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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.72 by root, Wed Mar 19 04:08:22 2008 UTC vs.
Revision 1.92 by root, Mon Sep 22 07:29:29 2008 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
…		…
64		59
65	#define IN_RANGE_INC(type,val,beg,end) \	60	#define IN_RANGE_INC(type,val,beg,end) \
66	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \	61	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
67	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))	62	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
68		63
		64	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
		65
69	#ifdef USE_ITHREADS	66	#ifdef USE_ITHREADS
70	# define JSON_SLOW 1	67	# define JSON_SLOW 1
71	# 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))
72	#else	69	#else
73	# define JSON_SLOW 0	70	# define JSON_SLOW 0
…		…
75	#endif	72	#endif
76		73
77	static HV json_stash, json_boolean_stash; // JSON::XS::	74	static HV json_stash, json_boolean_stash; // JSON::XS::
78	static SV json_true, json_false;	75	static SV json_true, json_false;
79		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
80	typedef struct {	86	typedef struct {
81	U32 flags;	87	U32 flags;
		88	U32 max_depth;
		89	STRLEN max_size;
		90
82	SV *cb_object;	91	SV *cb_object;
83	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;
84	} JSON;	99	} JSON;
		100
		101	INLINE void
		102	json_init (JSON *json)
		103	{
		104	Zero (json, 1, JSON);
		105	json->max_depth = 512;
		106	}
85		107
86	/////////////////////////////////////////////////////////////////////////////	108	/////////////////////////////////////////////////////////////////////////////
87	// utility functions	109	// utility functions
88		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
89	INLINE void	122	INLINE void
90	shrink (SV *sv)	123	shrink (SV *sv)
91	{	124	{
92	sv_utf8_downgrade (sv, 1);	125	sv_utf8_downgrade (sv, 1);
		126
93	if (SvLEN (sv) > SvCUR (sv) + 1)	127	if (SvLEN (sv) > SvCUR (sv) + 1)
94	{	128	{
95	#ifdef SvPV_shrink_to_cur	129	#ifdef SvPV_shrink_to_cur
96	SvPV_shrink_to_cur (sv);	130	SvPV_shrink_to_cur (sv);
97	#elif defined (SvPV_renew)	131	#elif defined (SvPV_renew)
…		…
123	// this function takes advantage of this fact, although current gccs	157	// this function takes advantage of this fact, although current gccs
124	// seem to optimise the check for >= 0x80 away anyways	158	// seem to optimise the check for >= 0x80 away anyways
125	INLINE unsigned char *	159	INLINE unsigned char *
126	encode_utf8 (unsigned char *s, UV ch)	160	encode_utf8 (unsigned char *s, UV ch)
127	{	161	{
128	if (ch <= 0x7FF)	162	if (expect_false (ch < 0x000080))
129	{	163	*s++ = ch;
130	*s++ = (ch >> 6) \| 0xc0;	164	else if (expect_true (ch < 0x000800))
131	*s++ = (ch & 0x3f) \| 0x80;	165	*s++ = 0xc0 \| ( ch >> 6),
132	}	166	*s++ = 0x80 \| ( ch & 0x3f);
133	else	167	else if ( ch < 0x010000)
134	s = uvuni_to_utf8_flags (s, ch, 0);	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);
135		176
136	return s;	177	return s;
137	}	178	}
138		179
139	/////////////////////////////////////////////////////////////////////////////	180	/////////////////////////////////////////////////////////////////////////////
…		…
145	char *cur; // SvPVX (sv) + current output position	186	char *cur; // SvPVX (sv) + current output position
146	char *end; // SvEND (sv)	187	char *end; // SvEND (sv)
147	SV *sv; // result scalar	188	SV *sv; // result scalar
148	JSON json;	189	JSON json;
149	U32 indent; // indentation level	190	U32 indent; // indentation level
150	U32 maxdepth; // max. indentation/recursion level
151	UV limit; // escape character values >= this value when encoding	191	UV limit; // escape character values >= this value when encoding
152	} enc_t;	192	} enc_t;
153		193
154	INLINE void	194	INLINE void
155	need (enc_t *enc, STRLEN len)	195	need (enc_t *enc, STRLEN len)
…		…
227	clen = 1;	267	clen = 1;
228	}	268	}
229		269
230	if (uch < 0x80/0x20/ \|\| uch >= enc->limit)	270	if (uch < 0x80/0x20/ \|\| uch >= enc->limit)
231	{	271	{
232	if (uch > 0xFFFFUL)	272	if (uch >= 0x10000UL)
233	{	273	{
234	if (uch > 0x10FFFFUL)	274	if (uch >= 0x110000UL)
235	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);	275	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
236		276
237	need (enc, len += 11);	277	need (enc, len += 11);
238	sprintf (enc->cur, "\\u%04x\\u%04x",	278	sprintf (enc->cur, "\\u%04x\\u%04x",
239	(int)((uch - 0x10000) / 0x400 + 0xD800),	279	(int)((uch - 0x10000) / 0x400 + 0xD800),
240	(int)((uch - 0x10000) % 0x400 + 0xDC00));	280	(int)((uch - 0x10000) % 0x400 + 0xDC00));
241	enc->cur += 12;	281	enc->cur += 12;
242	}	282	}
243	else	283	else
244	{	284	{
245	static char hexdigit [16] = "0123456789abcdef";
246	need (enc, len += 5);	285	need (enc, len += 5);
247	*enc->cur++ = '\\';	286	*enc->cur++ = '\\';
248	*enc->cur++ = 'u';	287	*enc->cur++ = 'u';
249	*enc->cur++ = hexdigit [ uch >> 12 ];	288	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
250	*enc->cur++ = hexdigit [(uch >> 8) & 15];	289	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
251	*enc->cur++ = hexdigit [(uch >> 4) & 15];	290	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
252	*enc->cur++ = hexdigit [(uch >> 0) & 15];	291	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
253	}	292	}
254		293
255	str += clen;	294	str += clen;
256	}	295	}
257	else if (enc->json.flags & F_LATIN1)	296	else if (enc->json.flags & F_LATIN1)
…		…
328	static void	367	static void
329	encode_av (enc_t enc, AV av)	368	encode_av (enc_t enc, AV av)
330	{	369	{
331	int i, len = av_len (av);	370	int i, len = av_len (av);
332		371
333	if (enc->indent >= enc->maxdepth)	372	if (enc->indent >= enc->json.max_depth)
334	croak ("data structure too deep (hit recursion limit)");	373	croak (ERR_NESTING_EXCEEDED);
335		374
336	encode_ch (enc, '[');	375	encode_ch (enc, '[');
337		376
338	if (len >= 0)	377	if (len >= 0)
339	{	378	{
…		…
413		452
414	static void	453	static void
415	encode_hv (enc_t enc, HV hv)	454	encode_hv (enc_t enc, HV hv)
416	{	455	{
417	HE *he;	456	HE *he;
418	int count;
419		457
420	if (enc->indent >= enc->maxdepth)	458	if (enc->indent >= enc->json.max_depth)
421	croak ("data structure too deep (hit recursion limit)");	459	croak (ERR_NESTING_EXCEEDED);
422		460
423	encode_ch (enc, '{');	461	encode_ch (enc, '{');
424		462
425	// for canonical output we have to sort by keys first	463	// for canonical output we have to sort by keys first
426	// actually, this is mostly due to the stupid so-called	464	// actually, this is mostly due to the stupid so-called
427	// security workaround added somewhere in 5.8.x.	465	// security workaround added somewhere in 5.8.x
428	// that randomises hash orderings	466	// that randomises hash orderings
429	if (enc->json.flags & F_CANONICAL)	467	if (enc->json.flags & F_CANONICAL)
430	{	468	{
431	int count = hv_iterinit (hv);	469	int count = hv_iterinit (hv);
432		470
…		…
607		645
608	if (len == 1 && *pv == '1')	646	if (len == 1 && *pv == '1')
609	encode_str (enc, "true", 4, 0);	647	encode_str (enc, "true", 4, 0);
610	else if (len == 1 && *pv == '0')	648	else if (len == 1 && *pv == '0')
611	encode_str (enc, "false", 5, 0);	649	encode_str (enc, "false", 5, 0);
		650	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		651	encode_str (enc, "null", 4, 0);
612	else	652	else
613	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	653	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
614	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	654	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
615	}	655	}
		656	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		657	encode_str (enc, "null", 4, 0);
616	else	658	else
617	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	659	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
618	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	660	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
619	}	661	}
620		662
…		…
638	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	680	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
639	enc->cur += strlen (enc->cur);	681	enc->cur += strlen (enc->cur);
640	}	682	}
641	else if (SvIOKp (sv))	683	else if (SvIOKp (sv))
642	{	684	{
643	// we assume we can always read an IV as a UV	685	// we assume we can always read an IV as a UV and vice versa
644	if (SvUV (sv) & ~(UV)0x7fff)	686	// we assume two's complement
645	{	687	// we assume no aliasing issues in the union
646	// large integer, use the (rather slow) snprintf way.	688	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
647	need (enc, sizeof (UV) * 3);	689	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
648	enc->cur +=
649	SvIsUV(sv)
650	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
651	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
652	}
653	else
654	{	690	{
655	// optimise the "small number case"	691	// optimise the "small number case"
656	// code will likely be branchless and use only a single multiplication	692	// code will likely be branchless and use only a single multiplication
		693	// works for numbers up to 59074
657	I32 i = SvIV (sv);	694	I32 i = SvIVX (sv);
658	U32 u;	695	U32 u;
659	char digit, nz = 0;	696	char digit, nz = 0;
660		697
661	need (enc, 6);	698	need (enc, 6);
662		699
…		…
668		705
669	// now output digit by digit, each time masking out the integer part	706	// now output digit by digit, each time masking out the integer part
670	// and multiplying by 5 while moving the decimal point one to the right,	707	// and multiplying by 5 while moving the decimal point one to the right,
671	// resulting in a net multiplication by 10.	708	// resulting in a net multiplication by 10.
672	// we always write the digit to memory but conditionally increment	709	// we always write the digit to memory but conditionally increment
673	// the pointer, to ease the usage of conditional move instructions.	710	// the pointer, to enable the use of conditional move instructions.
674	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffff) 5;	711	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffffUL) 5;
675	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffff) 5;	712	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffffUL) 5;
676	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffff) 5;	713	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffffUL) 5;
677	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffff) 5;	714	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffffUL) 5;
678	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'	715	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
679	}	716	}
		717	else
		718	{
		719	// large integer, use the (rather slow) snprintf way.
		720	need (enc, IVUV_MAXCHARS);
		721	enc->cur +=
		722	SvIsUV(sv)
		723	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
		724	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		725	}
680	}	726	}
681	else if (SvROK (sv))	727	else if (SvROK (sv))
682	encode_rv (enc, SvRV (sv));	728	encode_rv (enc, SvRV (sv));
683	else if (!SvOK (sv))	729	else if (!SvOK (sv) \|\| enc->json.flags & F_ALLOW_UNKNOWN)
684	encode_str (enc, "null", 4, 0);	730	encode_str (enc, "null", 4, 0);
685	else	731	else
686	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	732	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
687	SvPV_nolen (sv), SvFLAGS (sv));	733	SvPV_nolen (sv), SvFLAGS (sv));
688	}	734	}
…		…
698	enc.json = *json;	744	enc.json = *json;
699	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	745	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
700	enc.cur = SvPVX (enc.sv);	746	enc.cur = SvPVX (enc.sv);
701	enc.end = SvEND (enc.sv);	747	enc.end = SvEND (enc.sv);
702	enc.indent = 0;	748	enc.indent = 0;
703	enc.maxdepth = DEC_DEPTH (enc.json.flags);
704	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL	749	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
705	: enc.json.flags & F_LATIN1 ? 0x000100UL	750	: enc.json.flags & F_LATIN1 ? 0x000100UL
706	: 0x10FFFFUL;	751	: 0x110000UL;
707		752
708	SvPOK_only (enc.sv);	753	SvPOK_only (enc.sv);
709	encode_sv (&enc, scalar);	754	encode_sv (&enc, scalar);
710		755
711	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	756	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
…		…
775	if (*dec->cur != ch) \	820	if (*dec->cur != ch) \
776	ERR (# ch " expected"); \	821	ERR (# ch " expected"); \
777	++dec->cur; \	822	++dec->cur; \
778	SE	823	SE
779		824
780	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	825	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
781	#define DEC_DEC_DEPTH --dec->depth	826	#define DEC_DEC_DEPTH --dec->depth
782		827
783	static SV decode_sv (dec_t dec);	828	static SV decode_sv (dec_t dec);
784		829
785	static signed char decode_hexdigit[256];	830	static signed char decode_hexdigit[256];
…		…
889	default:	934	default:
890	--dec_cur;	935	--dec_cur;
891	ERR ("illegal backslash escape sequence in string");	936	ERR ("illegal backslash escape sequence in string");
892	}	937	}
893	}	938	}
894	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	939	else if (expect_true (ch >= 0x20 && ch < 0x80))
895	*cur++ = ch;	940	*cur++ = ch;
896	else if (ch >= 0x80)	941	else if (ch >= 0x80)
897	{	942	{
898	STRLEN clen;	943	STRLEN clen;
899	UV uch;	944	UV uch;
…		…
1157	char *p = dec->cur;	1202	char *p = dec->cur;
1158	char *e = p + 24; // only try up to 24 bytes	1203	char *e = p + 24; // only try up to 24 bytes
1159		1204
1160	for (;;)	1205	for (;;)
1161	{	1206	{
1162	// the >= 0x80 is true on most architectures	1207	// the >= 0x80 is false on most architectures
1163	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')	1208	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
1164	{	1209	{
1165	// slow path, back up and use decode_str	1210	// slow path, back up and use decode_str
1166	SV *key = decode_str (dec);	1211	SV *key = decode_str (dec);
1167	if (!key)	1212	if (!key)
…		…
1298		1343
1299	static SV *	1344	static SV *
1300	decode_sv (dec_t *dec)	1345	decode_sv (dec_t *dec)
1301	{	1346	{
1302	// the beauty of JSON: you need exactly one character lookahead	1347	// the beauty of JSON: you need exactly one character lookahead
1303	// to parse anything.	1348	// to parse everything.
1304	switch (*dec->cur)	1349	switch (*dec->cur)
1305	{	1350	{
1306	case '"': ++dec->cur; return decode_str (dec);	1351	case '"': ++dec->cur; return decode_str (dec);
1307	case '[': ++dec->cur; return decode_av (dec);	1352	case '[': ++dec->cur; return decode_av (dec);
1308	case '{': ++dec->cur; return decode_hv (dec);	1353	case '{': ++dec->cur; return decode_hv (dec);
1309		1354
1310	case '-':	1355	case '-':
1311	case '0': case '1': case '2': case '3': case '4':	1356	case '0': case '1': case '2': case '3': case '4':
1312	case '5': case '6': case '7': case '8': case '9':	1357	case '5': case '6': case '7': case '8': case '9':
1313	return decode_num (dec);	1358	return decode_num (dec);
…		…
1315	case 't':	1360	case 't':
1316	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1361	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1317	{	1362	{
1318	dec->cur += 4;	1363	dec->cur += 4;
1319	#if JSON_SLOW	1364	#if JSON_SLOW
1320	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);	1365	json_true = get_bool ("JSON::XS::true");
1321	#endif	1366	#endif
1322	return SvREFCNT_inc (json_true);	1367	return newSVsv (json_true);
1323	}	1368	}
1324	else	1369	else
1325	ERR ("'true' expected");	1370	ERR ("'true' expected");
1326		1371
1327	break;	1372	break;
…		…
1329	case 'f':	1374	case 'f':
1330	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1375	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1331	{	1376	{
1332	dec->cur += 5;	1377	dec->cur += 5;
1333	#if JSON_SLOW	1378	#if JSON_SLOW
1334	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1379	json_false = get_bool ("JSON::XS::false");
1335	#endif	1380	#endif
1336	return SvREFCNT_inc (json_false);	1381	return newSVsv (json_false);
1337	}	1382	}
1338	else	1383	else
1339	ERR ("'false' expected");	1384	ERR ("'false' expected");
1340		1385
1341	break;	1386	break;
…		…
1359	fail:	1404	fail:
1360	return 0;	1405	return 0;
1361	}	1406	}
1362		1407
1363	static SV *	1408	static SV *
1364	decode_json (SV string, JSON json, UV *offset_return)	1409	decode_json (SV string, JSON json, STRLEN *offset_return)
1365	{	1410	{
1366	dec_t dec;	1411	dec_t dec;
1367	UV offset;	1412	STRLEN offset;
1368	SV *sv;	1413	SV *sv;
1369		1414
1370	SvGETMAGIC (string);	1415	SvGETMAGIC (string);
1371	SvUPGRADE (string, SVt_PV);	1416	SvUPGRADE (string, SVt_PV);
1372		1417
1373	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))	1418	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1419	* assertion with -DDEBUGGING, although SvCUR is documented to
		1420	* return the xpv_cur field which certainly exists after upgrading.
		1421	* according to nicholas clark, calling SvPOK fixes this.
		1422	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1423	* and hope for the best.
		1424	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1425	* assertion business is seriously broken, try yet another workaround
		1426	* for the broken -DDEBUGGING.
		1427	*/
		1428	#ifdef DEBUGGING
		1429	offset = SvOK (string) ? sv_len (string) : 0;
		1430	#else
		1431	offset = SvCUR (string);
		1432	#endif
		1433
		1434	if (offset > json->max_size && json->max_size)
1374	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",	1435	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1375	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));	1436	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
1376		1437
1377	if (json->flags & F_UTF8)	1438	if (json->flags & F_UTF8)
1378	sv_utf8_downgrade (string, 0);	1439	sv_utf8_downgrade (string, 0);
1379	else	1440	else
1380	sv_utf8_upgrade (string);	1441	sv_utf8_upgrade (string);
1381		1442
1382	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1443	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1383		1444
1384	dec.json = *json;	1445	dec.json = *json;
1385	dec.cur = SvPVX (string);	1446	dec.cur = SvPVX (string);
1386	dec.end = SvEND (string);	1447	dec.end = SvEND (string);
1387	dec.err = 0;	1448	dec.err = 0;
1388	dec.depth = 0;	1449	dec.depth = 0;
1389	dec.maxdepth = DEC_DEPTH (dec.json.flags);
1390		1450
1391	if (dec.json.cb_object \|\| dec.json.cb_sk_object)	1451	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
1392	dec.json.flags \|= F_HOOK;	1452	dec.json.flags \|= F_HOOK;
1393		1453
1394	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1454	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
…		…
1445		1505
1446	return sv;	1506	return sv;
1447	}	1507	}
1448		1508
1449	/////////////////////////////////////////////////////////////////////////////	1509	/////////////////////////////////////////////////////////////////////////////
		1510	// incremental parser
		1511
		1512	static void
		1513	incr_parse (JSON *self)
		1514	{
		1515	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1516
		1517	for (;;)
		1518	{
		1519	//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
		1520	switch (self->incr_mode)
		1521	{
		1522	// only used for intiial whitespace skipping
		1523	case INCR_M_WS:
		1524	for (;;)
		1525	{
		1526	if (*p > 0x20)
		1527	{
		1528	self->incr_mode = INCR_M_JSON;
		1529	goto incr_m_json;
		1530	}
		1531	else if (!*p)
		1532	goto interrupt;
		1533
		1534	++p;
		1535	}
		1536
		1537	// skip a single char inside a string (for \\-processing)
		1538	case INCR_M_BS:
		1539	if (!*p)
		1540	goto interrupt;
		1541
		1542	++p;
		1543	self->incr_mode = INCR_M_STR;
		1544	goto incr_m_str;
		1545
		1546	// inside a string
		1547	case INCR_M_STR:
		1548	incr_m_str:
		1549	for (;;)
		1550	{
		1551	if (*p == '"')
		1552	{
		1553	++p;
		1554	self->incr_mode = INCR_M_JSON;
		1555
		1556	if (!self->incr_nest)
		1557	goto interrupt;
		1558
		1559	goto incr_m_json;
		1560	}
		1561	else if (*p == '\\')
		1562	{
		1563	++p; // "virtually" consumes character after \
		1564
		1565	if (!*p) // if at end of string we have to switch modes
		1566	{
		1567	self->incr_mode = INCR_M_BS;
		1568	goto interrupt;
		1569	}
		1570	}
		1571	else if (!*p)
		1572	goto interrupt;
		1573
		1574	++p;
		1575	}
		1576
		1577	// after initial ws, outside string
		1578	case INCR_M_JSON:
		1579	incr_m_json:
		1580	for (;;)
		1581	{
		1582	switch (*p++)
		1583	{
		1584	case 0:
		1585	--p;
		1586	goto interrupt;
		1587
		1588	case 0x09:
		1589	case 0x0a:
		1590	case 0x0d:
		1591	case 0x20:
		1592	if (!self->incr_nest)
		1593	{
		1594	--p; // do not eat the whitespace, let the next round do it
		1595	goto interrupt;
		1596	}
		1597	break;
		1598
		1599	case '"':
		1600	self->incr_mode = INCR_M_STR;
		1601	goto incr_m_str;
		1602
		1603	case '[':
		1604	case '{':
		1605	if (++self->incr_nest > self->max_depth)
		1606	croak (ERR_NESTING_EXCEEDED);
		1607	break;
		1608
		1609	case ']':
		1610	case '}':
		1611	if (--self->incr_nest <= 0)
		1612	goto interrupt;
		1613	}
		1614	}
		1615	}
		1616
		1617	modechange:
		1618	;
		1619	}
		1620
		1621	interrupt:
		1622	self->incr_pos = p - SvPVX (self->incr_text);
		1623	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
		1624	}
		1625
		1626	/////////////////////////////////////////////////////////////////////////////
1450	// XS interface functions	1627	// XS interface functions
1451		1628
1452	MODULE = JSON::XS PACKAGE = JSON::XS	1629	MODULE = JSON::XS PACKAGE = JSON::XS
1453		1630
1454	BOOT:	1631	BOOT:
…		…
1463	: -1;	1640	: -1;
1464		1641
1465	json_stash = gv_stashpv ("JSON::XS" , 1);	1642	json_stash = gv_stashpv ("JSON::XS" , 1);
1466	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);	1643	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1467		1644
1468	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );	1645	json_true = get_bool ("JSON::XS::true");
1469	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1646	json_false = get_bool ("JSON::XS::false");
1470	}	1647	}
1471		1648
1472	PROTOTYPES: DISABLE	1649	PROTOTYPES: DISABLE
1473		1650
1474	void CLONE (...)	1651	void CLONE (...)
…		…
1479	void new (char *klass)	1656	void new (char *klass)
1480	PPCODE:	1657	PPCODE:
1481	{	1658	{
1482	SV *pv = NEWSV (0, sizeof (JSON));	1659	SV *pv = NEWSV (0, sizeof (JSON));
1483	SvPOK_only (pv);	1660	SvPOK_only (pv);
1484	Zero (SvPVX (pv), 1, JSON);	1661	json_init ((JSON *)SvPVX (pv));
1485	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1486	XPUSHs (sv_2mortal (sv_bless (	1662	XPUSHs (sv_2mortal (sv_bless (
1487	newRV_noinc (pv),	1663	newRV_noinc (pv),
1488	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)	1664	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1489	)));	1665	)));
1490	}	1666	}
…		…
1502	allow_nonref = F_ALLOW_NONREF	1678	allow_nonref = F_ALLOW_NONREF
1503	shrink = F_SHRINK	1679	shrink = F_SHRINK
1504	allow_blessed = F_ALLOW_BLESSED	1680	allow_blessed = F_ALLOW_BLESSED
1505	convert_blessed = F_CONV_BLESSED	1681	convert_blessed = F_CONV_BLESSED
1506	relaxed = F_RELAXED	1682	relaxed = F_RELAXED
		1683	allow_unknown = F_ALLOW_UNKNOWN
1507	PPCODE:	1684	PPCODE:
1508	{	1685	{
1509	if (enable)	1686	if (enable)
1510	self->flags \|= ix;	1687	self->flags \|= ix;
1511	else	1688	else
…		…
1526	get_allow_nonref = F_ALLOW_NONREF	1703	get_allow_nonref = F_ALLOW_NONREF
1527	get_shrink = F_SHRINK	1704	get_shrink = F_SHRINK
1528	get_allow_blessed = F_ALLOW_BLESSED	1705	get_allow_blessed = F_ALLOW_BLESSED
1529	get_convert_blessed = F_CONV_BLESSED	1706	get_convert_blessed = F_CONV_BLESSED
1530	get_relaxed = F_RELAXED	1707	get_relaxed = F_RELAXED
		1708	get_allow_unknown = F_ALLOW_UNKNOWN
1531	PPCODE:	1709	PPCODE:
1532	XPUSHs (boolSV (self->flags & ix));	1710	XPUSHs (boolSV (self->flags & ix));
1533		1711
1534	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1712	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1535	PPCODE:	1713	PPCODE:
1536	{	1714	self->max_depth = max_depth;
1537	UV log2 = 0;
1538
1539	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1540
1541	while ((1UL << log2) < max_depth)
1542	++log2;
1543
1544	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1545
1546	XPUSHs (ST (0));	1715	XPUSHs (ST (0));
1547	}
1548		1716
1549	U32 get_max_depth (JSON *self)	1717	U32 get_max_depth (JSON *self)
1550	CODE:	1718	CODE:
1551	RETVAL = DEC_DEPTH (self->flags);	1719	RETVAL = self->max_depth;
1552	OUTPUT:	1720	OUTPUT:
1553	RETVAL	1721	RETVAL
1554		1722
1555	void max_size (JSON *self, UV max_size = 0)	1723	void max_size (JSON *self, U32 max_size = 0)
1556	PPCODE:	1724	PPCODE:
1557	{	1725	self->max_size = max_size;
1558	UV log2 = 0;
1559
1560	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1561	if (max_size == 1) max_size = 2;
1562
1563	while ((1UL << log2) < max_size)
1564	++log2;
1565
1566	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
1567
1568	XPUSHs (ST (0));	1726	XPUSHs (ST (0));
1569	}
1570		1727
1571	int get_max_size (JSON *self)	1728	int get_max_size (JSON *self)
1572	CODE:	1729	CODE:
1573	RETVAL = DEC_SIZE (self->flags);	1730	RETVAL = self->max_size;
1574	OUTPUT:	1731	OUTPUT:
1575	RETVAL	1732	RETVAL
1576		1733
1577	void filter_json_object (JSON self, SV cb = &PL_sv_undef)	1734	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1578	PPCODE:	1735	PPCODE:
…		…
1614	XPUSHs (decode_json (jsonstr, self, 0));	1771	XPUSHs (decode_json (jsonstr, self, 0));
1615		1772
1616	void decode_prefix (JSON self, SV jsonstr)	1773	void decode_prefix (JSON self, SV jsonstr)
1617	PPCODE:	1774	PPCODE:
1618	{	1775	{
1619	UV offset;	1776	STRLEN offset;
1620	EXTEND (SP, 2);	1777	EXTEND (SP, 2);
1621	PUSHs (decode_json (jsonstr, self, &offset));	1778	PUSHs (decode_json (jsonstr, self, &offset));
1622	PUSHs (sv_2mortal (newSVuv (offset)));	1779	PUSHs (sv_2mortal (newSVuv (offset)));
		1780	}
		1781
		1782	void incr_parse (JSON self, SV jsonstr = 0)
		1783	PPCODE:
		1784	{
		1785	if (!self->incr_text)
		1786	self->incr_text = newSVpvn ("", 0);
		1787
		1788	// append data, if any
		1789	if (jsonstr)
		1790	{
		1791	if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text))
		1792	{
		1793	/* utf-8-ness differs, need to upgrade */
		1794	sv_utf8_upgrade (self->incr_text);
		1795
		1796	if (self->incr_pos)
		1797	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1798	- (U8 *)SvPVX (self->incr_text);
		1799	}
		1800
		1801	{
		1802	STRLEN len;
		1803	const char *str = SvPV (jsonstr, len);
		1804	SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1);
		1805	Move (str, SvEND (self->incr_text), len, char);
		1806	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1807	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1808	}
		1809	}
		1810
		1811	if (GIMME_V != G_VOID)
		1812	do
		1813	{
		1814	STRLEN offset;
		1815
		1816	if (!INCR_DONE (self))
		1817	{
		1818	incr_parse (self);
		1819
		1820	if (self->incr_pos > self->max_size && self->max_size)
		1821	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1822	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1823
		1824	if (!INCR_DONE (self))
		1825	break;
		1826	}
		1827
		1828	XPUSHs (decode_json (self->incr_text, self, &offset));
		1829
		1830	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
		1831	self->incr_pos -= offset;
		1832	self->incr_nest = 0;
		1833	self->incr_mode = 0;
		1834	}
		1835	while (GIMME_V == G_ARRAY);
		1836	}
		1837
		1838	SV incr_text (JSON self)
		1839	ATTRS: lvalue
		1840	CODE:
		1841	{
		1842	if (self->incr_pos)
		1843	croak ("incr_text can not be called when the incremental parser already started parsing");
		1844
		1845	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		1846	}
		1847	OUTPUT:
		1848	RETVAL
		1849
		1850	void incr_skip (JSON *self)
		1851	CODE:
		1852	{
		1853	if (self->incr_pos)
		1854	{
		1855	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1856	self->incr_pos = 0;
		1857	self->incr_nest = 0;
		1858	self->incr_mode = 0;
		1859	}
		1860	}
		1861
		1862	void incr_reset (JSON *self)
		1863	CODE:
		1864	{
		1865	SvREFCNT_dec (self->incr_text);
		1866	self->incr_text = 0;
		1867	self->incr_pos = 0;
		1868	self->incr_nest = 0;
		1869	self->incr_mode = 0;
1623	}	1870	}
1624		1871
1625	void DESTROY (JSON *self)	1872	void DESTROY (JSON *self)
1626	CODE:	1873	CODE:
1627	SvREFCNT_dec (self->cb_sk_object);	1874	SvREFCNT_dec (self->cb_sk_object);
1628	SvREFCNT_dec (self->cb_object);	1875	SvREFCNT_dec (self->cb_object);
		1876	SvREFCNT_dec (self->incr_text);
1629		1877
1630	PROTOTYPES: ENABLE	1878	PROTOTYPES: ENABLE
1631		1879
1632	void encode_json (SV *scalar)	1880	void encode_json (SV *scalar)
		1881	ALIAS:
		1882	to_json_ = 0
		1883	encode_json = F_UTF8
1633	PPCODE:	1884	PPCODE:
1634	{	1885	{
1635	JSON json = { F_DEFAULT \| F_UTF8 };	1886	JSON json;
		1887	json_init (&json);
		1888	json.flags \|= ix;
1636	XPUSHs (encode_json (scalar, &json));	1889	XPUSHs (encode_json (scalar, &json));
1637	}	1890	}
1638		1891
1639	void decode_json (SV *jsonstr)	1892	void decode_json (SV *jsonstr)
		1893	ALIAS:
		1894	from_json_ = 0
		1895	decode_json = F_UTF8
1640	PPCODE:	1896	PPCODE:
1641	{	1897	{
1642	JSON json = { F_DEFAULT \| F_UTF8 };	1898	JSON json;
		1899	json_init (&json);
		1900	json.flags \|= ix;
1643	XPUSHs (decode_json (jsonstr, &json, 0));	1901	XPUSHs (decode_json (jsonstr, &json, 0));
1644	}	1902	}
1645		1903
		1904

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.72 by root, Wed Mar 19 04:08:22 2008 UTC vs. Revision 1.92 by root, Mon Sep 22 07:29:29 2008 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.72 by root, Wed Mar 19 04:08:22 2008 UTC vs.
Revision 1.92 by root, Mon Sep 22 07:29:29 2008 UTC