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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.71 by root, Wed Mar 19 03:17:38 2008 UTC vs.
Revision 1.97 by root, Fri Jul 3 09:38:30 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
…		…
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	// guarentees, 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
…		…
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)
64		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?)"
		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
69	# define JSON_SLOW 0	70	# define JSON_SLOW 0
…		…
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
		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
85	INLINE void	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)
…		…
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
133	UV limit; // escape character values >= this value when encoding	200	UV limit; // escape character values >= this value when encoding
134	} enc_t;	201	} enc_t;
135		202
136	INLINE void	203	INLINE void
137	need (enc_t *enc, STRLEN len)	204	need (enc_t *enc, STRLEN len)
138	{	205	{
139	if (expect_false (enc->cur + len >= enc->end))	206	if (expect_false (enc->cur + len >= enc->end))
140	{	207	{
141	STRLEN cur = enc->cur - SvPVX (enc->sv);	208	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
142	SvGROW (enc->sv, cur + len + 1);	209	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
143	enc->cur = SvPVX (enc->sv) + cur;	210	enc->cur = SvPVX (enc->sv) + cur;
144	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	211	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
145	}	212	}
146	}	213	}
147		214
…		…
207	{	274	{
208	uch = ch;	275	uch = ch;
209	clen = 1;	276	clen = 1;
210	}	277	}
211		278
212	if (uch < 0x20 \|\| uch >= enc->limit)	279	if (uch < 0x80/0x20/ \|\| uch >= enc->limit)
213	{	280	{
214	if (uch > 0xFFFFUL)	281	if (uch >= 0x10000UL)
215	{	282	{
216	if (uch > 0x10FFFFUL)	283	if (uch >= 0x110000UL)
217	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);	284	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
218		285
219	need (enc, len += 11);	286	need (enc, len += 11);
220	sprintf (enc->cur, "\\u%04x\\u%04x",	287	sprintf (enc->cur, "\\u%04x\\u%04x",
221	(int)((uch - 0x10000) / 0x400 + 0xD800),	288	(int)((uch - 0x10000) / 0x400 + 0xD800),
222	(int)((uch - 0x10000) % 0x400 + 0xDC00));	289	(int)((uch - 0x10000) % 0x400 + 0xDC00));
223	enc->cur += 12;	290	enc->cur += 12;
224	}	291	}
225	else	292	else
226	{	293	{
227	static char hexdigit [16] = "0123456789abcdef";
228	need (enc, len += 5);	294	need (enc, len += 5);
229	*enc->cur++ = '\\';	295	*enc->cur++ = '\\';
230	*enc->cur++ = 'u';	296	*enc->cur++ = 'u';
231	*enc->cur++ = hexdigit [ uch >> 12 ];	297	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
232	*enc->cur++ = hexdigit [(uch >> 8) & 15];	298	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
233	*enc->cur++ = hexdigit [(uch >> 4) & 15];	299	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
234	*enc->cur++ = hexdigit [(uch >> 0) & 15];	300	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
235	}	301	}
236		302
237	str += clen;	303	str += clen;
238	}	304	}
239	else if (enc->json.flags & F_LATIN1)	305	else if (enc->json.flags & F_LATIN1)
…		…
251	while (--clen);	317	while (--clen);
252	}	318	}
253	else	319	else
254	{	320	{
255	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
256	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	322	enc->cur = encode_utf8 (enc->cur, uch);
257	++str;	323	++str;
258	}	324	}
259	}	325	}
260	}	326	}
261	}	327	}
…		…
310	static void	376	static void
311	encode_av (enc_t enc, AV av)	377	encode_av (enc_t enc, AV av)
312	{	378	{
313	int i, len = av_len (av);	379	int i, len = av_len (av);
314		380
315	if (enc->indent >= enc->maxdepth)	381	if (enc->indent >= enc->json.max_depth)
316	croak ("data structure too deep (hit recursion limit)");	382	croak (ERR_NESTING_EXCEEDED);
317		383
318	encode_ch (enc, '[');	384	encode_ch (enc, '[');
319		385
320	if (len >= 0)	386	if (len >= 0)
321	{	387	{
…		…
395		461
396	static void	462	static void
397	encode_hv (enc_t enc, HV hv)	463	encode_hv (enc_t enc, HV hv)
398	{	464	{
399	HE *he;	465	HE *he;
400	int count;
401		466
402	if (enc->indent >= enc->maxdepth)	467	if (enc->indent >= enc->json.max_depth)
403	croak ("data structure too deep (hit recursion limit)");	468	croak (ERR_NESTING_EXCEEDED);
404		469
405	encode_ch (enc, '{');	470	encode_ch (enc, '{');
406		471
407	// for canonical output we have to sort by keys first	472	// for canonical output we have to sort by keys first
408	// actually, this is mostly due to the stupid so-called	473	// actually, this is mostly due to the stupid so-called
409	// security workaround added somewhere in 5.8.x.	474	// security workaround added somewhere in 5.8.x
410	// that randomises hash orderings	475	// that randomises hash orderings
411	if (enc->json.flags & F_CANONICAL)	476	if (enc->json.flags & F_CANONICAL)
412	{	477	{
413	int count = hv_iterinit (hv);	478	int count = hv_iterinit (hv);
414		479
…		…
589		654
590	if (len == 1 && *pv == '1')	655	if (len == 1 && *pv == '1')
591	encode_str (enc, "true", 4, 0);	656	encode_str (enc, "true", 4, 0);
592	else if (len == 1 && *pv == '0')	657	else if (len == 1 && *pv == '0')
593	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);
594	else	661	else
595	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",
596	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	663	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
597	}	664	}
		665	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		666	encode_str (enc, "null", 4, 0);
598	else	667	else
599	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",
600	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	669	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
601	}	670	}
602		671
…		…
620	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	689	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
621	enc->cur += strlen (enc->cur);	690	enc->cur += strlen (enc->cur);
622	}	691	}
623	else if (SvIOKp (sv))	692	else if (SvIOKp (sv))
624	{	693	{
625	// 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
626	if (SvUV (sv) & ~(UV)0x7fff)	695	// we assume two's complement
627	{	696	// we assume no aliasing issues in the union
628	// large integer, use the (rather slow) snprintf way.	697	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
629	need (enc, sizeof (UV) * 3);	698	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
630	enc->cur +=
631	SvIsUV(sv)
632	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
633	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
634	}
635	else
636	{	699	{
637	// optimise the "small number case"	700	// optimise the "small number case"
638	// 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
639	I32 i = SvIV (sv);	703	I32 i = SvIVX (sv);
640	U32 u;	704	U32 u;
641	char digit, nz = 0;	705	char digit, nz = 0;
642		706
643	need (enc, 6);	707	need (enc, 6);
644		708
…		…
650		714
651	// 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
652	// 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,
653	// resulting in a net multiplication by 10.	717	// resulting in a net multiplication by 10.
654	// we always write the digit to memory but conditionally increment	718	// we always write the digit to memory but conditionally increment
655	// the pointer, to ease the usage of conditional move instructions.	719	// the pointer, to enable the use of conditional move instructions.
656	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;
657	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;
658	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;
659	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;
660	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'
661	}	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	}
662	}	735	}
663	else if (SvROK (sv))	736	else if (SvROK (sv))
664	encode_rv (enc, SvRV (sv));	737	encode_rv (enc, SvRV (sv));
665	else if (!SvOK (sv))	738	else if (!SvOK (sv) \|\| enc->json.flags & F_ALLOW_UNKNOWN)
666	encode_str (enc, "null", 4, 0);	739	encode_str (enc, "null", 4, 0);
667	else	740	else
668	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",
669	SvPV_nolen (sv), SvFLAGS (sv));	742	SvPV_nolen (sv), SvFLAGS (sv));
670	}	743	}
…		…
680	enc.json = *json;	753	enc.json = *json;
681	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	754	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
682	enc.cur = SvPVX (enc.sv);	755	enc.cur = SvPVX (enc.sv);
683	enc.end = SvEND (enc.sv);	756	enc.end = SvEND (enc.sv);
684	enc.indent = 0;	757	enc.indent = 0;
685	enc.maxdepth = DEC_DEPTH (enc.json.flags);
686	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL	758	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
687	: enc.json.flags & F_LATIN1 ? 0x000100UL	759	: enc.json.flags & F_LATIN1 ? 0x000100UL
688	: 0x10FFFFUL;	760	: 0x110000UL;
689		761
690	SvPOK_only (enc.sv);	762	SvPOK_only (enc.sv);
691	encode_sv (&enc, scalar);	763	encode_sv (&enc, scalar);
692		764
693	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	765	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
…		…
757	if (*dec->cur != ch) \	829	if (*dec->cur != ch) \
758	ERR (# ch " expected"); \	830	ERR (# ch " expected"); \
759	++dec->cur; \	831	++dec->cur; \
760	SE	832	SE
761		833
762	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	834	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
763	#define DEC_DEC_DEPTH --dec->depth	835	#define DEC_DEC_DEPTH --dec->depth
764		836
765	static SV decode_sv (dec_t dec);	837	static SV decode_sv (dec_t dec);
766		838
767	static signed char decode_hexdigit[256];	839	static signed char decode_hexdigit[256];
…		…
859		931
860	if (hi >= 0x80)	932	if (hi >= 0x80)
861	{	933	{
862	utf8 = 1;	934	utf8 = 1;
863		935
864	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	936	cur = encode_utf8 (cur, hi);
865	}	937	}
866	else	938	else
867	*cur++ = hi;	939	*cur++ = hi;
868	}	940	}
869	break;	941	break;
…		…
871	default:	943	default:
872	--dec_cur;	944	--dec_cur;
873	ERR ("illegal backslash escape sequence in string");	945	ERR ("illegal backslash escape sequence in string");
874	}	946	}
875	}	947	}
876	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	948	else if (expect_true (ch >= 0x20 && ch < 0x80))
877	*cur++ = ch;	949	*cur++ = ch;
878	else if (ch >= 0x80)	950	else if (ch >= 0x80)
879	{	951	{
880	STRLEN clen;	952	STRLEN clen;
881	UV uch;	953	UV uch;
…		…
907	{	979	{
908	STRLEN len = cur - buf;	980	STRLEN len = cur - buf;
909		981
910	if (sv)	982	if (sv)
911	{	983	{
912	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
913	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	989	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
914	SvCUR_set (sv, SvCUR (sv) + len);	990	SvCUR_set (sv, SvCUR (sv) + len);
915	}	991	}
916	else	992	else
917	sv = newSVpvn (buf, len);	993	sv = newSVpvn (buf, len);
…		…
1008		1084
1009	// special case the rather common 1..5-digit-int case	1085	// special case the rather common 1..5-digit-int case
1010	if (*start == '-')	1086	if (*start == '-')
1011	switch (len)	1087	switch (len)
1012	{	1088	{
1013	case 2: return newSViv (-( start [1] - '0' * 1));	1089	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1014	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1090	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1015	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));
1016	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));
1017	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));	1093	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1018	}	1094	}
1019	else	1095	else
1020	switch (len)	1096	switch (len)
1021	{	1097	{
1022	case 1: return newSViv ( start [0] - '0' * 1);	1098	case 1: return newSViv ( start [0] - '0' * 1);
1023	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1099	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1024	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);
1025	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);
1026	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);	1102	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1027	}	1103	}
1028		1104
1029	{	1105	{
1030	UV uv;	1106	UV uv;
1031	int numtype = grok_number (start, len, &uv);	1107	int numtype = grok_number (start, len, &uv);
…		…
1139	char *p = dec->cur;	1215	char *p = dec->cur;
1140	char *e = p + 24; // only try up to 24 bytes	1216	char *e = p + 24; // only try up to 24 bytes
1141		1217
1142	for (;;)	1218	for (;;)
1143	{	1219	{
1144	// the >= 0x80 is true on most architectures	1220	// the >= 0x80 is false on most architectures
1145	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')	1221	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
1146	{	1222	{
1147	// slow path, back up and use decode_str	1223	// slow path, back up and use decode_str
1148	SV *key = decode_str (dec);	1224	SV *key = decode_str (dec);
1149	if (!key)	1225	if (!key)
…		…
1280		1356
1281	static SV *	1357	static SV *
1282	decode_sv (dec_t *dec)	1358	decode_sv (dec_t *dec)
1283	{	1359	{
1284	// the beauty of JSON: you need exactly one character lookahead	1360	// the beauty of JSON: you need exactly one character lookahead
1285	// to parse anything.	1361	// to parse everything.
1286	switch (*dec->cur)	1362	switch (*dec->cur)
1287	{	1363	{
1288	case '"': ++dec->cur; return decode_str (dec);	1364	case '"': ++dec->cur; return decode_str (dec);
1289	case '[': ++dec->cur; return decode_av (dec);	1365	case '[': ++dec->cur; return decode_av (dec);
1290	case '{': ++dec->cur; return decode_hv (dec);	1366	case '{': ++dec->cur; return decode_hv (dec);
1291		1367
1292	case '-':	1368	case '-':
1293	case '0': case '1': case '2': case '3': case '4':	1369	case '0': case '1': case '2': case '3': case '4':
1294	case '5': case '6': case '7': case '8': case '9':	1370	case '5': case '6': case '7': case '8': case '9':
1295	return decode_num (dec);	1371	return decode_num (dec);
…		…
1297	case 't':	1373	case 't':
1298	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1374	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1299	{	1375	{
1300	dec->cur += 4;	1376	dec->cur += 4;
1301	#if JSON_SLOW	1377	#if JSON_SLOW
1302	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);	1378	json_true = get_bool ("JSON::XS::true");
1303	#endif	1379	#endif
1304	return SvREFCNT_inc (json_true);	1380	return newSVsv (json_true);
1305	}	1381	}
1306	else	1382	else
1307	ERR ("'true' expected");	1383	ERR ("'true' expected");
1308		1384
1309	break;	1385	break;
…		…
1311	case 'f':	1387	case 'f':
1312	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1388	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1313	{	1389	{
1314	dec->cur += 5;	1390	dec->cur += 5;
1315	#if JSON_SLOW	1391	#if JSON_SLOW
1316	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1392	json_false = get_bool ("JSON::XS::false");
1317	#endif	1393	#endif
1318	return SvREFCNT_inc (json_false);	1394	return newSVsv (json_false);
1319	}	1395	}
1320	else	1396	else
1321	ERR ("'false' expected");	1397	ERR ("'false' expected");
1322		1398
1323	break;	1399	break;
…		…
1341	fail:	1417	fail:
1342	return 0;	1418	return 0;
1343	}	1419	}
1344		1420
1345	static SV *	1421	static SV *
1346	decode_json (SV string, JSON json, UV *offset_return)	1422	decode_json (SV string, JSON json, char **offset_return)
1347	{	1423	{
1348	dec_t dec;	1424	dec_t dec;
1349	UV offset;
1350	SV *sv;	1425	SV *sv;
1351		1426
		1427	/* work around bugs in 5.10 where manipulating magic values
		1428	* will perl ignore the magic in subsequent accesses
		1429	*/
1352	SvGETMAGIC (string);	1430	/SvGETMAGIC (string);/
		1431	if (SvMAGICAL (string))
		1432	string = sv_2mortal (newSVsv (string));
		1433
1353	SvUPGRADE (string, SVt_PV);	1434	SvUPGRADE (string, SVt_PV);
1354		1435
1355	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)
1356	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",
1357	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));	1455	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1456	}
1358		1457
1359	if (json->flags & F_UTF8)	1458	if (json->flags & F_UTF8)
1360	sv_utf8_downgrade (string, 0);	1459	sv_utf8_downgrade (string, 0);
1361	else	1460	else
1362	sv_utf8_upgrade (string);	1461	sv_utf8_upgrade (string);
1363		1462
1364	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1463	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1365		1464
1366	dec.json = *json;	1465	dec.json = *json;
1367	dec.cur = SvPVX (string);	1466	dec.cur = SvPVX (string);
1368	dec.end = SvEND (string);	1467	dec.end = SvEND (string);
1369	dec.err = 0;	1468	dec.err = 0;
1370	dec.depth = 0;	1469	dec.depth = 0;
1371	dec.maxdepth = DEC_DEPTH (dec.json.flags);
1372		1470
1373	if (dec.json.cb_object \|\| dec.json.cb_sk_object)	1471	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
1374	dec.json.flags \|= F_HOOK;	1472	dec.json.flags \|= F_HOOK;
1375		1473
1376	*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
1377		1475
1378	decode_ws (&dec);	1476	decode_ws (&dec);
1379	sv = decode_sv (&dec);	1477	sv = decode_sv (&dec);
		1478
		1479	if (offset_return)
		1480	*offset_return = dec.cur;
1380		1481
1381	if (!(offset_return \|\| !sv))	1482	if (!(offset_return \|\| !sv))
1382	{	1483	{
1383	// check for trailing garbage	1484	// check for trailing garbage
1384	decode_ws (&dec);	1485	decode_ws (&dec);
…		…
1387	{	1488	{
1388	dec.err = "garbage after JSON object";	1489	dec.err = "garbage after JSON object";
1389	SvREFCNT_dec (sv);	1490	SvREFCNT_dec (sv);
1390	sv = 0;	1491	sv = 0;
1391	}	1492	}
1392	}
1393
1394	if (offset_return \|\| !sv)
1395	{
1396	offset = dec.json.flags & F_UTF8
1397	? dec.cur - SvPVX (string)
1398	: utf8_distance (dec.cur, SvPVX (string));
1399
1400	if (offset_return)
1401	*offset_return = offset;
1402	}	1493	}
1403		1494
1404	if (!sv)	1495	if (!sv)
1405	{	1496	{
1406	SV *uni = sv_newmortal ();	1497	SV *uni = sv_newmortal ();
…		…
1412	SAVEVPTR (PL_curcop);	1503	SAVEVPTR (PL_curcop);
1413	PL_curcop = &cop;	1504	PL_curcop = &cop;
1414	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);
1415	LEAVE;	1506	LEAVE;
1416		1507
1417	croak ("%s, at character offset %d [\"%s\"]",	1508	croak ("%s, at character offset %d (before \"%s\")",
1418	dec.err,	1509	dec.err,
1419	(int)offset,	1510	ptr_to_index (string, dec.cur),
1420	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1511	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1421	}	1512	}
1422		1513
1423	sv = sv_2mortal (sv);	1514	sv = sv_2mortal (sv);
1424		1515
1425	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))	1516	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1426	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)");
1427		1518
1428	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
1429	}	1637	}
1430		1638
1431	/////////////////////////////////////////////////////////////////////////////	1639	/////////////////////////////////////////////////////////////////////////////
1432	// XS interface functions	1640	// XS interface functions
1433		1641
…		…
1445	: -1;	1653	: -1;
1446		1654
1447	json_stash = gv_stashpv ("JSON::XS" , 1);	1655	json_stash = gv_stashpv ("JSON::XS" , 1);
1448	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);	1656	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1449		1657
1450	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );	1658	json_true = get_bool ("JSON::XS::true");
1451	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1659	json_false = get_bool ("JSON::XS::false");
1452	}	1660	}
1453		1661
1454	PROTOTYPES: DISABLE	1662	PROTOTYPES: DISABLE
1455		1663
1456	void CLONE (...)	1664	void CLONE (...)
…		…
1461	void new (char *klass)	1669	void new (char *klass)
1462	PPCODE:	1670	PPCODE:
1463	{	1671	{
1464	SV *pv = NEWSV (0, sizeof (JSON));	1672	SV *pv = NEWSV (0, sizeof (JSON));
1465	SvPOK_only (pv);	1673	SvPOK_only (pv);
1466	Zero (SvPVX (pv), 1, JSON);	1674	json_init ((JSON *)SvPVX (pv));
1467	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1468	XPUSHs (sv_2mortal (sv_bless (	1675	XPUSHs (sv_2mortal (sv_bless (
1469	newRV_noinc (pv),	1676	newRV_noinc (pv),
1470	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)	1677	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1471	)));	1678	)));
1472	}	1679	}
…		…
1484	allow_nonref = F_ALLOW_NONREF	1691	allow_nonref = F_ALLOW_NONREF
1485	shrink = F_SHRINK	1692	shrink = F_SHRINK
1486	allow_blessed = F_ALLOW_BLESSED	1693	allow_blessed = F_ALLOW_BLESSED
1487	convert_blessed = F_CONV_BLESSED	1694	convert_blessed = F_CONV_BLESSED
1488	relaxed = F_RELAXED	1695	relaxed = F_RELAXED
		1696	allow_unknown = F_ALLOW_UNKNOWN
1489	PPCODE:	1697	PPCODE:
1490	{	1698	{
1491	if (enable)	1699	if (enable)
1492	self->flags \|= ix;	1700	self->flags \|= ix;
1493	else	1701	else
…		…
1508	get_allow_nonref = F_ALLOW_NONREF	1716	get_allow_nonref = F_ALLOW_NONREF
1509	get_shrink = F_SHRINK	1717	get_shrink = F_SHRINK
1510	get_allow_blessed = F_ALLOW_BLESSED	1718	get_allow_blessed = F_ALLOW_BLESSED
1511	get_convert_blessed = F_CONV_BLESSED	1719	get_convert_blessed = F_CONV_BLESSED
1512	get_relaxed = F_RELAXED	1720	get_relaxed = F_RELAXED
		1721	get_allow_unknown = F_ALLOW_UNKNOWN
1513	PPCODE:	1722	PPCODE:
1514	XPUSHs (boolSV (self->flags & ix));	1723	XPUSHs (boolSV (self->flags & ix));
1515		1724
1516	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1725	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1517	PPCODE:	1726	PPCODE:
1518	{	1727	self->max_depth = max_depth;
1519	UV log2 = 0;
1520
1521	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1522
1523	while ((1UL << log2) < max_depth)
1524	++log2;
1525
1526	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1527
1528	XPUSHs (ST (0));	1728	XPUSHs (ST (0));
1529	}
1530		1729
1531	U32 get_max_depth (JSON *self)	1730	U32 get_max_depth (JSON *self)
1532	CODE:	1731	CODE:
1533	RETVAL = DEC_DEPTH (self->flags);	1732	RETVAL = self->max_depth;
1534	OUTPUT:	1733	OUTPUT:
1535	RETVAL	1734	RETVAL
1536		1735
1537	void max_size (JSON *self, UV max_size = 0)	1736	void max_size (JSON *self, U32 max_size = 0)
1538	PPCODE:	1737	PPCODE:
1539	{	1738	self->max_size = max_size;
1540	UV log2 = 0;
1541
1542	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1543	if (max_size == 1) max_size = 2;
1544
1545	while ((1UL << log2) < max_size)
1546	++log2;
1547
1548	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
1549
1550	XPUSHs (ST (0));	1739	XPUSHs (ST (0));
1551	}
1552		1740
1553	int get_max_size (JSON *self)	1741	int get_max_size (JSON *self)
1554	CODE:	1742	CODE:
1555	RETVAL = DEC_SIZE (self->flags);	1743	RETVAL = self->max_size;
1556	OUTPUT:	1744	OUTPUT:
1557	RETVAL	1745	RETVAL
1558		1746
1559	void filter_json_object (JSON self, SV cb = &PL_sv_undef)	1747	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1560	PPCODE:	1748	PPCODE:
…		…
1596	XPUSHs (decode_json (jsonstr, self, 0));	1784	XPUSHs (decode_json (jsonstr, self, 0));
1597		1785
1598	void decode_prefix (JSON self, SV jsonstr)	1786	void decode_prefix (JSON self, SV jsonstr)
1599	PPCODE:	1787	PPCODE:
1600	{	1788	{
1601	UV offset;	1789	char *offset;
1602	EXTEND (SP, 2);	1790	EXTEND (SP, 2);
1603	PUSHs (decode_json (jsonstr, self, &offset));	1791	PUSHs (decode_json (jsonstr, self, &offset));
1604	PUSHs (sv_2mortal (newSVuv (offset)));	1792	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		1793	}
		1794
		1795	void incr_parse (JSON self, SV jsonstr = 0)
		1796	PPCODE:
		1797	{
		1798	if (!self->incr_text)
		1799	self->incr_text = newSVpvn ("", 0);
		1800
		1801	// append data, if any
		1802	if (jsonstr)
		1803	{
		1804	if (SvUTF8 (jsonstr))
		1805	{
		1806	if (!SvUTF8 (self->incr_text))
		1807	{
		1808	/* utf-8-ness differs, need to upgrade */
		1809	sv_utf8_upgrade (self->incr_text);
		1810
		1811	if (self->incr_pos)
		1812	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1813	- (U8 *)SvPVX (self->incr_text);
		1814	}
		1815	}
		1816	else if (SvUTF8 (self->incr_text))
		1817	sv_utf8_upgrade (jsonstr);
		1818
		1819	{
		1820	STRLEN len;
		1821	const char *str = SvPV (jsonstr, len);
		1822	STRLEN cur = SvCUR (self->incr_text);
		1823
		1824	if (SvLEN (self->incr_text) <= cur + len)
		1825	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1826
		1827	Move (str, SvEND (self->incr_text), len, char);
		1828	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1829	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1830	}
		1831	}
		1832
		1833	if (GIMME_V != G_VOID)
		1834	do
		1835	{
		1836	char *offset;
		1837
		1838	if (!INCR_DONE (self))
		1839	{
		1840	incr_parse (self);
		1841
		1842	if (self->incr_pos > self->max_size && self->max_size)
		1843	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1844	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1845
		1846	if (!INCR_DONE (self))
		1847	break;
		1848	}
		1849
		1850	XPUSHs (decode_json (self->incr_text, self, &offset));
		1851
		1852	self->incr_pos -= offset - SvPVX (self->incr_text);
		1853	self->incr_nest = 0;
		1854	self->incr_mode = 0;
		1855
		1856	sv_chop (self->incr_text, offset);
		1857	}
		1858	while (GIMME_V == G_ARRAY);
		1859	}
		1860
		1861	SV incr_text (JSON self)
		1862	ATTRS: lvalue
		1863	CODE:
		1864	{
		1865	if (self->incr_pos)
		1866	croak ("incr_text can not be called when the incremental parser already started parsing");
		1867
		1868	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		1869	}
		1870	OUTPUT:
		1871	RETVAL
		1872
		1873	void incr_skip (JSON *self)
		1874	CODE:
		1875	{
		1876	if (self->incr_pos)
		1877	{
		1878	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1879	self->incr_pos = 0;
		1880	self->incr_nest = 0;
		1881	self->incr_mode = 0;
		1882	}
		1883	}
		1884
		1885	void incr_reset (JSON *self)
		1886	CODE:
		1887	{
		1888	SvREFCNT_dec (self->incr_text);
		1889	self->incr_text = 0;
		1890	self->incr_pos = 0;
		1891	self->incr_nest = 0;
		1892	self->incr_mode = 0;
1605	}	1893	}
1606		1894
1607	void DESTROY (JSON *self)	1895	void DESTROY (JSON *self)
1608	CODE:	1896	CODE:
1609	SvREFCNT_dec (self->cb_sk_object);	1897	SvREFCNT_dec (self->cb_sk_object);
1610	SvREFCNT_dec (self->cb_object);	1898	SvREFCNT_dec (self->cb_object);
		1899	SvREFCNT_dec (self->incr_text);
1611		1900
1612	PROTOTYPES: ENABLE	1901	PROTOTYPES: ENABLE
1613		1902
1614	void encode_json (SV *scalar)	1903	void encode_json (SV *scalar)
		1904	ALIAS:
		1905	to_json_ = 0
		1906	encode_json = F_UTF8
1615	PPCODE:	1907	PPCODE:
1616	{	1908	{
1617	JSON json = { F_DEFAULT \| F_UTF8 };	1909	JSON json;
		1910	json_init (&json);
		1911	json.flags \|= ix;
1618	XPUSHs (encode_json (scalar, &json));	1912	XPUSHs (encode_json (scalar, &json));
1619	}	1913	}
1620		1914
1621	void decode_json (SV *jsonstr)	1915	void decode_json (SV *jsonstr)
		1916	ALIAS:
		1917	from_json_ = 0
		1918	decode_json = F_UTF8
1622	PPCODE:	1919	PPCODE:
1623	{	1920	{
1624	JSON json = { F_DEFAULT \| F_UTF8 };	1921	JSON json;
		1922	json_init (&json);
		1923	json.flags \|= ix;
1625	XPUSHs (decode_json (jsonstr, &json, 0));	1924	XPUSHs (decode_json (jsonstr, &json, 0));
1626	}	1925	}
1627		1926
		1927

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.71 by root, Wed Mar 19 03:17:38 2008 UTC vs. Revision 1.97 by root, Fri Jul 3 09:38:30 2009 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.71 by root, Wed Mar 19 03:17:38 2008 UTC vs.
Revision 1.97 by root, Fri Jul 3 09:38:30 2009 UTC