…		…
638		714
639	// now output digit by digit, each time masking out the integer part	715	// 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,	716	// and multiplying by 5 while moving the decimal point one to the right,
641	// resulting in a net multiplication by 10.	717	// resulting in a net multiplication by 10.
642	// we always write the digit to memory but conditionally increment	718	// we always write the digit to memory but conditionally increment
643	// the pointer, to ease the usage of conditional move instructions.	719	// 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;	720	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;	721	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;	722	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;	723	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'	724	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
649	}	725	}
		726	else
		727	{
		728	// large integer, use the (rather slow) snprintf way.
		729	need (enc, IVUV_MAXCHARS);
		730	enc->cur +=
		731	SvIsUV(sv)
		732	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
		733	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		734	}
650	}	735	}
651	else if (SvROK (sv))	736	else if (SvROK (sv))
652	encode_rv (enc, SvRV (sv));	737	encode_rv (enc, SvRV (sv));
653	else if (!SvOK (sv))	738	else if (!SvOK (sv) \|\| enc->json.flags & F_ALLOW_UNKNOWN)
654	encode_str (enc, "null", 4, 0);	739	encode_str (enc, "null", 4, 0);
655	else	740	else
656	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	741	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
657	SvPV_nolen (sv), SvFLAGS (sv));	742	SvPV_nolen (sv), SvFLAGS (sv));
658	}	743	}
…		…
668	enc.json = *json;	753	enc.json = *json;
669	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	754	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
670	enc.cur = SvPVX (enc.sv);	755	enc.cur = SvPVX (enc.sv);
671	enc.end = SvEND (enc.sv);	756	enc.end = SvEND (enc.sv);
672	enc.indent = 0;	757	enc.indent = 0;
673	enc.maxdepth = DEC_DEPTH (enc.json.flags);	758	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		759	: enc.json.flags & F_LATIN1 ? 0x000100UL
		760	: 0x110000UL;
674		761
675	SvPOK_only (enc.sv);	762	SvPOK_only (enc.sv);
676	encode_sv (&enc, scalar);	763	encode_sv (&enc, scalar);
		764	encode_nl (&enc);
677		765
678	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	766	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
679	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	767	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
680		768
681	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))	769	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))
…		…
699	JSON json;	787	JSON json;
700	U32 depth; // recursion depth	788	U32 depth; // recursion depth
701	U32 maxdepth; // recursion depth limit	789	U32 maxdepth; // recursion depth limit
702	} dec_t;	790	} dec_t;
703		791
704	inline void	792	INLINE void
705	decode_comment (dec_t *dec)	793	decode_comment (dec_t *dec)
706	{	794	{
707	// only '#'-style comments allowed a.t.m.	795	// only '#'-style comments allowed a.t.m.
708		796
709	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)	797	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)
710	++dec->cur;	798	++dec->cur;
711	}	799	}
712		800
713	inline void	801	INLINE void
714	decode_ws (dec_t *dec)	802	decode_ws (dec_t *dec)
715	{	803	{
716	for (;;)	804	for (;;)
717	{	805	{
718	char ch = *dec->cur;	806	char ch = *dec->cur;
…		…
742	if (*dec->cur != ch) \	830	if (*dec->cur != ch) \
743	ERR (# ch " expected"); \	831	ERR (# ch " expected"); \
744	++dec->cur; \	832	++dec->cur; \
745	SE	833	SE
746		834
747	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	835	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
748	#define DEC_DEC_DEPTH --dec->depth	836	#define DEC_DEC_DEPTH --dec->depth
749		837
750	static SV decode_sv (dec_t dec);	838	static SV decode_sv (dec_t dec);
751		839
752	static signed char decode_hexdigit[256];	840	static signed char decode_hexdigit[256];
…		…
844		932
845	if (hi >= 0x80)	933	if (hi >= 0x80)
846	{	934	{
847	utf8 = 1;	935	utf8 = 1;
848		936
849	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	937	cur = encode_utf8 (cur, hi);
850	}	938	}
851	else	939	else
852	*cur++ = hi;	940	*cur++ = hi;
853	}	941	}
854	break;	942	break;
…		…
856	default:	944	default:
857	--dec_cur;	945	--dec_cur;
858	ERR ("illegal backslash escape sequence in string");	946	ERR ("illegal backslash escape sequence in string");
859	}	947	}
860	}	948	}
861	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	949	else if (expect_true (ch >= 0x20 && ch < 0x80))
862	*cur++ = ch;	950	*cur++ = ch;
863	else if (ch >= 0x80)	951	else if (ch >= 0x80)
864	{	952	{
865	STRLEN clen;	953	STRLEN clen;
866	UV uch;
867		954
868	--dec_cur;	955	--dec_cur;
869		956
870	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);	957	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
871	if (clen == (STRLEN)-1)	958	if (clen == (STRLEN)-1)
872	ERR ("malformed UTF-8 character in JSON string");	959	ERR ("malformed UTF-8 character in JSON string");
873		960
874	do	961	do
875	cur++ = dec_cur++;	962	cur++ = dec_cur++;
…		…
892	{	979	{
893	STRLEN len = cur - buf;	980	STRLEN len = cur - buf;
894		981
895	if (sv)	982	if (sv)
896	{	983	{
897	SvGROW (sv, SvCUR (sv) + len + 1);	984	STRLEN cur = SvCUR (sv);
		985
		986	if (SvLEN (sv) <= cur + len)
		987	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		988
898	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	989	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
899	SvCUR_set (sv, SvCUR (sv) + len);	990	SvCUR_set (sv, SvCUR (sv) + len);
900	}	991	}
901	else	992	else
902	sv = newSVpvn (buf, len);	993	sv = newSVpvn (buf, len);
…		…
989		1080
990	if (!is_nv)	1081	if (!is_nv)
991	{	1082	{
992	int len = dec->cur - start;	1083	int len = dec->cur - start;
993		1084
994	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1085	// special case the rather common 1..5-digit-int case
995	if (*start == '-')	1086	if (*start == '-')
996	switch (len)	1087	switch (len)
997	{	1088	{
998	case 2: return newSViv (-( start [1] - '0' * 1));	1089	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
999	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1090	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));	1091	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));	1092	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1093	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1002	}	1094	}
1003	else	1095	else
1004	switch (len)	1096	switch (len)
1005	{	1097	{
1006	case 1: return newSViv ( start [0] - '0' * 1);	1098	case 1: return newSViv ( start [0] - '0' * 1);
1007	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1099	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);	1100	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);	1101	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1102	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1010	}	1103	}
1011		1104
1012	{	1105	{
1013	UV uv;	1106	UV uv;
1014	int numtype = grok_number (start, len, &uv);	1107	int numtype = grok_number (start, len, &uv);
…		…
1122	char *p = dec->cur;	1215	char *p = dec->cur;
1123	char *e = p + 24; // only try up to 24 bytes	1216	char *e = p + 24; // only try up to 24 bytes
1124		1217
1125	for (;;)	1218	for (;;)
1126	{	1219	{
1127	// the >= 0x80 is true on most architectures	1220	// the >= 0x80 is false on most architectures
1128	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')	1221	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
1129	{	1222	{
1130	// slow path, back up and use decode_str	1223	// slow path, back up and use decode_str
1131	SV *key = decode_str (dec);	1224	SV *key = decode_str (dec);
1132	if (!key)	1225	if (!key)
…		…
1263		1356
1264	static SV *	1357	static SV *
1265	decode_sv (dec_t *dec)	1358	decode_sv (dec_t *dec)
1266	{	1359	{
1267	// the beauty of JSON: you need exactly one character lookahead	1360	// the beauty of JSON: you need exactly one character lookahead
1268	// to parse anything.	1361	// to parse everything.
1269	switch (*dec->cur)	1362	switch (*dec->cur)
1270	{	1363	{
1271	case '"': ++dec->cur; return decode_str (dec);	1364	case '"': ++dec->cur; return decode_str (dec);
1272	case '[': ++dec->cur; return decode_av (dec);	1365	case '[': ++dec->cur; return decode_av (dec);
1273	case '{': ++dec->cur; return decode_hv (dec);	1366	case '{': ++dec->cur; return decode_hv (dec);
1274		1367
1275	case '-':	1368	case '-':
1276	case '0': case '1': case '2': case '3': case '4':	1369	case '0': case '1': case '2': case '3': case '4':
1277	case '5': case '6': case '7': case '8': case '9':	1370	case '5': case '6': case '7': case '8': case '9':
1278	return decode_num (dec);	1371	return decode_num (dec);
…		…
1280	case 't':	1373	case 't':
1281	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1374	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1282	{	1375	{
1283	dec->cur += 4;	1376	dec->cur += 4;
1284	#if JSON_SLOW	1377	#if JSON_SLOW
1285	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);	1378	json_true = get_bool ("JSON::XS::true");
1286	#endif	1379	#endif
1287	return SvREFCNT_inc (json_true);	1380	return newSVsv (json_true);
1288	}	1381	}
1289	else	1382	else
1290	ERR ("'true' expected");	1383	ERR ("'true' expected");
1291		1384
1292	break;	1385	break;
…		…
1294	case 'f':	1387	case 'f':
1295	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1388	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1296	{	1389	{
1297	dec->cur += 5;	1390	dec->cur += 5;
1298	#if JSON_SLOW	1391	#if JSON_SLOW
1299	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1392	json_false = get_bool ("JSON::XS::false");
1300	#endif	1393	#endif
1301	return SvREFCNT_inc (json_false);	1394	return newSVsv (json_false);
1302	}	1395	}
1303	else	1396	else
1304	ERR ("'false' expected");	1397	ERR ("'false' expected");
1305		1398
1306	break;	1399	break;
…		…
1324	fail:	1417	fail:
1325	return 0;	1418	return 0;
1326	}	1419	}
1327		1420
1328	static SV *	1421	static SV *
1329	decode_json (SV string, JSON json, UV *offset_return)	1422	decode_json (SV string, JSON json, char **offset_return)
1330	{	1423	{
1331	dec_t dec;	1424	dec_t dec;
1332	UV offset;
1333	SV *sv;	1425	SV *sv;
1334		1426
		1427	/* work around bugs in 5.10 where manipulating magic values
		1428	* will perl ignore the magic in subsequent accesses
		1429	*/
1335	SvGETMAGIC (string);	1430	/SvGETMAGIC (string);/
		1431	if (SvMAGICAL (string))
		1432	string = sv_2mortal (newSVsv (string));
		1433
1336	SvUPGRADE (string, SVt_PV);	1434	SvUPGRADE (string, SVt_PV);
1337		1435
1338	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))	1436	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1437	* assertion with -DDEBUGGING, although SvCUR is documented to
		1438	* return the xpv_cur field which certainly exists after upgrading.
		1439	* according to nicholas clark, calling SvPOK fixes this.
		1440	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1441	* and hope for the best.
		1442	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1443	* assertion business is seriously broken, try yet another workaround
		1444	* for the broken -DDEBUGGING.
		1445	*/
		1446	{
		1447	#ifdef DEBUGGING
		1448	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1449	#else
		1450	STRLEN offset = SvCUR (string);
		1451	#endif
		1452
		1453	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",	1454	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));	1455	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1456	}
1341		1457
1342	if (json->flags & F_UTF8)	1458	if (json->flags & F_UTF8)
1343	sv_utf8_downgrade (string, 0);	1459	sv_utf8_downgrade (string, 0);
1344	else	1460	else
1345	sv_utf8_upgrade (string);	1461	sv_utf8_upgrade (string);
1346		1462
1347	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1463	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1348		1464
1349	dec.json = *json;	1465	dec.json = *json;
1350	dec.cur = SvPVX (string);	1466	dec.cur = SvPVX (string);
1351	dec.end = SvEND (string);	1467	dec.end = SvEND (string);
1352	dec.err = 0;	1468	dec.err = 0;
1353	dec.depth = 0;	1469	dec.depth = 0;
1354	dec.maxdepth = DEC_DEPTH (dec.json.flags);
1355		1470
1356	if (dec.json.cb_object \|\| dec.json.cb_sk_object)	1471	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
1357	dec.json.flags \|= F_HOOK;	1472	dec.json.flags \|= F_HOOK;
1358		1473
1359	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1474	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
1360		1475
1361	decode_ws (&dec);	1476	decode_ws (&dec);
1362	sv = decode_sv (&dec);	1477	sv = decode_sv (&dec);
		1478
		1479	if (offset_return)
		1480	*offset_return = dec.cur;
1363		1481
1364	if (!(offset_return \|\| !sv))	1482	if (!(offset_return \|\| !sv))
1365	{	1483	{
1366	// check for trailing garbage	1484	// check for trailing garbage
1367	decode_ws (&dec);	1485	decode_ws (&dec);
…		…
1370	{	1488	{
1371	dec.err = "garbage after JSON object";	1489	dec.err = "garbage after JSON object";
1372	SvREFCNT_dec (sv);	1490	SvREFCNT_dec (sv);
1373	sv = 0;	1491	sv = 0;
1374	}	1492	}
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	}	1493	}
1386		1494
1387	if (!sv)	1495	if (!sv)
1388	{	1496	{
1389	SV *uni = sv_newmortal ();	1497	SV *uni = sv_newmortal ();
…		…
1395	SAVEVPTR (PL_curcop);	1503	SAVEVPTR (PL_curcop);
1396	PL_curcop = &cop;	1504	PL_curcop = &cop;
1397	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1505	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1398	LEAVE;	1506	LEAVE;
1399		1507
1400	croak ("%s, at character offset %d [\"%s\"]",	1508	croak ("%s, at character offset %d (before \"%s\")",
1401	dec.err,	1509	dec.err,
1402	(int)offset,	1510	ptr_to_index (string, dec.cur),
1403	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1511	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1404	}	1512	}
1405		1513
1406	sv = sv_2mortal (sv);	1514	sv = sv_2mortal (sv);
1407		1515
1408	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))	1516	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)");	1517	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1410		1518
1411	return sv;	1519	return sv;
		1520	}
		1521
		1522	/////////////////////////////////////////////////////////////////////////////
		1523	// incremental parser
		1524
		1525	static void
		1526	incr_parse (JSON *self)
		1527	{
		1528	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1529
		1530	for (;;)
		1531	{
		1532	//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
		1533	switch (self->incr_mode)
		1534	{
		1535	// only used for intiial whitespace skipping
		1536	case INCR_M_WS:
		1537	for (;;)
		1538	{
		1539	if (*p > 0x20)
		1540	{
		1541	self->incr_mode = INCR_M_JSON;
		1542	goto incr_m_json;
		1543	}
		1544	else if (!*p)
		1545	goto interrupt;
		1546
		1547	++p;
		1548	}
		1549
		1550	// skip a single char inside a string (for \\-processing)
		1551	case INCR_M_BS:
		1552	if (!*p)
		1553	goto interrupt;
		1554
		1555	++p;
		1556	self->incr_mode = INCR_M_STR;
		1557	goto incr_m_str;
		1558
		1559	// inside a string
		1560	case INCR_M_STR:
		1561	incr_m_str:
		1562	for (;;)
		1563	{
		1564	if (*p == '"')
		1565	{
		1566	++p;
		1567	self->incr_mode = INCR_M_JSON;
		1568
		1569	if (!self->incr_nest)
		1570	goto interrupt;
		1571
		1572	goto incr_m_json;
		1573	}
		1574	else if (*p == '\\')
		1575	{
		1576	++p; // "virtually" consumes character after \
		1577
		1578	if (!*p) // if at end of string we have to switch modes
		1579	{
		1580	self->incr_mode = INCR_M_BS;
		1581	goto interrupt;
		1582	}
		1583	}
		1584	else if (!*p)
		1585	goto interrupt;
		1586
		1587	++p;
		1588	}
		1589
		1590	// after initial ws, outside string
		1591	case INCR_M_JSON:
		1592	incr_m_json:
		1593	for (;;)
		1594	{
		1595	switch (*p++)
		1596	{
		1597	case 0:
		1598	--p;
		1599	goto interrupt;
		1600
		1601	case 0x09:
		1602	case 0x0a:
		1603	case 0x0d:
		1604	case 0x20:
		1605	if (!self->incr_nest)
		1606	{
		1607	--p; // do not eat the whitespace, let the next round do it
		1608	goto interrupt;
		1609	}
		1610	break;
		1611
		1612	case '"':
		1613	self->incr_mode = INCR_M_STR;
		1614	goto incr_m_str;
		1615
		1616	case '[':
		1617	case '{':
		1618	if (++self->incr_nest > self->max_depth)
		1619	croak (ERR_NESTING_EXCEEDED);
		1620	break;
		1621
		1622	case ']':
		1623	case '}':
		1624	if (--self->incr_nest <= 0)
		1625	goto interrupt;
		1626	}
		1627	}
		1628	}
		1629
		1630	modechange:
		1631	;
		1632	}
		1633
		1634	interrupt:
		1635	self->incr_pos = p - SvPVX (self->incr_text);
		1636	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1412	}	1637	}
1413		1638
1414	/////////////////////////////////////////////////////////////////////////////	1639	/////////////////////////////////////////////////////////////////////////////
1415	// XS interface functions	1640	// XS interface functions
1416		1641
…		…
1428	: -1;	1653	: -1;
1429		1654
1430	json_stash = gv_stashpv ("JSON::XS" , 1);	1655	json_stash = gv_stashpv ("JSON::XS" , 1);
1431	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);	1656	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1432		1657
1433	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );	1658	json_true = get_bool ("JSON::XS::true");
1434	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1659	json_false = get_bool ("JSON::XS::false");
		1660
		1661	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1435	}	1662	}
1436		1663
1437	PROTOTYPES: DISABLE	1664	PROTOTYPES: DISABLE
1438		1665
1439	void CLONE (...)	1666	void CLONE (...)
…		…
1444	void new (char *klass)	1671	void new (char *klass)
1445	PPCODE:	1672	PPCODE:
1446	{	1673	{
1447	SV *pv = NEWSV (0, sizeof (JSON));	1674	SV *pv = NEWSV (0, sizeof (JSON));
1448	SvPOK_only (pv);	1675	SvPOK_only (pv);
1449	Zero (SvPVX (pv), 1, JSON);	1676	json_init ((JSON *)SvPVX (pv));
1450	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	1677	XPUSHs (sv_2mortal (sv_bless (
1451	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH)));	1678	newRV_noinc (pv),
		1679	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1680	)));
1452	}	1681	}
1453		1682
1454	void ascii (JSON *self, int enable = 1)	1683	void ascii (JSON *self, int enable = 1)
1455	ALIAS:	1684	ALIAS:
1456	ascii = F_ASCII	1685	ascii = F_ASCII
…		…
1464	allow_nonref = F_ALLOW_NONREF	1693	allow_nonref = F_ALLOW_NONREF
1465	shrink = F_SHRINK	1694	shrink = F_SHRINK
1466	allow_blessed = F_ALLOW_BLESSED	1695	allow_blessed = F_ALLOW_BLESSED
1467	convert_blessed = F_CONV_BLESSED	1696	convert_blessed = F_CONV_BLESSED
1468	relaxed = F_RELAXED	1697	relaxed = F_RELAXED
		1698	allow_unknown = F_ALLOW_UNKNOWN
1469	PPCODE:	1699	PPCODE:
1470	{	1700	{
1471	if (enable)	1701	if (enable)
1472	self->flags \|= ix;	1702	self->flags \|= ix;
1473	else	1703	else
1474	self->flags &= ~ix;	1704	self->flags &= ~ix;
1475		1705
1476	XPUSHs (ST (0));	1706	XPUSHs (ST (0));
1477	}	1707	}
1478		1708
1479	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1709	void get_ascii (JSON *self)
		1710	ALIAS:
		1711	get_ascii = F_ASCII
		1712	get_latin1 = F_LATIN1
		1713	get_utf8 = F_UTF8
		1714	get_indent = F_INDENT
		1715	get_canonical = F_CANONICAL
		1716	get_space_before = F_SPACE_BEFORE
		1717	get_space_after = F_SPACE_AFTER
		1718	get_allow_nonref = F_ALLOW_NONREF
		1719	get_shrink = F_SHRINK
		1720	get_allow_blessed = F_ALLOW_BLESSED
		1721	get_convert_blessed = F_CONV_BLESSED
		1722	get_relaxed = F_RELAXED
		1723	get_allow_unknown = F_ALLOW_UNKNOWN
1480	PPCODE:	1724	PPCODE:
1481	{	1725	XPUSHs (boolSV (self->flags & ix));
1482	UV log2 = 0;
1483		1726
1484	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	1727	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1485		1728	PPCODE:
1486	while ((1UL << log2) < max_depth)	1729	self->max_depth = max_depth;
1487	++log2;
1488
1489	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1490
1491	XPUSHs (ST (0));	1730	XPUSHs (ST (0));
1492	}
1493		1731
		1732	U32 get_max_depth (JSON *self)
		1733	CODE:
		1734	RETVAL = self->max_depth;
		1735	OUTPUT:
		1736	RETVAL
		1737
1494	void max_size (JSON *self, UV max_size = 0)	1738	void max_size (JSON *self, U32 max_size = 0)
1495	PPCODE:	1739	PPCODE:
1496	{	1740	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));	1741	XPUSHs (ST (0));
1508	}	1742
		1743	int get_max_size (JSON *self)
		1744	CODE:
		1745	RETVAL = self->max_size;
		1746	OUTPUT:
		1747	RETVAL
1509		1748
1510	void filter_json_object (JSON self, SV cb = &PL_sv_undef)	1749	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1511	PPCODE:	1750	PPCODE:
1512	{	1751	{
1513	SvREFCNT_dec (self->cb_object);	1752	SvREFCNT_dec (self->cb_object);
…		…
1547	XPUSHs (decode_json (jsonstr, self, 0));	1786	XPUSHs (decode_json (jsonstr, self, 0));
1548		1787
1549	void decode_prefix (JSON self, SV jsonstr)	1788	void decode_prefix (JSON self, SV jsonstr)
1550	PPCODE:	1789	PPCODE:
1551	{	1790	{
1552	UV offset;	1791	char *offset;
1553	EXTEND (SP, 2);	1792	EXTEND (SP, 2);
1554	PUSHs (decode_json (jsonstr, self, &offset));	1793	PUSHs (decode_json (jsonstr, self, &offset));
1555	PUSHs (sv_2mortal (newSVuv (offset)));	1794	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		1795	}
		1796
		1797	void incr_parse (JSON self, SV jsonstr = 0)
		1798	PPCODE:
		1799	{
		1800	if (!self->incr_text)
		1801	self->incr_text = newSVpvn ("", 0);
		1802
		1803	// append data, if any
		1804	if (jsonstr)
		1805	{
		1806	if (SvUTF8 (jsonstr))
		1807	{
		1808	if (!SvUTF8 (self->incr_text))
		1809	{
		1810	/* utf-8-ness differs, need to upgrade */
		1811	sv_utf8_upgrade (self->incr_text);
		1812
		1813	if (self->incr_pos)
		1814	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1815	- (U8 *)SvPVX (self->incr_text);
		1816	}
		1817	}
		1818	else if (SvUTF8 (self->incr_text))
		1819	sv_utf8_upgrade (jsonstr);
		1820
		1821	{
		1822	STRLEN len;
		1823	const char *str = SvPV (jsonstr, len);
		1824	STRLEN cur = SvCUR (self->incr_text);
		1825
		1826	if (SvLEN (self->incr_text) <= cur + len)
		1827	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1828
		1829	Move (str, SvEND (self->incr_text), len, char);
		1830	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1831	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1832	}
		1833	}
		1834
		1835	if (GIMME_V != G_VOID)
		1836	do
		1837	{
		1838	char *offset;
		1839
		1840	if (!INCR_DONE (self))
		1841	{
		1842	incr_parse (self);
		1843
		1844	if (self->incr_pos > self->max_size && self->max_size)
		1845	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1846	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1847
		1848	if (!INCR_DONE (self))
		1849	break;
		1850	}
		1851
		1852	XPUSHs (decode_json (self->incr_text, self, &offset));
		1853
		1854	self->incr_pos -= offset - SvPVX (self->incr_text);
		1855	self->incr_nest = 0;
		1856	self->incr_mode = 0;
		1857
		1858	sv_chop (self->incr_text, offset);
		1859	}
		1860	while (GIMME_V == G_ARRAY);
		1861	}
		1862
		1863	SV incr_text (JSON self)
		1864	ATTRS: lvalue
		1865	CODE:
		1866	{
		1867	if (self->incr_pos)
		1868	croak ("incr_text can not be called when the incremental parser already started parsing");
		1869
		1870	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		1871	}
		1872	OUTPUT:
		1873	RETVAL
		1874
		1875	void incr_skip (JSON *self)
		1876	CODE:
		1877	{
		1878	if (self->incr_pos)
		1879	{
		1880	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1881	self->incr_pos = 0;
		1882	self->incr_nest = 0;
		1883	self->incr_mode = 0;
		1884	}
		1885	}
		1886
		1887	void incr_reset (JSON *self)
		1888	CODE:
		1889	{
		1890	SvREFCNT_dec (self->incr_text);
		1891	self->incr_text = 0;
		1892	self->incr_pos = 0;
		1893	self->incr_nest = 0;
		1894	self->incr_mode = 0;
1556	}	1895	}
1557		1896
1558	void DESTROY (JSON *self)	1897	void DESTROY (JSON *self)
1559	CODE:	1898	CODE:
1560	SvREFCNT_dec (self->cb_sk_object);	1899	SvREFCNT_dec (self->cb_sk_object);
1561	SvREFCNT_dec (self->cb_object);	1900	SvREFCNT_dec (self->cb_object);
		1901	SvREFCNT_dec (self->incr_text);
1562		1902
1563	PROTOTYPES: ENABLE	1903	PROTOTYPES: ENABLE
1564		1904
1565	void to_json (SV *scalar)	1905	void encode_json (SV *scalar)
		1906	ALIAS:
		1907	to_json_ = 0
		1908	encode_json = F_UTF8
1566	PPCODE:	1909	PPCODE:
1567	{	1910	{
1568	JSON json = { F_DEFAULT \| F_UTF8 };	1911	JSON json;
		1912	json_init (&json);
		1913	json.flags \|= ix;
1569	XPUSHs (encode_json (scalar, &json));	1914	XPUSHs (encode_json (scalar, &json));
1570	}	1915	}
1571		1916
1572	void from_json (SV *jsonstr)	1917	void decode_json (SV *jsonstr)
		1918	ALIAS:
		1919	from_json_ = 0
		1920	decode_json = F_UTF8
1573	PPCODE:	1921	PPCODE:
1574	{	1922	{
1575	JSON json = { F_DEFAULT \| F_UTF8 };	1923	JSON json;
		1924	json_init (&json);
		1925	json.flags \|= ix;
1576	XPUSHs (decode_json (jsonstr, &json, 0));	1926	XPUSHs (decode_json (jsonstr, &json, 0));
1577	}	1927	}
1578		1928
		1929

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 1.101 by root, Tue Sep 8 18:00:03 2009 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 1.101 by root, Tue Sep 8 18:00:03 2009 UTC