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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.74 by root, Wed Mar 19 15:17:53 2008 UTC vs.
Revision 1.109 by root, Tue Mar 8 10:18:50 2011 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
		46
		47	#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
50		48
51	#define SB do {	49	#define SB do {
52	#define SE } while (0)	50	#define SE } while (0)
53		51
54	#if __GNUC__ >= 3	52	#if __GNUC__ >= 3
…		…
64		62
65	#define IN_RANGE_INC(type,val,beg,end) \	63	#define IN_RANGE_INC(type,val,beg,end) \
66	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \	64	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
67	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))	65	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
68		66
		67	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
		68
69	#ifdef USE_ITHREADS	69	#ifdef USE_ITHREADS
70	# define JSON_SLOW 1	70	# define JSON_SLOW 1
71	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))	71	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
72	#else	72	#else
73	# define JSON_SLOW 0	73	# define JSON_SLOW 0
…		…
75	#endif	75	#endif
76		76
77	static HV json_stash, json_boolean_stash; // JSON::XS::	77	static HV json_stash, json_boolean_stash; // JSON::XS::
78	static SV json_true, json_false;	78	static SV json_true, json_false;
79		79
		80	enum {
		81	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		82	INCR_M_STR, // inside string
		83	INCR_M_BS, // inside backslash
		84	INCR_M_C0, // inside comment in initial whitespace sequence
		85	INCR_M_C1, // inside comment in other places
		86	INCR_M_JSON // outside anything, count nesting
		87	};
		88
		89	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		90
80	typedef struct {	91	typedef struct {
81	U32 flags;	92	U32 flags;
		93	U32 max_depth;
		94	STRLEN max_size;
		95
82	SV *cb_object;	96	SV *cb_object;
83	HV *cb_sk_object;	97	HV *cb_sk_object;
		98
		99	// for the incremental parser
		100	SV *incr_text; // the source text so far
		101	STRLEN incr_pos; // the current offset into the text
		102	int incr_nest; // {[]}-nesting level
		103	unsigned char incr_mode;
84	} JSON;	104	} JSON;
		105
		106	INLINE void
		107	json_init (JSON *json)
		108	{
		109	Zero (json, 1, JSON);
		110	json->max_depth = 512;
		111	}
85		112
86	/////////////////////////////////////////////////////////////////////////////	113	/////////////////////////////////////////////////////////////////////////////
87	// utility functions	114	// utility functions
88		115
		116	INLINE SV *
		117	get_bool (const char *name)
		118	{
		119	SV *sv = get_sv (name, 1);
		120
		121	SvREADONLY_on (sv);
		122	SvREADONLY_on (SvRV (sv));
		123
		124	return sv;
		125	}
		126
89	INLINE void	127	INLINE void
90	shrink (SV *sv)	128	shrink (SV *sv)
91	{	129	{
92	sv_utf8_downgrade (sv, 1);	130	sv_utf8_downgrade (sv, 1);
		131
93	if (SvLEN (sv) > SvCUR (sv) + 1)	132	if (SvLEN (sv) > SvCUR (sv) + 1)
94	{	133	{
95	#ifdef SvPV_shrink_to_cur	134	#ifdef SvPV_shrink_to_cur
96	SvPV_shrink_to_cur (sv);	135	SvPV_shrink_to_cur (sv);
97	#elif defined (SvPV_renew)	136	#elif defined (SvPV_renew)
…		…
141	*s++ = 0x80 \| ( ch & 0x3f);	180	*s++ = 0x80 \| ( ch & 0x3f);
142		181
143	return s;	182	return s;
144	}	183	}
145		184
		185	// convert offset pointer to character index, sv must be string
		186	static STRLEN
		187	ptr_to_index (SV sv, char offset)
		188	{
		189	return SvUTF8 (sv)
		190	? utf8_distance (offset, SvPVX (sv))
		191	: offset - SvPVX (sv);
		192	}
		193
		194	/////////////////////////////////////////////////////////////////////////////
		195	// fp hell
		196
		197	// scan a group of digits, and a trailing exponent
		198	static void
		199	json_atof_scan1 (const char s, NV accum, int *expo, int postdp, int maxdepth)
		200	{
		201	UV uaccum = 0;
		202	int eaccum = 0;
		203
		204	// if we recurse too deep, skip all remaining digits
		205	// to avoid a stack overflow attack
		206	if (expect_false (--maxdepth <= 0))
		207	while (((U8)*s - '0') < 10)
		208	++s;
		209
		210	for (;;)
		211	{
		212	U8 dig = (U8)*s - '0';
		213
		214	if (expect_false (dig >= 10))
		215	{
		216	if (dig == (U8)((U8)'.' - (U8)'0'))
		217	{
		218	++s;
		219	json_atof_scan1 (s, accum, expo, 1, maxdepth);
		220	}
		221	else if ((dig \| ' ') == 'e' - '0')
		222	{
		223	int exp2 = 0;
		224	int neg = 0;
		225
		226	++s;
		227
		228	if (*s == '-')
		229	{
		230	++s;
		231	neg = 1;
		232	}
		233	else if (*s == '+')
		234	++s;
		235
		236	while ((dig = (U8)*s - '0') < 10)
		237	exp2 = exp2 * 10 + *s++ - '0';
		238
		239	*expo += neg ? -exp2 : exp2;
		240	}
		241
		242	break;
		243	}
		244
		245	++s;
		246
		247	uaccum = uaccum * 10 + dig;
		248	++eaccum;
		249
		250	// if we have too many digits, then recurse for more
		251	// we actually do this for rather few digits
		252	if (uaccum >= (UV_MAX - 9) / 10)
		253	{
		254	if (postdp) *expo -= eaccum;
		255	json_atof_scan1 (s, accum, expo, postdp, maxdepth);
		256	if (postdp) *expo += eaccum;
		257
		258	break;
		259	}
		260	}
		261
		262	// this relies greatly on the quality of the pow ()
		263	// implementation of the platform, but a good
		264	// implementation is hard to beat.
		265	if (postdp) *expo -= eaccum;
		266	accum += uaccum Perl_pow (10., *expo);
		267	*expo += eaccum;
		268	}
		269
		270	static NV
		271	json_atof (const char *s)
		272	{
		273	NV accum = 0.;
		274	int expo = 0;
		275	int neg = 0;
		276
		277	if (*s == '-')
		278	{
		279	++s;
		280	neg = 1;
		281	}
		282
		283	// a recursion depth of ten gives us >>500 bits
		284	json_atof_scan1 (s, &accum, &expo, 0, 10);
		285
		286	return neg ? -accum : accum;
		287	}
146	/////////////////////////////////////////////////////////////////////////////	288	/////////////////////////////////////////////////////////////////////////////
147	// encoder	289	// encoder
148		290
149	// structure used for encoding JSON	291	// structure used for encoding JSON
150	typedef struct	292	typedef struct
…		…
152	char *cur; // SvPVX (sv) + current output position	294	char *cur; // SvPVX (sv) + current output position
153	char *end; // SvEND (sv)	295	char *end; // SvEND (sv)
154	SV *sv; // result scalar	296	SV *sv; // result scalar
155	JSON json;	297	JSON json;
156	U32 indent; // indentation level	298	U32 indent; // indentation level
157	U32 maxdepth; // max. indentation/recursion level
158	UV limit; // escape character values >= this value when encoding	299	UV limit; // escape character values >= this value when encoding
159	} enc_t;	300	} enc_t;
160		301
161	INLINE void	302	INLINE void
162	need (enc_t *enc, STRLEN len)	303	need (enc_t *enc, STRLEN len)
163	{	304	{
164	if (expect_false (enc->cur + len >= enc->end))	305	if (expect_false (enc->cur + len >= enc->end))
165	{	306	{
166	STRLEN cur = enc->cur - SvPVX (enc->sv);	307	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
167	SvGROW (enc->sv, cur + len + 1);	308	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
168	enc->cur = SvPVX (enc->sv) + cur;	309	enc->cur = SvPVX (enc->sv) + cur;
169	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	310	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
170	}	311	}
171	}	312	}
172		313
…		…
247	(int)((uch - 0x10000) % 0x400 + 0xDC00));	388	(int)((uch - 0x10000) % 0x400 + 0xDC00));
248	enc->cur += 12;	389	enc->cur += 12;
249	}	390	}
250	else	391	else
251	{	392	{
252	static char hexdigit [16] = "0123456789abcdef";
253	need (enc, len += 5);	393	need (enc, len += 5);
254	*enc->cur++ = '\\';	394	*enc->cur++ = '\\';
255	*enc->cur++ = 'u';	395	*enc->cur++ = 'u';
256	*enc->cur++ = hexdigit [ uch >> 12 ];	396	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
257	*enc->cur++ = hexdigit [(uch >> 8) & 15];	397	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
258	*enc->cur++ = hexdigit [(uch >> 4) & 15];	398	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
259	*enc->cur++ = hexdigit [(uch >> 0) & 15];	399	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
260	}	400	}
261		401
262	str += clen;	402	str += clen;
263	}	403	}
264	else if (enc->json.flags & F_LATIN1)	404	else if (enc->json.flags & F_LATIN1)
…		…
335	static void	475	static void
336	encode_av (enc_t enc, AV av)	476	encode_av (enc_t enc, AV av)
337	{	477	{
338	int i, len = av_len (av);	478	int i, len = av_len (av);
339		479
340	if (enc->indent >= enc->maxdepth)	480	if (enc->indent >= enc->json.max_depth)
341	croak ("data structure too deep (hit recursion limit)");	481	croak (ERR_NESTING_EXCEEDED);
342		482
343	encode_ch (enc, '[');	483	encode_ch (enc, '[');
344		484
345	if (len >= 0)	485	if (len >= 0)
346	{	486	{
…		…
420		560
421	static void	561	static void
422	encode_hv (enc_t enc, HV hv)	562	encode_hv (enc_t enc, HV hv)
423	{	563	{
424	HE *he;	564	HE *he;
425	int count;
426		565
427	if (enc->indent >= enc->maxdepth)	566	if (enc->indent >= enc->json.max_depth)
428	croak ("data structure too deep (hit recursion limit)");	567	croak (ERR_NESTING_EXCEEDED);
429		568
430	encode_ch (enc, '{');	569	encode_ch (enc, '{');
431		570
432	// for canonical output we have to sort by keys first	571	// for canonical output we have to sort by keys first
433	// actually, this is mostly due to the stupid so-called	572	// actually, this is mostly due to the stupid so-called
434	// security workaround added somewhere in 5.8.x.	573	// security workaround added somewhere in 5.8.x
435	// that randomises hash orderings	574	// that randomises hash orderings
436	if (enc->json.flags & F_CANONICAL)	575	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
437	{	576	{
438	int count = hv_iterinit (hv);	577	int count = hv_iterinit (hv);
439		578
440	if (SvMAGICAL (hv))	579	if (SvMAGICAL (hv))
441	{	580	{
…		…
614		753
615	if (len == 1 && *pv == '1')	754	if (len == 1 && *pv == '1')
616	encode_str (enc, "true", 4, 0);	755	encode_str (enc, "true", 4, 0);
617	else if (len == 1 && *pv == '0')	756	else if (len == 1 && *pv == '0')
618	encode_str (enc, "false", 5, 0);	757	encode_str (enc, "false", 5, 0);
		758	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		759	encode_str (enc, "null", 4, 0);
619	else	760	else
620	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	761	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
621	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	762	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
622	}	763	}
		764	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		765	encode_str (enc, "null", 4, 0);
623	else	766	else
624	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	767	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
625	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	768	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
626	}	769	}
627		770
…		…
648	else if (SvIOKp (sv))	791	else if (SvIOKp (sv))
649	{	792	{
650	// we assume we can always read an IV as a UV and vice versa	793	// we assume we can always read an IV as a UV and vice versa
651	// we assume two's complement	794	// we assume two's complement
652	// we assume no aliasing issues in the union	795	// we assume no aliasing issues in the union
653	if (SvIsUV (sv) ? SvUVX (sv) > 59000	796	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
654	: SvIVX (sv) > 59000 \|\| SvIVX (sv) < -59000)	797	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
655	{
656	// large integer, use the (rather slow) snprintf way.
657	need (enc, sizeof (UV) * 5 / 2 + 1); // CHAR_BIT is at least 8
658	enc->cur +=
659	SvIsUV(sv)
660	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
661	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
662	}
663	else
664	{	798	{
665	// optimise the "small number case"	799	// optimise the "small number case"
666	// code will likely be branchless and use only a single multiplication	800	// code will likely be branchless and use only a single multiplication
667	// works for numbers up to 59074	801	// works for numbers up to 59074
668	I32 i = SvIVX (sv);	802	I32 i = SvIVX (sv);
…		…
679		813
680	// now output digit by digit, each time masking out the integer part	814	// now output digit by digit, each time masking out the integer part
681	// and multiplying by 5 while moving the decimal point one to the right,	815	// and multiplying by 5 while moving the decimal point one to the right,
682	// resulting in a net multiplication by 10.	816	// resulting in a net multiplication by 10.
683	// we always write the digit to memory but conditionally increment	817	// we always write the digit to memory but conditionally increment
684	// the pointer, to ease the usage of conditional move instructions.	818	// the pointer, to enable the use of conditional move instructions.
685	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffff) 5;	819	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffffUL) 5;
686	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffff) 5;	820	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffffUL) 5;
687	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffff) 5;	821	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffffUL) 5;
688	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffff) 5;	822	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffffUL) 5;
689	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'	823	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
690	}	824	}
		825	else
		826	{
		827	// large integer, use the (rather slow) snprintf way.
		828	need (enc, IVUV_MAXCHARS);
		829	enc->cur +=
		830	SvIsUV(sv)
		831	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
		832	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		833	}
691	}	834	}
692	else if (SvROK (sv))	835	else if (SvROK (sv))
693	encode_rv (enc, SvRV (sv));	836	encode_rv (enc, SvRV (sv));
694	else if (!SvOK (sv))	837	else if (!SvOK (sv) \|\| enc->json.flags & F_ALLOW_UNKNOWN)
695	encode_str (enc, "null", 4, 0);	838	encode_str (enc, "null", 4, 0);
696	else	839	else
697	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	840	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
698	SvPV_nolen (sv), SvFLAGS (sv));	841	SvPV_nolen (sv), SvFLAGS (sv));
699	}	842	}
…		…
709	enc.json = *json;	852	enc.json = *json;
710	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	853	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
711	enc.cur = SvPVX (enc.sv);	854	enc.cur = SvPVX (enc.sv);
712	enc.end = SvEND (enc.sv);	855	enc.end = SvEND (enc.sv);
713	enc.indent = 0;	856	enc.indent = 0;
714	enc.maxdepth = DEC_DEPTH (enc.json.flags);
715	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL	857	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
716	: enc.json.flags & F_LATIN1 ? 0x000100UL	858	: enc.json.flags & F_LATIN1 ? 0x000100UL
717	: 0x110000UL;	859	: 0x110000UL;
718		860
719	SvPOK_only (enc.sv);	861	SvPOK_only (enc.sv);
720	encode_sv (&enc, scalar);	862	encode_sv (&enc, scalar);
		863	encode_nl (&enc);
721		864
722	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	865	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
723	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	866	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
724		867
725	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))	868	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))
…		…
786	if (*dec->cur != ch) \	929	if (*dec->cur != ch) \
787	ERR (# ch " expected"); \	930	ERR (# ch " expected"); \
788	++dec->cur; \	931	++dec->cur; \
789	SE	932	SE
790		933
791	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	934	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
792	#define DEC_DEC_DEPTH --dec->depth	935	#define DEC_DEC_DEPTH --dec->depth
793		936
794	static SV decode_sv (dec_t dec);	937	static SV decode_sv (dec_t dec);
795		938
796	static signed char decode_hexdigit[256];	939	static signed char decode_hexdigit[256];
…		…
905	else if (expect_true (ch >= 0x20 && ch < 0x80))	1048	else if (expect_true (ch >= 0x20 && ch < 0x80))
906	*cur++ = ch;	1049	*cur++ = ch;
907	else if (ch >= 0x80)	1050	else if (ch >= 0x80)
908	{	1051	{
909	STRLEN clen;	1052	STRLEN clen;
910	UV uch;
911		1053
912	--dec_cur;	1054	--dec_cur;
913		1055
914	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);	1056	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
915	if (clen == (STRLEN)-1)	1057	if (clen == (STRLEN)-1)
916	ERR ("malformed UTF-8 character in JSON string");	1058	ERR ("malformed UTF-8 character in JSON string");
917		1059
918	do	1060	do
919	cur++ = dec_cur++;	1061	cur++ = dec_cur++;
…		…
936	{	1078	{
937	STRLEN len = cur - buf;	1079	STRLEN len = cur - buf;
938		1080
939	if (sv)	1081	if (sv)
940	{	1082	{
941	SvGROW (sv, SvCUR (sv) + len + 1);	1083	STRLEN cur = SvCUR (sv);
		1084
		1085	if (SvLEN (sv) <= cur + len)
		1086	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1087
942	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	1088	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
943	SvCUR_set (sv, SvCUR (sv) + len);	1089	SvCUR_set (sv, SvCUR (sv) + len);
944	}	1090	}
945	else	1091	else
946	sv = newSVpvn (buf, len);	1092	sv = newSVpvn (buf, len);
…		…
1037		1183
1038	// special case the rather common 1..5-digit-int case	1184	// special case the rather common 1..5-digit-int case
1039	if (*start == '-')	1185	if (*start == '-')
1040	switch (len)	1186	switch (len)
1041	{	1187	{
1042	case 2: return newSViv (-( start [1] - '0' * 1));	1188	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1043	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1189	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1044	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));	1190	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1045	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));	1191	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
1046	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));	1192	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1047	}	1193	}
1048	else	1194	else
1049	switch (len)	1195	switch (len)
1050	{	1196	{
1051	case 1: return newSViv ( start [0] - '0' * 1);	1197	case 1: return newSViv ( start [0] - '0' * 1);
1052	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1198	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1053	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);	1199	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1054	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);	1200	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1055	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);	1201	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1056	}	1202	}
1057		1203
1058	{	1204	{
1059	UV uv;	1205	UV uv;
1060	int numtype = grok_number (start, len, &uv);	1206	int numtype = grok_number (start, len, &uv);
…		…
1069	}	1215	}
1070		1216
1071	len -= *start == '-' ? 1 : 0;	1217	len -= *start == '-' ? 1 : 0;
1072		1218
1073	// does not fit into IV or UV, try NV	1219	// does not fit into IV or UV, try NV
1074	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)	1220	if (len <= NV_DIG)
1075	#if defined (LDBL_DIG)
1076	\|\| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1077	#endif
1078	)
1079	// fits into NV without loss of precision	1221	// fits into NV without loss of precision
1080	return newSVnv (Atof (start));	1222	return newSVnv (json_atof (start));
1081		1223
1082	// everything else fails, convert it to a string	1224	// everything else fails, convert it to a string
1083	return newSVpvn (start, dec->cur - start);	1225	return newSVpvn (start, dec->cur - start);
1084	}	1226	}
1085		1227
1086	// loss of precision here	1228	// loss of precision here
1087	return newSVnv (Atof (start));	1229	return newSVnv (json_atof (start));
1088		1230
1089	fail:	1231	fail:
1090	return 0;	1232	return 0;
1091	}	1233	}
1092		1234
…		…
1168	char *p = dec->cur;	1310	char *p = dec->cur;
1169	char *e = p + 24; // only try up to 24 bytes	1311	char *e = p + 24; // only try up to 24 bytes
1170		1312
1171	for (;;)	1313	for (;;)
1172	{	1314	{
1173	// the >= 0x80 is true on most architectures	1315	// the >= 0x80 is false on most architectures
1174	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')	1316	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
1175	{	1317	{
1176	// slow path, back up and use decode_str	1318	// slow path, back up and use decode_str
1177	SV *key = decode_str (dec);	1319	SV *key = decode_str (dec);
1178	if (!key)	1320	if (!key)
…		…
1261	dSP;	1403	dSP;
1262	int count;	1404	int count;
1263		1405
1264	ENTER; SAVETMPS; PUSHMARK (SP);	1406	ENTER; SAVETMPS; PUSHMARK (SP);
1265	XPUSHs (HeVAL (he));	1407	XPUSHs (HeVAL (he));
		1408	sv_2mortal (sv);
1266		1409
1267	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;	1410	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1268		1411
1269	if (count == 1)	1412	if (count == 1)
1270	{	1413	{
1271	sv = newSVsv (POPs);	1414	sv = newSVsv (POPs);
1272	FREETMPS; LEAVE;	1415	FREETMPS; LEAVE;
1273	return sv;	1416	return sv;
1274	}	1417	}
1275		1418
		1419	SvREFCNT_inc (sv);
1276	FREETMPS; LEAVE;	1420	FREETMPS; LEAVE;
1277	}	1421	}
1278	}	1422	}
1279		1423
1280	if (dec->json.cb_object)	1424	if (dec->json.cb_object)
…		…
1326	case 't':	1470	case 't':
1327	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1471	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1328	{	1472	{
1329	dec->cur += 4;	1473	dec->cur += 4;
1330	#if JSON_SLOW	1474	#if JSON_SLOW
1331	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);	1475	json_true = get_bool ("JSON::XS::true");
1332	#endif	1476	#endif
1333	return SvREFCNT_inc (json_true);	1477	return newSVsv (json_true);
1334	}	1478	}
1335	else	1479	else
1336	ERR ("'true' expected");	1480	ERR ("'true' expected");
1337		1481
1338	break;	1482	break;
…		…
1340	case 'f':	1484	case 'f':
1341	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1485	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1342	{	1486	{
1343	dec->cur += 5;	1487	dec->cur += 5;
1344	#if JSON_SLOW	1488	#if JSON_SLOW
1345	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1489	json_false = get_bool ("JSON::XS::false");
1346	#endif	1490	#endif
1347	return SvREFCNT_inc (json_false);	1491	return newSVsv (json_false);
1348	}	1492	}
1349	else	1493	else
1350	ERR ("'false' expected");	1494	ERR ("'false' expected");
1351		1495
1352	break;	1496	break;
…		…
1370	fail:	1514	fail:
1371	return 0;	1515	return 0;
1372	}	1516	}
1373		1517
1374	static SV *	1518	static SV *
1375	decode_json (SV string, JSON json, UV *offset_return)	1519	decode_json (SV string, JSON json, char **offset_return)
1376	{	1520	{
1377	dec_t dec;	1521	dec_t dec;
1378	UV offset;
1379	SV *sv;	1522	SV *sv;
1380		1523
		1524	/* work around bugs in 5.10 where manipulating magic values
		1525	* will perl ignore the magic in subsequent accesses.
		1526	* also make a copy of non-PV values, to get them into a clean
		1527	* state (SvPV should do that, but it's buggy, see below).
		1528	*/
1381	SvGETMAGIC (string);	1529	/SvGETMAGIC (string);/
		1530	if (SvMAGICAL (string) \|\| !SvPOK (string))
		1531	string = sv_2mortal (newSVsv (string));
		1532
1382	SvUPGRADE (string, SVt_PV);	1533	SvUPGRADE (string, SVt_PV);
1383		1534
1384	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))	1535	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1536	* assertion with -DDEBUGGING, although SvCUR is documented to
		1537	* return the xpv_cur field which certainly exists after upgrading.
		1538	* according to nicholas clark, calling SvPOK fixes this.
		1539	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1540	* and hope for the best.
		1541	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1542	* assertion business is seriously broken, try yet another workaround
		1543	* for the broken -DDEBUGGING.
		1544	*/
		1545	{
		1546	#ifdef DEBUGGING
		1547	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1548	#else
		1549	STRLEN offset = SvCUR (string);
		1550	#endif
		1551
		1552	if (offset > json->max_size && json->max_size)
1385	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",	1553	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1386	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));	1554	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1555	}
1387		1556
1388	if (json->flags & F_UTF8)	1557	if (DECODE_WANTS_OCTETS (json))
1389	sv_utf8_downgrade (string, 0);	1558	sv_utf8_downgrade (string, 0);
1390	else	1559	else
1391	sv_utf8_upgrade (string);	1560	sv_utf8_upgrade (string);
1392		1561
1393	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1562	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1394		1563
1395	dec.json = *json;	1564	dec.json = *json;
1396	dec.cur = SvPVX (string);	1565	dec.cur = SvPVX (string);
1397	dec.end = SvEND (string);	1566	dec.end = SvEND (string);
1398	dec.err = 0;	1567	dec.err = 0;
1399	dec.depth = 0;	1568	dec.depth = 0;
1400	dec.maxdepth = DEC_DEPTH (dec.json.flags);
1401		1569
1402	if (dec.json.cb_object \|\| dec.json.cb_sk_object)	1570	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
1403	dec.json.flags \|= F_HOOK;	1571	dec.json.flags \|= F_HOOK;
1404		1572
1405	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1573	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
1406		1574
1407	decode_ws (&dec);	1575	decode_ws (&dec);
1408	sv = decode_sv (&dec);	1576	sv = decode_sv (&dec);
		1577
		1578	if (offset_return)
		1579	*offset_return = dec.cur;
1409		1580
1410	if (!(offset_return \|\| !sv))	1581	if (!(offset_return \|\| !sv))
1411	{	1582	{
1412	// check for trailing garbage	1583	// check for trailing garbage
1413	decode_ws (&dec);	1584	decode_ws (&dec);
…		…
1416	{	1587	{
1417	dec.err = "garbage after JSON object";	1588	dec.err = "garbage after JSON object";
1418	SvREFCNT_dec (sv);	1589	SvREFCNT_dec (sv);
1419	sv = 0;	1590	sv = 0;
1420	}	1591	}
1421	}
1422
1423	if (offset_return \|\| !sv)
1424	{
1425	offset = dec.json.flags & F_UTF8
1426	? dec.cur - SvPVX (string)
1427	: utf8_distance (dec.cur, SvPVX (string));
1428
1429	if (offset_return)
1430	*offset_return = offset;
1431	}	1592	}
1432		1593
1433	if (!sv)	1594	if (!sv)
1434	{	1595	{
1435	SV *uni = sv_newmortal ();	1596	SV *uni = sv_newmortal ();
…		…
1441	SAVEVPTR (PL_curcop);	1602	SAVEVPTR (PL_curcop);
1442	PL_curcop = &cop;	1603	PL_curcop = &cop;
1443	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1604	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1444	LEAVE;	1605	LEAVE;
1445		1606
1446	croak ("%s, at character offset %d [\"%s\"]",	1607	croak ("%s, at character offset %d (before \"%s\")",
1447	dec.err,	1608	dec.err,
1448	(int)offset,	1609	ptr_to_index (string, dec.cur),
1449	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1610	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1450	}	1611	}
1451		1612
1452	sv = sv_2mortal (sv);	1613	sv = sv_2mortal (sv);
1453		1614
1454	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))	1615	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1455	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1616	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1456		1617
1457	return sv;	1618	return sv;
		1619	}
		1620
		1621	/////////////////////////////////////////////////////////////////////////////
		1622	// incremental parser
		1623
		1624	static void
		1625	incr_parse (JSON *self)
		1626	{
		1627	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1628
		1629	// the state machine here is a bit convoluted and could be simplified a lot
		1630	// but this would make it slower, so...
		1631
		1632	for (;;)
		1633	{
		1634	//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
		1635	switch (self->incr_mode)
		1636	{
		1637	// only used for initial whitespace skipping
		1638	case INCR_M_WS:
		1639	for (;;)
		1640	{
		1641	if (*p > 0x20)
		1642	{
		1643	if (*p == '#')
		1644	{
		1645	self->incr_mode = INCR_M_C0;
		1646	goto incr_m_c;
		1647	}
		1648	else
		1649	{
		1650	self->incr_mode = INCR_M_JSON;
		1651	goto incr_m_json;
		1652	}
		1653	}
		1654	else if (!*p)
		1655	goto interrupt;
		1656
		1657	++p;
		1658	}
		1659
		1660	// skip a single char inside a string (for \\-processing)
		1661	case INCR_M_BS:
		1662	if (!*p)
		1663	goto interrupt;
		1664
		1665	++p;
		1666	self->incr_mode = INCR_M_STR;
		1667	goto incr_m_str;
		1668
		1669	// inside #-style comments
		1670	case INCR_M_C0:
		1671	case INCR_M_C1:
		1672	incr_m_c:
		1673	for (;;)
		1674	{
		1675	if (*p == '\n')
		1676	{
		1677	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1678	break;
		1679	}
		1680	else if (!*p)
		1681	goto interrupt;
		1682
		1683	++p;
		1684	}
		1685
		1686	break;
		1687
		1688	// inside a string
		1689	case INCR_M_STR:
		1690	incr_m_str:
		1691	for (;;)
		1692	{
		1693	if (*p == '"')
		1694	{
		1695	++p;
		1696	self->incr_mode = INCR_M_JSON;
		1697
		1698	if (!self->incr_nest)
		1699	goto interrupt;
		1700
		1701	goto incr_m_json;
		1702	}
		1703	else if (*p == '\\')
		1704	{
		1705	++p; // "virtually" consumes character after \
		1706
		1707	if (!*p) // if at end of string we have to switch modes
		1708	{
		1709	self->incr_mode = INCR_M_BS;
		1710	goto interrupt;
		1711	}
		1712	}
		1713	else if (!*p)
		1714	goto interrupt;
		1715
		1716	++p;
		1717	}
		1718
		1719	// after initial ws, outside string
		1720	case INCR_M_JSON:
		1721	incr_m_json:
		1722	for (;;)
		1723	{
		1724	switch (*p++)
		1725	{
		1726	case 0:
		1727	--p;
		1728	goto interrupt;
		1729
		1730	case 0x09:
		1731	case 0x0a:
		1732	case 0x0d:
		1733	case 0x20:
		1734	if (!self->incr_nest)
		1735	{
		1736	--p; // do not eat the whitespace, let the next round do it
		1737	goto interrupt;
		1738	}
		1739	break;
		1740
		1741	case '"':
		1742	self->incr_mode = INCR_M_STR;
		1743	goto incr_m_str;
		1744
		1745	case '[':
		1746	case '{':
		1747	if (++self->incr_nest > self->max_depth)
		1748	croak (ERR_NESTING_EXCEEDED);
		1749	break;
		1750
		1751	case ']':
		1752	case '}':
		1753	if (--self->incr_nest <= 0)
		1754	goto interrupt;
		1755	break;
		1756
		1757	case '#':
		1758	self->incr_mode = INCR_M_C1;
		1759	goto incr_m_c;
		1760	}
		1761	}
		1762	}
		1763
		1764	modechange:
		1765	;
		1766	}
		1767
		1768	interrupt:
		1769	self->incr_pos = p - SvPVX (self->incr_text);
		1770	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1771	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1458	}	1772	}
1459		1773
1460	/////////////////////////////////////////////////////////////////////////////	1774	/////////////////////////////////////////////////////////////////////////////
1461	// XS interface functions	1775	// XS interface functions
1462		1776
…		…
1474	: -1;	1788	: -1;
1475		1789
1476	json_stash = gv_stashpv ("JSON::XS" , 1);	1790	json_stash = gv_stashpv ("JSON::XS" , 1);
1477	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);	1791	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1478		1792
1479	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );	1793	json_true = get_bool ("JSON::XS::true");
1480	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1794	json_false = get_bool ("JSON::XS::false");
		1795
		1796	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1481	}	1797	}
1482		1798
1483	PROTOTYPES: DISABLE	1799	PROTOTYPES: DISABLE
1484		1800
1485	void CLONE (...)	1801	void CLONE (...)
…		…
1490	void new (char *klass)	1806	void new (char *klass)
1491	PPCODE:	1807	PPCODE:
1492	{	1808	{
1493	SV *pv = NEWSV (0, sizeof (JSON));	1809	SV *pv = NEWSV (0, sizeof (JSON));
1494	SvPOK_only (pv);	1810	SvPOK_only (pv);
1495	Zero (SvPVX (pv), 1, JSON);	1811	json_init ((JSON *)SvPVX (pv));
1496	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1497	XPUSHs (sv_2mortal (sv_bless (	1812	XPUSHs (sv_2mortal (sv_bless (
1498	newRV_noinc (pv),	1813	newRV_noinc (pv),
1499	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)	1814	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1500	)));	1815	)));
1501	}	1816	}
…		…
1513	allow_nonref = F_ALLOW_NONREF	1828	allow_nonref = F_ALLOW_NONREF
1514	shrink = F_SHRINK	1829	shrink = F_SHRINK
1515	allow_blessed = F_ALLOW_BLESSED	1830	allow_blessed = F_ALLOW_BLESSED
1516	convert_blessed = F_CONV_BLESSED	1831	convert_blessed = F_CONV_BLESSED
1517	relaxed = F_RELAXED	1832	relaxed = F_RELAXED
		1833	allow_unknown = F_ALLOW_UNKNOWN
1518	PPCODE:	1834	PPCODE:
1519	{	1835	{
1520	if (enable)	1836	if (enable)
1521	self->flags \|= ix;	1837	self->flags \|= ix;
1522	else	1838	else
…		…
1537	get_allow_nonref = F_ALLOW_NONREF	1853	get_allow_nonref = F_ALLOW_NONREF
1538	get_shrink = F_SHRINK	1854	get_shrink = F_SHRINK
1539	get_allow_blessed = F_ALLOW_BLESSED	1855	get_allow_blessed = F_ALLOW_BLESSED
1540	get_convert_blessed = F_CONV_BLESSED	1856	get_convert_blessed = F_CONV_BLESSED
1541	get_relaxed = F_RELAXED	1857	get_relaxed = F_RELAXED
		1858	get_allow_unknown = F_ALLOW_UNKNOWN
1542	PPCODE:	1859	PPCODE:
1543	XPUSHs (boolSV (self->flags & ix));	1860	XPUSHs (boolSV (self->flags & ix));
1544		1861
1545	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1862	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1546	PPCODE:	1863	PPCODE:
1547	{	1864	self->max_depth = max_depth;
1548	UV log2 = 0;
1549
1550	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1551
1552	while ((1UL << log2) < max_depth)
1553	++log2;
1554
1555	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1556
1557	XPUSHs (ST (0));	1865	XPUSHs (ST (0));
1558	}
1559		1866
1560	U32 get_max_depth (JSON *self)	1867	U32 get_max_depth (JSON *self)
1561	CODE:	1868	CODE:
1562	RETVAL = DEC_DEPTH (self->flags);	1869	RETVAL = self->max_depth;
1563	OUTPUT:	1870	OUTPUT:
1564	RETVAL	1871	RETVAL
1565		1872
1566	void max_size (JSON *self, UV max_size = 0)	1873	void max_size (JSON *self, U32 max_size = 0)
1567	PPCODE:	1874	PPCODE:
1568	{	1875	self->max_size = max_size;
1569	UV log2 = 0;
1570
1571	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1572	if (max_size == 1) max_size = 2;
1573
1574	while ((1UL << log2) < max_size)
1575	++log2;
1576
1577	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
1578
1579	XPUSHs (ST (0));	1876	XPUSHs (ST (0));
1580	}
1581		1877
1582	int get_max_size (JSON *self)	1878	int get_max_size (JSON *self)
1583	CODE:	1879	CODE:
1584	RETVAL = DEC_SIZE (self->flags);	1880	RETVAL = self->max_size;
1585	OUTPUT:	1881	OUTPUT:
1586	RETVAL	1882	RETVAL
1587		1883
1588	void filter_json_object (JSON self, SV cb = &PL_sv_undef)	1884	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1589	PPCODE:	1885	PPCODE:
…		…
1625	XPUSHs (decode_json (jsonstr, self, 0));	1921	XPUSHs (decode_json (jsonstr, self, 0));
1626		1922
1627	void decode_prefix (JSON self, SV jsonstr)	1923	void decode_prefix (JSON self, SV jsonstr)
1628	PPCODE:	1924	PPCODE:
1629	{	1925	{
1630	UV offset;	1926	char *offset;
1631	EXTEND (SP, 2);	1927	EXTEND (SP, 2);
1632	PUSHs (decode_json (jsonstr, self, &offset));	1928	PUSHs (decode_json (jsonstr, self, &offset));
1633	PUSHs (sv_2mortal (newSVuv (offset)));	1929	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		1930	}
		1931
		1932	void incr_parse (JSON self, SV jsonstr = 0)
		1933	PPCODE:
		1934	{
		1935	if (!self->incr_text)
		1936	self->incr_text = newSVpvn ("", 0);
		1937
		1938	/* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
		1939	if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
		1940	if (DECODE_WANTS_OCTETS (self))
		1941	{
		1942	if (self->incr_pos)
		1943	self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
		1944	(U8 *)SvPVX (self->incr_text) + self->incr_pos);
		1945
		1946	sv_utf8_downgrade (self->incr_text, 0);
		1947	}
		1948	else
		1949	{
		1950	sv_utf8_upgrade (self->incr_text);
		1951
		1952	if (self->incr_pos)
		1953	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1954	- (U8 *)SvPVX (self->incr_text);
		1955	}
		1956
		1957	// append data, if any
		1958	if (jsonstr)
		1959	{
		1960	/* make sure both strings have same encoding */
		1961	if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
		1962	if (SvUTF8 (jsonstr))
		1963	sv_utf8_downgrade (jsonstr, 0);
		1964	else
		1965	sv_utf8_upgrade (jsonstr);
		1966
		1967	/* and then just blindly append */
		1968	{
		1969	STRLEN len;
		1970	const char *str = SvPV (jsonstr, len);
		1971	STRLEN cur = SvCUR (self->incr_text);
		1972
		1973	if (SvLEN (self->incr_text) <= cur + len)
		1974	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1975
		1976	Move (str, SvEND (self->incr_text), len, char);
		1977	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1978	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1979	}
		1980	}
		1981
		1982	if (GIMME_V != G_VOID)
		1983	do
		1984	{
		1985	char *offset;
		1986
		1987	if (!INCR_DONE (self))
		1988	{
		1989	incr_parse (self);
		1990
		1991	if (self->incr_pos > self->max_size && self->max_size)
		1992	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1993	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1994
		1995	if (!INCR_DONE (self))
		1996	break;
		1997	}
		1998
		1999	XPUSHs (decode_json (self->incr_text, self, &offset));
		2000
		2001	self->incr_pos -= offset - SvPVX (self->incr_text);
		2002	self->incr_nest = 0;
		2003	self->incr_mode = 0;
		2004
		2005	sv_chop (self->incr_text, offset);
		2006	}
		2007	while (GIMME_V == G_ARRAY);
		2008	}
		2009
		2010	SV incr_text (JSON self)
		2011	ATTRS: lvalue
		2012	CODE:
		2013	{
		2014	if (self->incr_pos)
		2015	croak ("incr_text can not be called when the incremental parser already started parsing");
		2016
		2017	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		2018	}
		2019	OUTPUT:
		2020	RETVAL
		2021
		2022	void incr_skip (JSON *self)
		2023	CODE:
		2024	{
		2025	if (self->incr_pos)
		2026	{
		2027	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2028	self->incr_pos = 0;
		2029	self->incr_nest = 0;
		2030	self->incr_mode = 0;
		2031	}
		2032	}
		2033
		2034	void incr_reset (JSON *self)
		2035	CODE:
		2036	{
		2037	SvREFCNT_dec (self->incr_text);
		2038	self->incr_text = 0;
		2039	self->incr_pos = 0;
		2040	self->incr_nest = 0;
		2041	self->incr_mode = 0;
1634	}	2042	}
1635		2043
1636	void DESTROY (JSON *self)	2044	void DESTROY (JSON *self)
1637	CODE:	2045	CODE:
1638	SvREFCNT_dec (self->cb_sk_object);	2046	SvREFCNT_dec (self->cb_sk_object);
1639	SvREFCNT_dec (self->cb_object);	2047	SvREFCNT_dec (self->cb_object);
		2048	SvREFCNT_dec (self->incr_text);
1640		2049
1641	PROTOTYPES: ENABLE	2050	PROTOTYPES: ENABLE
1642		2051
1643	void encode_json (SV *scalar)	2052	void encode_json (SV *scalar)
		2053	ALIAS:
		2054	to_json_ = 0
		2055	encode_json = F_UTF8
1644	PPCODE:	2056	PPCODE:
1645	{	2057	{
1646	JSON json = { F_DEFAULT \| F_UTF8 };	2058	JSON json;
		2059	json_init (&json);
		2060	json.flags \|= ix;
1647	XPUSHs (encode_json (scalar, &json));	2061	XPUSHs (encode_json (scalar, &json));
1648	}	2062	}
1649		2063
1650	void decode_json (SV *jsonstr)	2064	void decode_json (SV *jsonstr)
		2065	ALIAS:
		2066	from_json_ = 0
		2067	decode_json = F_UTF8
1651	PPCODE:	2068	PPCODE:
1652	{	2069	{
1653	JSON json = { F_DEFAULT \| F_UTF8 };	2070	JSON json;
		2071	json_init (&json);
		2072	json.flags \|= ix;
1654	XPUSHs (decode_json (jsonstr, &json, 0));	2073	XPUSHs (decode_json (jsonstr, &json, 0));
1655	}	2074	}
1656		2075

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.74 by root, Wed Mar 19 15:17:53 2008 UTC vs. Revision 1.109 by root, Tue Mar 8 10:18:50 2011 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.74 by root, Wed Mar 19 15:17:53 2008 UTC vs.
Revision 1.109 by root, Tue Mar 8 10:18:50 2011 UTC