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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.16 by root, Sun Mar 25 02:37:00 2007 UTC vs.
Revision 1.56 by root, Thu Jul 26 11:33:35 2007 UTC

…		…
1	#include "EXTERN.h"	1	#include "EXTERN.h"
2	#include "perl.h"	2	#include "perl.h"
3	#include "XSUB.h"	3	#include "XSUB.h"
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>
		9	#include <float.h>
8		10
		11	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
		12	# define snprintf _snprintf // C compilers have this in stdio.h
		13	#endif
		14
		15	// 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	#ifndef UTF8_MAXBYTES
		18	# define UTF8_MAXBYTES 13
		19	#endif
		20
9	#define F_ASCII 0x00000001	21	#define F_ASCII 0x00000001UL
		22	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	23	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	24	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	25	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	26	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	27	#define F_SPACE_AFTER 0x00000040UL
15	#define F_ALLOW_NONREF 0x00000080	28	#define F_ALLOW_NONREF 0x00000100UL
16	#define F_SHRINK 0x00000100	29	#define F_SHRINK 0x00000200UL
		30	#define F_ALLOW_BLESSED 0x00000400UL
		31	#define F_CONV_BLESSED 0x00000800UL
		32	#define F_MAXDEPTH 0xf8000000UL
		33	#define S_MAXDEPTH 27
		34	#define F_MAXSIZE 0x01f00000UL
		35	#define S_MAXSIZE 20
		36	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
17		37
18	// F_SKIPINVALID?	38	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
19	// F_EXECCODEREF?	39	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
20	// F_SELFCONVERT?
21		40
22	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER	41	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER
23	#define F_DEFAULT 0	42	#define F_DEFAULT (9UL << S_MAXDEPTH)
24		43
25	#define INIT_SIZE 32 // initial scalar size to be allocated	44	#define INIT_SIZE 32 // initial scalar size to be allocated
26	#define INDENT_STEP 3 // spaces per indentation level	45	#define INDENT_STEP 3 // spaces per indentation level
27		46
28	#define UTF8_MAX_LEN 11 // for perls UTF-X: max. number of octets per character
29	#define SHORT_STRING_LEN 512 // special-case strings of up to this size	47	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
30		48
31	#define SB do {	49	#define SB do {
32	#define SE } while (0)	50	#define SE } while (0)
33		51
		52	#if __GNUC__ >= 3
		53	# define expect(expr,value) __builtin_expect ((expr),(value))
		54	# define inline inline
		55	#else
		56	# define expect(expr,value) (expr)
		57	# define inline static
		58	#endif
		59
		60	#define expect_false(expr) expect ((expr) != 0, 0)
		61	#define expect_true(expr) expect ((expr) != 0, 1)
		62
34	static HV *json_stash; // JSON::XS::	63	static HV json_stash, json_boolean_stash; // JSON::XS::
		64	static SV json_true, json_false;
		65
		66	typedef struct {
		67	U32 flags;
		68	SV *cb_object;
		69	HV *cb_sk_object;
		70	} JSON;
35		71
36	/////////////////////////////////////////////////////////////////////////////	72	/////////////////////////////////////////////////////////////////////////////
37	// utility functions	73	// utility functions
38		74
39	static UV *	75	inline void
40	SvJSON (SV *sv)
41	{
42	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
43	croak ("object is not of type JSON::XS");
44
45	return &SvUVX (SvRV (sv));
46	}
47
48	static void
49	shrink (SV *sv)	76	shrink (SV *sv)
50	{	77	{
51	sv_utf8_downgrade (sv, 1);	78	sv_utf8_downgrade (sv, 1);
52	if (SvLEN (sv) > SvCUR (sv) + 1)	79	if (SvLEN (sv) > SvCUR (sv) + 1)
53	{	80	{
62	// decode an utf-8 character and return it, or (UV)-1 in	89	// decode an utf-8 character and return it, or (UV)-1 in
63	// case of an error.	90	// case of an error.
64	// we special-case "safe" characters from U+80 .. U+7FF,	91	// we special-case "safe" characters from U+80 .. U+7FF,
65	// but use the very good perl function to parse anything else.	92	// but use the very good perl function to parse anything else.
66	// note that we never call this function for a ascii codepoints	93	// note that we never call this function for a ascii codepoints
67	static UV	94	inline UV
68	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	95	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
69	{	96	{
70	if (s[0] > 0xdf \|\| s[0] < 0xc2)	97	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))
71	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	98	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
72	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	99	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
73	{	100	{
74	*clen = 2;	101	*clen = 2;
75	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);	102	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
76	}	103	}
77	else	104	else
		105	{
		106	*clen = (STRLEN)-1;
78	return (UV)-1;	107	return (UV)-1;
		108	}
79	}	109	}
80		110
81	/////////////////////////////////////////////////////////////////////////////	111	/////////////////////////////////////////////////////////////////////////////
82	// encoder	112	// encoder
83		113
…		…
85	typedef struct	115	typedef struct
86	{	116	{
87	char *cur; // SvPVX (sv) + current output position	117	char *cur; // SvPVX (sv) + current output position
88	char *end; // SvEND (sv)	118	char *end; // SvEND (sv)
89	SV *sv; // result scalar	119	SV *sv; // result scalar
90	UV flags; // F_*	120	JSON json;
91	int indent; // indentation level	121	U32 indent; // indentation level
92	int max_depth; // max. recursion level	122	U32 maxdepth; // max. indentation/recursion level
93	} enc_t;	123	} enc_t;
94		124
95	static void	125	inline void
96	need (enc_t *enc, STRLEN len)	126	need (enc_t *enc, STRLEN len)
97	{	127	{
98	if (enc->cur + len >= enc->end)	128	if (expect_false (enc->cur + len >= enc->end))
99	{	129	{
100	STRLEN cur = enc->cur - SvPVX (enc->sv);	130	STRLEN cur = enc->cur - SvPVX (enc->sv);
101	SvGROW (enc->sv, cur + len + 1);	131	SvGROW (enc->sv, cur + len + 1);
102	enc->cur = SvPVX (enc->sv) + cur;	132	enc->cur = SvPVX (enc->sv) + cur;
103	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	133	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
104	}	134	}
105	}	135	}
106		136
107	static void	137	inline void
108	encode_ch (enc_t *enc, char ch)	138	encode_ch (enc_t *enc, char ch)
109	{	139	{
110	need (enc, 1);	140	need (enc, 1);
111	*enc->cur++ = ch;	141	*enc->cur++ = ch;
112	}	142	}
…		…
120		150
121	while (str < end)	151	while (str < end)
122	{	152	{
123	unsigned char ch = (unsigned char )str;	153	unsigned char ch = (unsigned char )str;
124		154
125	if (ch >= 0x20 && ch < 0x80) // most common case	155	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
126	{	156	{
127	if (ch == '"') // but with slow exceptions	157	if (expect_false (ch == '"')) // but with slow exceptions
128	{	158	{
129	need (enc, len += 1);	159	need (enc, len += 1);
130	*enc->cur++ = '\\';	160	*enc->cur++ = '\\';
131	*enc->cur++ = '"';	161	*enc->cur++ = '"';
132	}	162	}
133	else if (ch == '\\')	163	else if (expect_false (ch == '\\'))
134	{	164	{
135	need (enc, len += 1);	165	need (enc, len += 1);
136	*enc->cur++ = '\\';	166	*enc->cur++ = '\\';
137	*enc->cur++ = '\\';	167	*enc->cur++ = '\\';
138	}	168	}
…		…
156	STRLEN clen;	186	STRLEN clen;
157	UV uch;	187	UV uch;
158		188
159	if (is_utf8)	189	if (is_utf8)
160	{	190	{
161	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
162	uch = decode_utf8 (str, end - str, &clen);	191	uch = decode_utf8 (str, end - str, &clen);
163	if (clen == (STRLEN)-1)	192	if (clen == (STRLEN)-1)
164	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	193	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
165	}	194	}
166	else	195	else
…		…
170	}	199	}
171		200
172	if (uch > 0x10FFFFUL)	201	if (uch > 0x10FFFFUL)
173	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);	202	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
174		203
175	if (uch < 0x80 \|\| enc->flags & F_ASCII)	204	if (uch < 0x80 \|\| enc->json.flags & F_ASCII \|\| (enc->json.flags & F_LATIN1 && uch > 0xFF))
176	{	205	{
177	if (uch > 0xFFFFUL)	206	if (uch > 0xFFFFUL)
178	{	207	{
179	need (enc, len += 11);	208	need (enc, len += 11);
180	sprintf (enc->cur, "\\u%04x\\u%04x",	209	sprintf (enc->cur, "\\u%04x\\u%04x",
…		…
194	*enc->cur++ = hexdigit [(uch >> 0) & 15];	223	*enc->cur++ = hexdigit [(uch >> 0) & 15];
195	}	224	}
196		225
197	str += clen;	226	str += clen;
198	}	227	}
		228	else if (enc->json.flags & F_LATIN1)
		229	{
		230	*enc->cur++ = uch;
		231	str += clen;
		232	}
199	else if (is_utf8)	233	else if (is_utf8)
200	{	234	{
201	need (enc, len += clen);	235	need (enc, len += clen);
202	do	236	do
203	{	237	{
…		…
205	}	239	}
206	while (--clen);	240	while (--clen);
207	}	241	}
208	else	242	else
209	{	243	{
210	need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed	244	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
211	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	245	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
212	++str;	246	++str;
213	}	247	}
214	}	248	}
215	}	249	}
…		…
217		251
218	--len;	252	--len;
219	}	253	}
220	}	254	}
221		255
222	static void	256	inline void
223	encode_indent (enc_t *enc)	257	encode_indent (enc_t *enc)
224	{	258	{
225	if (enc->flags & F_INDENT)	259	if (enc->json.flags & F_INDENT)
226	{	260	{
227	int spaces = enc->indent * INDENT_STEP;	261	int spaces = enc->indent * INDENT_STEP;
228		262
229	need (enc, spaces);	263	need (enc, spaces);
230	memset (enc->cur, ' ', spaces);	264	memset (enc->cur, ' ', spaces);
231	enc->cur += spaces;	265	enc->cur += spaces;
232	}	266	}
233	}	267	}
234		268
235	static void	269	inline void
236	encode_space (enc_t *enc)	270	encode_space (enc_t *enc)
237	{	271	{
238	need (enc, 1);	272	need (enc, 1);
239	encode_ch (enc, ' ');	273	encode_ch (enc, ' ');
240	}	274	}
241		275
242	static void	276	inline void
243	encode_nl (enc_t *enc)	277	encode_nl (enc_t *enc)
244	{	278	{
245	if (enc->flags & F_INDENT)	279	if (enc->json.flags & F_INDENT)
246	{	280	{
247	need (enc, 1);	281	need (enc, 1);
248	encode_ch (enc, '\n');	282	encode_ch (enc, '\n');
249	}	283	}
250	}	284	}
251		285
252	static void	286	inline void
253	encode_comma (enc_t *enc)	287	encode_comma (enc_t *enc)
254	{	288	{
255	encode_ch (enc, ',');	289	encode_ch (enc, ',');
256		290
257	if (enc->flags & F_INDENT)	291	if (enc->json.flags & F_INDENT)
258	encode_nl (enc);	292	encode_nl (enc);
259	else if (enc->flags & F_SPACE_AFTER)	293	else if (enc->json.flags & F_SPACE_AFTER)
260	encode_space (enc);	294	encode_space (enc);
261	}	295	}
262		296
263	static void encode_sv (enc_t enc, SV sv);	297	static void encode_sv (enc_t enc, SV sv);
264		298
265	static void	299	static void
266	encode_av (enc_t enc, AV av)	300	encode_av (enc_t enc, AV av)
267	{	301	{
268	int i, len = av_len (av);	302	int i, len = av_len (av);
269		303
		304	if (enc->indent >= enc->maxdepth)
		305	croak ("data structure too deep (hit recursion limit)");
		306
270	encode_ch (enc, '['); encode_nl (enc);	307	encode_ch (enc, '['); encode_nl (enc);
271	++enc->indent;	308	++enc->indent;
272		309
273	for (i = 0; i <= len; ++i)	310	for (i = 0; i <= len; ++i)
274	{	311	{
		312	SV **svp = av_fetch (av, i, 0);
		313
275	encode_indent (enc);	314	encode_indent (enc);
276	encode_sv (enc, *av_fetch (av, i, 0));	315
		316	if (svp)
		317	encode_sv (enc, *svp);
		318	else
		319	encode_str (enc, "null", 4, 0);
277		320
278	if (i < len)	321	if (i < len)
279	encode_comma (enc);	322	encode_comma (enc);
280	}	323	}
281		324
…		…
304	else	347	else
305	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	348	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
306		349
307	encode_ch (enc, '"');	350	encode_ch (enc, '"');
308		351
309	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	352	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
310	encode_ch (enc, ':');	353	encode_ch (enc, ':');
311	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	354	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
312	encode_sv (enc, HeVAL (he));	355	encode_sv (enc, HeVAL (he));
313	}	356	}
314		357
315	// compare hash entries, used when all keys are bytestrings	358	// compare hash entries, used when all keys are bytestrings
316	static int	359	static int
…		…
339		382
340	static void	383	static void
341	encode_hv (enc_t enc, HV hv)	384	encode_hv (enc_t enc, HV hv)
342	{	385	{
343	int count, i;	386	int count, i;
		387
		388	if (enc->indent >= enc->maxdepth)
		389	croak ("data structure too deep (hit recursion limit)");
344		390
345	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	391	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
346		392
347	if ((count = hv_iterinit (hv)))	393	if ((count = hv_iterinit (hv)))
348	{	394	{
349	// for canonical output we have to sort by keys first	395	// for canonical output we have to sort by keys first
350	// actually, this is mostly due to the stupid so-called	396	// actually, this is mostly due to the stupid so-called
351	// security workaround added somewhere in 5.8.x.	397	// security workaround added somewhere in 5.8.x.
352	// that randomises hash orderings	398	// that randomises hash orderings
353	if (enc->flags & F_CANONICAL)	399	if (enc->json.flags & F_CANONICAL)
354	{	400	{
355	HE he, hes [count]; // if your compiler dies here, you need to enable C99 mode
356	int fast = 1;	401	int fast = 1;
		402	HE *he;
		403	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
		404	HE *hes = _alloca (count sizeof (HE));
		405	#else
		406	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		407	#endif
357		408
358	i = 0;	409	i = 0;
359	while ((he = hv_iternext (hv)))	410	while ((he = hv_iternext (hv)))
360	{	411	{
361	hes [i++] = he;	412	hes [i++] = he;
…		…
396		447
397	encode_nl (enc);	448	encode_nl (enc);
398	}	449	}
399	else	450	else
400	{	451	{
401	SV *sv;
402	HE *he = hv_iternext (hv);	452	HE *he = hv_iternext (hv);
403		453
404	for (;;)	454	for (;;)
405	{	455	{
406	encode_indent (enc);	456	encode_indent (enc);
…		…
415	encode_nl (enc);	465	encode_nl (enc);
416	}	466	}
417	}	467	}
418		468
419	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	469	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
		470	}
		471
		472	// encode objects, arrays and special \0=false and \1=true values.
		473	static void
		474	encode_rv (enc_t enc, SV sv)
		475	{
		476	svtype svt;
		477
		478	SvGETMAGIC (sv);
		479	svt = SvTYPE (sv);
		480
		481	if (expect_false (SvOBJECT (sv)))
		482	{
		483	if (SvSTASH (sv) == json_boolean_stash)
		484	{
		485	if (SvIV (sv))
		486	encode_str (enc, "true", 4, 0);
		487	else
		488	encode_str (enc, "false", 5, 0);
		489	}
		490	else
		491	{
		492	#if 0
		493	if (0 && sv_derived_from (rv, "JSON::Literal"))
		494	{
		495	// not yet
		496	}
		497	#endif
		498	if (enc->json.flags & F_CONV_BLESSED)
		499	{
		500	// we re-bless the reference to get overload and other niceties right
		501	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
		502
		503	if (to_json)
		504	{
		505	int count;
		506	dSP;
		507
		508	ENTER; SAVETMPS; PUSHMARK (SP);
		509	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		510
		511	// calling with G_SCALAR ensures that we always get a 1 return value
		512	PUTBACK;
		513	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		514	SPAGAIN;
		515
		516	// catch this surprisingly common error
		517	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		518	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		519
		520	sv = POPs;
		521	PUTBACK;
		522
		523	encode_sv (enc, sv);
		524
		525	FREETMPS; LEAVE;
		526	}
		527	else if (enc->json.flags & F_ALLOW_BLESSED)
		528	encode_str (enc, "null", 4, 0);
		529	else
		530	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		531	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		532	}
		533	else if (enc->json.flags & F_ALLOW_BLESSED)
		534	encode_str (enc, "null", 4, 0);
		535	else
		536	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		537	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		538	}
		539	}
		540	else if (svt == SVt_PVHV)
		541	encode_hv (enc, (HV *)sv);
		542	else if (svt == SVt_PVAV)
		543	encode_av (enc, (AV *)sv);
		544	else if (svt < SVt_PVAV)
		545	{
		546	STRLEN len = 0;
		547	char *pv = svt ? SvPV (sv, len) : 0;
		548
		549	if (len == 1 && *pv == '1')
		550	encode_str (enc, "true", 4, 0);
		551	else if (len == 1 && *pv == '0')
		552	encode_str (enc, "false", 5, 0);
		553	else
		554	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		555	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		556	}
		557	else
		558	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		559	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
420	}	560	}
421		561
422	static void	562	static void
423	encode_sv (enc_t enc, SV sv)	563	encode_sv (enc_t enc, SV sv)
424	{	564	{
…		…
432	encode_str (enc, str, len, SvUTF8 (sv));	572	encode_str (enc, str, len, SvUTF8 (sv));
433	encode_ch (enc, '"');	573	encode_ch (enc, '"');
434	}	574	}
435	else if (SvNOKp (sv))	575	else if (SvNOKp (sv))
436	{	576	{
		577	// trust that perl will do the right thing w.r.t. JSON syntax.
437	need (enc, NV_DIG + 32);	578	need (enc, NV_DIG + 32);
438	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	579	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
439	enc->cur += strlen (enc->cur);	580	enc->cur += strlen (enc->cur);
440	}	581	}
441	else if (SvIOKp (sv))	582	else if (SvIOKp (sv))
442	{	583	{
443	need (enc, 64);	584	// we assume we can always read an IV as a UV
		585	if (SvUV (sv) & ~(UV)0x7fff)
		586	{
		587	// large integer, use the (rather slow) snprintf way.
		588	need (enc, sizeof (UV) * 3);
444	enc->cur +=	589	enc->cur +=
445	SvIsUV(sv)	590	SvIsUV(sv)
446	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	591	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
447	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	592	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		593	}
		594	else
		595	{
		596	// optimise the "small number case"
		597	// code will likely be branchless and use only a single multiplication
		598	I32 i = SvIV (sv);
		599	U32 u;
		600	char digit, nz = 0;
		601
		602	need (enc, 6);
		603
		604	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		605	u = i < 0 ? -i : i;
		606
		607	// convert to 4.28 fixed-point representation
		608	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		609
		610	// now output digit by digit, each time masking out the integer part
		611	// and multiplying by 5 while moving the decimal point one to the right,
		612	// resulting in a net multiplication by 10.
		613	// we always write the digit to memory but conditionally increment
		614	// the pointer, to ease the usage of conditional move instructions.
		615	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffff) 5;
		616	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffff) 5;
		617	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffff) 5;
		618	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffff) 5;
		619	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		620	}
448	}	621	}
449	else if (SvROK (sv))	622	else if (SvROK (sv))
450	{	623	encode_rv (enc, SvRV (sv));
451	SV *rv = SvRV (sv);
452
453	if (enc->indent >= enc->max_depth)
454	croak ("data structure too deep (hit recursion limit)");
455
456	switch (SvTYPE (rv))
457	{
458	case SVt_PVAV: encode_av (enc, (AV *)rv); break;
459	case SVt_PVHV: encode_hv (enc, (HV *)rv); break;
460
461	default:
462	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
463	SvPV_nolen (sv));
464	}
465	}
466	else if (!SvOK (sv))	624	else if (!SvOK (sv))
467	encode_str (enc, "null", 4, 0);	625	encode_str (enc, "null", 4, 0);
468	else	626	else
469	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	627	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
470	SvPV_nolen (sv), SvFLAGS (sv));	628	SvPV_nolen (sv), SvFLAGS (sv));
471	}	629	}
472		630
473	static SV *	631	static SV *
474	encode_json (SV *scalar, UV flags)	632	encode_json (SV scalar, JSON json)
475	{	633	{
		634	enc_t enc;
		635
476	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	636	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
477	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	637	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
478		638
479	enc_t enc;	639	enc.json = *json;
480	enc.flags = flags;
481	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	640	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
482	enc.cur = SvPVX (enc.sv);	641	enc.cur = SvPVX (enc.sv);
483	enc.end = SvEND (enc.sv);	642	enc.end = SvEND (enc.sv);
484	enc.indent = 0;	643	enc.indent = 0;
485	enc.max_depth = 0x7fffffffUL;	644	enc.maxdepth = DEC_DEPTH (enc.json.flags);
486		645
487	SvPOK_only (enc.sv);	646	SvPOK_only (enc.sv);
488	encode_sv (&enc, scalar);	647	encode_sv (&enc, scalar);
489		648
		649	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		650	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		651
490	if (!(flags & (F_ASCII \| F_UTF8)))	652	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))
491	SvUTF8_on (enc.sv);	653	SvUTF8_on (enc.sv);
492		654
493	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
494
495	if (enc.flags & F_SHRINK)	655	if (enc.json.flags & F_SHRINK)
496	shrink (enc.sv);	656	shrink (enc.sv);
497		657
498	return enc.sv;	658	return enc.sv;
499	}	659	}
500		660
…		…
505	typedef struct	665	typedef struct
506	{	666	{
507	char *cur; // current parser pointer	667	char *cur; // current parser pointer
508	char *end; // end of input string	668	char *end; // end of input string
509	const char *err; // parse error, if != 0	669	const char *err; // parse error, if != 0
510	UV flags; // F_*	670	JSON json;
		671	U32 depth; // recursion depth
		672	U32 maxdepth; // recursion depth limit
511	} dec_t;	673	} dec_t;
512		674
513	static void	675	inline void
514	decode_ws (dec_t *dec)	676	decode_ws (dec_t *dec)
515	{	677	{
516	for (;;)	678	for (;;)
517	{	679	{
518	char ch = *dec->cur;	680	char ch = *dec->cur;
…		…
524	++dec->cur;	686	++dec->cur;
525	}	687	}
526	}	688	}
527		689
528	#define ERR(reason) SB dec->err = reason; goto fail; SE	690	#define ERR(reason) SB dec->err = reason; goto fail; SE
		691
529	#define EXPECT_CH(ch) SB \	692	#define EXPECT_CH(ch) SB \
530	if (*dec->cur != ch) \	693	if (*dec->cur != ch) \
531	ERR (# ch " expected"); \	694	ERR (# ch " expected"); \
532	++dec->cur; \	695	++dec->cur; \
533	SE	696	SE
534		697
		698	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		699	#define DEC_DEC_DEPTH --dec->depth
		700
535	static SV decode_sv (dec_t dec);	701	static SV decode_sv (dec_t dec);
536		702
537	static signed char decode_hexdigit[256];	703	static signed char decode_hexdigit[256];
538		704
539	static UV	705	static UV
540	decode_4hex (dec_t *dec)	706	decode_4hex (dec_t *dec)
541	{	707	{
542	signed char d1, d2, d3, d4;	708	signed char d1, d2, d3, d4;
543	unsigned char cur = (unsigned char )dec->cur;	709	unsigned char cur = (unsigned char )dec->cur;
544		710
545	d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected");	711	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
546	d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected");	712	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
547	d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected");	713	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
548	d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected");	714	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
549		715
550	dec->cur += 4;	716	dec->cur += 4;
551		717
552	return ((UV)d1) << 12	718	return ((UV)d1) << 12
553	\| ((UV)d2) << 8	719	\| ((UV)d2) << 8
…		…
561	static SV *	727	static SV *
562	decode_str (dec_t *dec)	728	decode_str (dec_t *dec)
563	{	729	{
564	SV *sv = 0;	730	SV *sv = 0;
565	int utf8 = 0;	731	int utf8 = 0;
		732	char *dec_cur = dec->cur;
566		733
567	do	734	do
568	{	735	{
569	char buf [SHORT_STRING_LEN + UTF8_MAX_LEN];	736	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
570	char *cur = buf;	737	char *cur = buf;
571		738
572	do	739	do
573	{	740	{
574	unsigned char ch = (unsigned char )dec->cur++;	741	unsigned char ch = (unsigned char )dec_cur++;
575		742
576	if (ch == '"')	743	if (expect_false (ch == '"'))
577	{	744	{
578	--dec->cur;	745	--dec_cur;
579	break;	746	break;
580	}	747	}
581	else if (ch == '\\')	748	else if (expect_false (ch == '\\'))
582	{	749	{
583	switch (*dec->cur)	750	switch (*dec_cur)
584	{	751	{
585	case '\\':	752	case '\\':
586	case '/':	753	case '/':
587	case '"': cur++ = dec->cur++; break;	754	case '"': cur++ = dec_cur++; break;
588		755
589	case 'b': ++dec->cur; *cur++ = '\010'; break;	756	case 'b': ++dec_cur; *cur++ = '\010'; break;
590	case 't': ++dec->cur; *cur++ = '\011'; break;	757	case 't': ++dec_cur; *cur++ = '\011'; break;
591	case 'n': ++dec->cur; *cur++ = '\012'; break;	758	case 'n': ++dec_cur; *cur++ = '\012'; break;
592	case 'f': ++dec->cur; *cur++ = '\014'; break;	759	case 'f': ++dec_cur; *cur++ = '\014'; break;
593	case 'r': ++dec->cur; *cur++ = '\015'; break;	760	case 'r': ++dec_cur; *cur++ = '\015'; break;
594		761
595	case 'u':	762	case 'u':
596	{	763	{
597	UV lo, hi;	764	UV lo, hi;
598	++dec->cur;	765	++dec_cur;
599		766
		767	dec->cur = dec_cur;
600	hi = decode_4hex (dec);	768	hi = decode_4hex (dec);
		769	dec_cur = dec->cur;
601	if (hi == (UV)-1)	770	if (hi == (UV)-1)
602	goto fail;	771	goto fail;
603		772
604	// possibly a surrogate pair	773	// possibly a surrogate pair
605	if (hi >= 0xd800)	774	if (hi >= 0xd800)
606	if (hi < 0xdc00)	775	if (hi < 0xdc00)
607	{	776	{
608	if (dec->cur [0] != '\\' \|\| dec->cur [1] != 'u')	777	if (dec_cur [0] != '\\' \|\| dec_cur [1] != 'u')
609	ERR ("missing low surrogate character in surrogate pair");	778	ERR ("missing low surrogate character in surrogate pair");
610		779
611	dec->cur += 2;	780	dec_cur += 2;
612		781
		782	dec->cur = dec_cur;
613	lo = decode_4hex (dec);	783	lo = decode_4hex (dec);
		784	dec_cur = dec->cur;
614	if (lo == (UV)-1)	785	if (lo == (UV)-1)
615	goto fail;	786	goto fail;
616		787
617	if (lo < 0xdc00 \|\| lo >= 0xe000)	788	if (lo < 0xdc00 \|\| lo >= 0xe000)
618	ERR ("surrogate pair expected");	789	ERR ("surrogate pair expected");
…		…
632	*cur++ = hi;	803	*cur++ = hi;
633	}	804	}
634	break;	805	break;
635		806
636	default:	807	default:
637	--dec->cur;	808	--dec_cur;
638	ERR ("illegal backslash escape sequence in string");	809	ERR ("illegal backslash escape sequence in string");
639	}	810	}
640	}	811	}
641	else if (ch >= 0x20 && ch <= 0x7f)	812	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
642	*cur++ = ch;	813	*cur++ = ch;
643	else if (ch >= 0x80)	814	else if (ch >= 0x80)
644	{	815	{
645	--dec->cur;
646
647	STRLEN clen;	816	STRLEN clen;
		817	UV uch;
		818
		819	--dec_cur;
		820
648	UV uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	821	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
649	if (clen == (STRLEN)-1)	822	if (clen == (STRLEN)-1)
650	ERR ("malformed UTF-8 character in JSON string");	823	ERR ("malformed UTF-8 character in JSON string");
651		824
652	do	825	do
653	{
654	cur++ = dec->cur++;	826	cur++ = dec_cur++;
655	}
656	while (--clen);	827	while (--clen);
657		828
658	utf8 = 1;	829	utf8 = 1;
659	}	830	}
660	else if (!ch)
661	ERR ("unexpected end of string while parsing json string");
662	else	831	else
		832	{
		833	--dec_cur;
		834
		835	if (!ch)
		836	ERR ("unexpected end of string while parsing JSON string");
		837	else
663	ERR ("invalid character encountered");	838	ERR ("invalid character encountered while parsing JSON string");
664		839	}
665	}	840	}
666	while (cur < buf + SHORT_STRING_LEN);	841	while (cur < buf + SHORT_STRING_LEN);
667		842
		843	{
668	STRLEN len = cur - buf;	844	STRLEN len = cur - buf;
669		845
670	if (sv)	846	if (sv)
671	{	847	{
672	SvGROW (sv, SvCUR (sv) + len + 1);	848	SvGROW (sv, SvCUR (sv) + len + 1);
673	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	849	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
674	SvCUR_set (sv, SvCUR (sv) + len);	850	SvCUR_set (sv, SvCUR (sv) + len);
675	}	851	}
676	else	852	else
677	sv = newSVpvn (buf, len);	853	sv = newSVpvn (buf, len);
678	}	854	}
		855	}
679	while (*dec->cur != '"');	856	while (*dec_cur != '"');
680		857
681	++dec->cur;	858	++dec_cur;
682		859
683	if (sv)	860	if (sv)
684	{	861	{
685	SvPOK_only (sv);	862	SvPOK_only (sv);
686	*SvEND (sv) = 0;	863	*SvEND (sv) = 0;
…		…
689	SvUTF8_on (sv);	866	SvUTF8_on (sv);
690	}	867	}
691	else	868	else
692	sv = newSVpvn ("", 0);	869	sv = newSVpvn ("", 0);
693		870
		871	dec->cur = dec_cur;
694	return sv;	872	return sv;
695		873
696	fail:	874	fail:
		875	dec->cur = dec_cur;
697	return 0;	876	return 0;
698	}	877	}
699		878
700	static SV *	879	static SV *
701	decode_num (dec_t *dec)	880	decode_num (dec_t *dec)
…		…
759	is_nv = 1;	938	is_nv = 1;
760	}	939	}
761		940
762	if (!is_nv)	941	if (!is_nv)
763	{	942	{
764	UV uv;	943	int len = dec->cur - start;
765	int numtype = grok_number (start, dec->cur - start, &uv);	944
766	if (numtype & IS_NUMBER_IN_UV)	945	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
767	if (numtype & IS_NUMBER_NEG)	946	if (*start == '-')
		947	switch (len)
768	{	948	{
769	if (uv < (UV)IV_MIN)	949	case 2: return newSViv (-( start [1] - '0' * 1));
770	return newSViv (-(IV)uv);	950	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		951	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		952	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
771	}	953	}
		954	else
		955	switch (len)
		956	{
		957	case 1: return newSViv ( start [0] - '0' * 1);
		958	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		959	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		960	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		961	}
		962
		963	{
		964	UV uv;
		965	int numtype = grok_number (start, len, &uv);
		966	if (numtype & IS_NUMBER_IN_UV)
		967	if (numtype & IS_NUMBER_NEG)
		968	{
		969	if (uv < (UV)IV_MIN)
		970	return newSViv (-(IV)uv);
		971	}
772	else	972	else
773	return newSVuv (uv);	973	return newSVuv (uv);
774	}	974	}
775		975
		976	len -= *start == '-' ? 1 : 0;
		977
		978	// does not fit into IV or UV, try NV
		979	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		980	#if defined (LDBL_DIG)
		981	\|\| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		982	#endif
		983	)
		984	// fits into NV without loss of precision
		985	return newSVnv (Atof (start));
		986
		987	// everything else fails, convert it to a string
		988	return newSVpvn (start, dec->cur - start);
		989	}
		990
		991	// loss of precision here
776	return newSVnv (Atof (start));	992	return newSVnv (Atof (start));
777		993
778	fail:	994	fail:
779	return 0;	995	return 0;
780	}	996	}
…		…
782	static SV *	998	static SV *
783	decode_av (dec_t *dec)	999	decode_av (dec_t *dec)
784	{	1000	{
785	AV *av = newAV ();	1001	AV *av = newAV ();
786		1002
		1003	DEC_INC_DEPTH;
787	decode_ws (dec);	1004	decode_ws (dec);
		1005
788	if (*dec->cur == ']')	1006	if (*dec->cur == ']')
789	++dec->cur;	1007	++dec->cur;
790	else	1008	else
791	for (;;)	1009	for (;;)
792	{	1010	{
…		…
810	ERR (", or ] expected while parsing array");	1028	ERR (", or ] expected while parsing array");
811		1029
812	++dec->cur;	1030	++dec->cur;
813	}	1031	}
814		1032
		1033	DEC_DEC_DEPTH;
815	return newRV_noinc ((SV *)av);	1034	return newRV_noinc ((SV *)av);
816		1035
817	fail:	1036	fail:
818	SvREFCNT_dec (av);	1037	SvREFCNT_dec (av);
		1038	DEC_DEC_DEPTH;
819	return 0;	1039	return 0;
820	}	1040	}
821		1041
822	static SV *	1042	static SV *
823	decode_hv (dec_t *dec)	1043	decode_hv (dec_t *dec)
824	{	1044	{
		1045	SV *sv;
825	HV *hv = newHV ();	1046	HV *hv = newHV ();
826		1047
		1048	DEC_INC_DEPTH;
827	decode_ws (dec);	1049	decode_ws (dec);
		1050
828	if (*dec->cur == '}')	1051	if (*dec->cur == '}')
829	++dec->cur;	1052	++dec->cur;
830	else	1053	else
831	for (;;)	1054	for (;;)
832	{	1055	{
833	SV key, value;
834
835	decode_ws (dec); EXPECT_CH ('"');	1056	decode_ws (dec); EXPECT_CH ('"');
836		1057
837	key = decode_str (dec);	1058	// heuristic: assume that
838	if (!key)	1059	// a) decode_str + hv_store_ent are abysmally slow.
839	goto fail;	1060	// b) most hash keys are short, simple ascii text.
		1061	// => try to "fast-match" such strings to avoid
		1062	// the overhead of decode_str + hv_store_ent.
		1063	{
		1064	SV *value;
		1065	char *p = dec->cur;
		1066	char *e = p + 24; // only try up to 24 bytes
840		1067
841	decode_ws (dec); EXPECT_CH (':');	1068	for (;;)
842
843	value = decode_sv (dec);
844	if (!value)
845	{	1069	{
		1070	// the >= 0x80 is true on most architectures
		1071	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
		1072	{
		1073	// slow path, back up and use decode_str
		1074	SV *key = decode_str (dec);
		1075	if (!key)
		1076	goto fail;
		1077
		1078	decode_ws (dec); EXPECT_CH (':');
		1079
		1080	value = decode_sv (dec);
		1081	if (!value)
		1082	{
		1083	SvREFCNT_dec (key);
		1084	goto fail;
		1085	}
		1086
		1087	hv_store_ent (hv, key, value, 0);
846	SvREFCNT_dec (key);	1088	SvREFCNT_dec (key);
		1089
		1090	break;
		1091	}
		1092	else if (*p == '"')
		1093	{
		1094	// fast path, got a simple key
		1095	char *key = dec->cur;
		1096	int len = p - key;
		1097	dec->cur = p + 1;
		1098
		1099	decode_ws (dec); EXPECT_CH (':');
		1100
		1101	value = decode_sv (dec);
		1102	if (!value)
847	goto fail;	1103	goto fail;
		1104
		1105	hv_store (hv, key, len, value, 0);
		1106
		1107	break;
		1108	}
		1109
		1110	++p;
848	}	1111	}
849		1112	}
850	//TODO: optimise
851	hv_store_ent (hv, key, value, 0);
852		1113
853	decode_ws (dec);	1114	decode_ws (dec);
854		1115
855	if (*dec->cur == '}')	1116	if (*dec->cur == '}')
856	{	1117	{
…		…
862	ERR (", or } expected while parsing object/hash");	1123	ERR (", or } expected while parsing object/hash");
863		1124
864	++dec->cur;	1125	++dec->cur;
865	}	1126	}
866		1127
		1128	DEC_DEC_DEPTH;
867	return newRV_noinc ((SV *)hv);	1129	sv = newRV_noinc ((SV *)hv);
		1130
		1131	// check filter callbacks
		1132	if (dec->json.flags & F_HOOK)
		1133	{
		1134	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1135	{
		1136	HE cb, he;
		1137
		1138	hv_iterinit (hv);
		1139	he = hv_iternext (hv);
		1140	hv_iterinit (hv);
		1141
		1142	// the next line creates a mortal sv each time its called.
		1143	// might want to optimise this for common cases.
		1144	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1145
		1146	if (cb)
		1147	{
		1148	dSP;
		1149	int count;
		1150
		1151	ENTER; SAVETMPS; PUSHMARK (SP);
		1152	XPUSHs (HeVAL (he));
		1153
		1154	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1155
		1156	if (count == 1)
		1157	{
		1158	sv = newSVsv (POPs);
		1159	FREETMPS; LEAVE;
		1160	return sv;
		1161	}
		1162
		1163	FREETMPS; LEAVE;
		1164	}
		1165	}
		1166
		1167	if (dec->json.cb_object)
		1168	{
		1169	dSP;
		1170	int count;
		1171
		1172	ENTER; SAVETMPS; PUSHMARK (SP);
		1173	XPUSHs (sv_2mortal (sv));
		1174
		1175	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1176
		1177	if (count == 1)
		1178	{
		1179	sv = newSVsv (POPs);
		1180	FREETMPS; LEAVE;
		1181	return sv;
		1182	}
		1183
		1184	SvREFCNT_inc (sv);
		1185	FREETMPS; LEAVE;
		1186	}
		1187	}
		1188
		1189	return sv;
868		1190
869	fail:	1191	fail:
870	SvREFCNT_dec (hv);	1192	SvREFCNT_dec (hv);
		1193	DEC_DEC_DEPTH;
871	return 0;	1194	return 0;
872	}	1195	}
873		1196
874	static SV *	1197	static SV *
875	decode_sv (dec_t *dec)	1198	decode_sv (dec_t *dec)
876	{	1199	{
877	decode_ws (dec);	1200	decode_ws (dec);
		1201
		1202	// the beauty of JSON: you need exactly one character lookahead
		1203	// to parse anything.
878	switch (*dec->cur)	1204	switch (*dec->cur)
879	{	1205	{
880	case '"': ++dec->cur; return decode_str (dec);	1206	case '"': ++dec->cur; return decode_str (dec);
881	case '[': ++dec->cur; return decode_av (dec);	1207	case '[': ++dec->cur; return decode_av (dec);
882	case '{': ++dec->cur; return decode_hv (dec);	1208	case '{': ++dec->cur; return decode_hv (dec);
…		…
888		1214
889	case 't':	1215	case 't':
890	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1216	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
891	{	1217	{
892	dec->cur += 4;	1218	dec->cur += 4;
893	return newSViv (1);	1219	return SvREFCNT_inc (json_true);
894	}	1220	}
895	else	1221	else
896	ERR ("'true' expected");	1222	ERR ("'true' expected");
897		1223
898	break;	1224	break;
899		1225
900	case 'f':	1226	case 'f':
901	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1227	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
902	{	1228	{
903	dec->cur += 5;	1229	dec->cur += 5;
904	return newSViv (0);	1230	return SvREFCNT_inc (json_false);
905	}	1231	}
906	else	1232	else
907	ERR ("'false' expected");	1233	ERR ("'false' expected");
908		1234
909	break;	1235	break;
…		…
918	ERR ("'null' expected");	1244	ERR ("'null' expected");
919		1245
920	break;	1246	break;
921		1247
922	default:	1248	default:
923	ERR ("malformed json string, neither array, object, number, string or atom");	1249	ERR ("malformed JSON string, neither array, object, number, string or atom");
924	break;	1250	break;
925	}	1251	}
926		1252
927	fail:	1253	fail:
928	return 0;	1254	return 0;
929	}	1255	}
930		1256
931	static SV *	1257	static SV *
932	decode_json (SV *string, UV flags)	1258	decode_json (SV string, JSON json, UV *offset_return)
933	{	1259	{
		1260	dec_t dec;
		1261	UV offset;
934	SV *sv;	1262	SV *sv;
935		1263
		1264	SvGETMAGIC (string);
		1265	SvUPGRADE (string, SVt_PV);
		1266
		1267	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1268	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1269	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1270
936	if (flags & F_UTF8)	1271	if (json->flags & F_UTF8)
937	sv_utf8_downgrade (string, 0);	1272	sv_utf8_downgrade (string, 0);
938	else	1273	else
939	sv_utf8_upgrade (string);	1274	sv_utf8_upgrade (string);
940		1275
941	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1276	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
942		1277
943	dec_t dec;	1278	dec.json = *json;
944	dec.flags = flags;
945	dec.cur = SvPVX (string);	1279	dec.cur = SvPVX (string);
946	dec.end = SvEND (string);	1280	dec.end = SvEND (string);
947	dec.err = 0;	1281	dec.err = 0;
		1282	dec.depth = 0;
		1283	dec.maxdepth = DEC_DEPTH (dec.json.flags);
948		1284
		1285	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
		1286	dec.json.flags \|= F_HOOK;
		1287
		1288	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
949	sv = decode_sv (&dec);	1289	sv = decode_sv (&dec);
950		1290
		1291	if (!(offset_return \|\| !sv))
		1292	{
		1293	// check for trailing garbage
		1294	decode_ws (&dec);
		1295
		1296	if (*dec.cur)
		1297	{
		1298	dec.err = "garbage after JSON object";
		1299	SvREFCNT_dec (sv);
		1300	sv = 0;
		1301	}
		1302	}
		1303
		1304	if (offset_return \|\| !sv)
		1305	{
		1306	offset = dec.json.flags & F_UTF8
		1307	? dec.cur - SvPVX (string)
		1308	: utf8_distance (dec.cur, SvPVX (string));
		1309
		1310	if (offset_return)
		1311	*offset_return = offset;
		1312	}
		1313
951	if (!sv)	1314	if (!sv)
952	{	1315	{
953	IV offset = dec.flags & F_UTF8
954	? dec.cur - SvPVX (string)
955	: utf8_distance (dec.cur, SvPVX (string));
956	SV *uni = sv_newmortal ();	1316	SV *uni = sv_newmortal ();
957		1317
958	// horrible hack to silence warning inside pv_uni_display	1318	// horrible hack to silence warning inside pv_uni_display
959	COP cop = *PL_curcop;	1319	COP cop = *PL_curcop;
960	cop.cop_warnings = pWARN_NONE;	1320	cop.cop_warnings = pWARN_NONE;
…		…
962	SAVEVPTR (PL_curcop);	1322	SAVEVPTR (PL_curcop);
963	PL_curcop = &cop;	1323	PL_curcop = &cop;
964	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1324	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
965	LEAVE;	1325	LEAVE;
966		1326
967	croak ("%s, at character offset %d (%s)",	1327	croak ("%s, at character offset %d [\"%s\"]",
968	dec.err,	1328	dec.err,
969	(int)offset,	1329	(int)offset,
970	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1330	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
971	}	1331	}
972		1332
973	sv = sv_2mortal (sv);	1333	sv = sv_2mortal (sv);
974		1334
975	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1335	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
976	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1336	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
977		1337
978	return sv;	1338	return sv;
979	}	1339	}
980		1340
…		…
985		1345
986	BOOT:	1346	BOOT:
987	{	1347	{
988	int i;	1348	int i;
989		1349
990	memset (decode_hexdigit, 0xff, 256);
991	for (i = 10; i--; )	1350	for (i = 0; i < 256; ++i)
992	decode_hexdigit ['0' + i] = i;	1351	decode_hexdigit [i] =
		1352	i >= '0' && i <= '9' ? i - '0'
		1353	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1354	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1355	: -1;
993		1356
994	for (i = 7; i--; )	1357	json_stash = gv_stashpv ("JSON::XS" , 1);
		1358	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1359
		1360	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1361	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1362	}
		1363
		1364	PROTOTYPES: DISABLE
		1365
		1366	void new (char *klass)
		1367	PPCODE:
		1368	{
		1369	SV *pv = NEWSV (0, sizeof (JSON));
		1370	SvPOK_only (pv);
		1371	Zero (SvPVX (pv), 1, JSON);
		1372	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1373	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash)));
		1374	}
		1375
		1376	void ascii (JSON *self, int enable = 1)
		1377	ALIAS:
		1378	ascii = F_ASCII
		1379	latin1 = F_LATIN1
		1380	utf8 = F_UTF8
		1381	indent = F_INDENT
		1382	canonical = F_CANONICAL
		1383	space_before = F_SPACE_BEFORE
		1384	space_after = F_SPACE_AFTER
		1385	pretty = F_PRETTY
		1386	allow_nonref = F_ALLOW_NONREF
		1387	shrink = F_SHRINK
		1388	allow_blessed = F_ALLOW_BLESSED
		1389	convert_blessed = F_CONV_BLESSED
		1390	PPCODE:
		1391	{
		1392	if (enable)
		1393	self->flags \|= ix;
		1394	else
		1395	self->flags &= ~ix;
		1396
		1397	XPUSHs (ST (0));
		1398	}
		1399
		1400	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1401	PPCODE:
		1402	{
		1403	UV log2 = 0;
		1404
		1405	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1406
		1407	while ((1UL << log2) < max_depth)
		1408	++log2;
		1409
		1410	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
		1411
		1412	XPUSHs (ST (0));
		1413	}
		1414
		1415	void max_size (JSON *self, UV max_size = 0)
		1416	PPCODE:
		1417	{
		1418	UV log2 = 0;
		1419
		1420	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1421	if (max_size == 1) max_size = 2;
		1422
		1423	while ((1UL << log2) < max_size)
		1424	++log2;
		1425
		1426	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
		1427
		1428	XPUSHs (ST (0));
		1429	}
		1430
		1431	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
		1432	PPCODE:
		1433	{
		1434	SvREFCNT_dec (self->cb_object);
		1435	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1436
		1437	XPUSHs (ST (0));
		1438	}
		1439
		1440	void filter_json_single_key_object (JSON self, SV key, SV *cb = &PL_sv_undef)
		1441	PPCODE:
		1442	{
		1443	if (!self->cb_sk_object)
		1444	self->cb_sk_object = newHV ();
		1445
		1446	if (SvOK (cb))
		1447	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1448	else
995	{	1449	{
996	decode_hexdigit ['a' + i] = 10 + i;	1450	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
997	decode_hexdigit ['A' + i] = 10 + i;	1451
		1452	if (!HvKEYS (self->cb_sk_object))
		1453	{
		1454	SvREFCNT_dec (self->cb_sk_object);
		1455	self->cb_sk_object = 0;
		1456	}
998	}	1457	}
999		1458
1000	json_stash = gv_stashpv ("JSON::XS", 1);	1459	XPUSHs (ST (0));
1001	}	1460	}
1002		1461
1003	PROTOTYPES: DISABLE	1462	void encode (JSON self, SV scalar)
		1463	PPCODE:
		1464	XPUSHs (encode_json (scalar, self));
1004		1465
1005	SV new (char dummy)	1466	void decode (JSON self, SV jsonstr)
		1467	PPCODE:
		1468	XPUSHs (decode_json (jsonstr, self, 0));
		1469
		1470	void decode_prefix (JSON self, SV jsonstr)
		1471	PPCODE:
		1472	{
		1473	UV offset;
		1474	EXTEND (SP, 2);
		1475	PUSHs (decode_json (jsonstr, self, &offset));
		1476	PUSHs (sv_2mortal (newSVuv (offset)));
		1477	}
		1478
		1479	void DESTROY (JSON *self)
1006	CODE:	1480	CODE:
1007	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1481	SvREFCNT_dec (self->cb_sk_object);
1008	OUTPUT:	1482	SvREFCNT_dec (self->cb_object);
1009	RETVAL
1010		1483
1011	SV ascii (SV self, int enable = 1)	1484	PROTOTYPES: ENABLE
1012	ALIAS:
1013	ascii = F_ASCII
1014	utf8 = F_UTF8
1015	indent = F_INDENT
1016	canonical = F_CANONICAL
1017	space_before = F_SPACE_BEFORE
1018	space_after = F_SPACE_AFTER
1019	pretty = F_PRETTY
1020	allow_nonref = F_ALLOW_NONREF
1021	shrink = F_SHRINK
1022	CODE:
1023	{
1024	UV *uv = SvJSON (self);
1025	if (enable)
1026	*uv \|= ix;
1027	else
1028	*uv &= ~ix;
1029		1485
1030	RETVAL = newSVsv (self);	1486	void to_json (SV *scalar)
1031	}
1032	OUTPUT:
1033	RETVAL
1034
1035	void encode (SV self, SV scalar)
1036	PPCODE:	1487	PPCODE:
		1488	{
		1489	JSON json = { F_DEFAULT \| F_UTF8 };
1037	XPUSHs (encode_json (scalar, *SvJSON (self)));	1490	XPUSHs (encode_json (scalar, &json));
		1491	}
1038		1492
1039	void decode (SV self, SV jsonstr)	1493	void from_json (SV *jsonstr)
1040	PPCODE:	1494	PPCODE:
1041	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1495	{
1042		1496	JSON json = { F_DEFAULT \| F_UTF8 };
1043	PROTOTYPES: ENABLE
1044
1045	void to_json (SV *scalar)
1046	ALIAS:
1047	objToJson = 0
1048	PPCODE:
1049	XPUSHs (encode_json (scalar, F_UTF8));
1050
1051	void from_json (SV *jsonstr)
1052	ALIAS:
1053	jsonToObj = 0
1054	PPCODE:
1055	XPUSHs (decode_json (jsonstr, F_UTF8));	1497	XPUSHs (decode_json (jsonstr, &json, 0));
		1498	}
1056		1499

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.16 by root, Sun Mar 25 02:37:00 2007 UTC vs. Revision 1.56 by root, Thu Jul 26 11:33:35 2007 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.16 by root, Sun Mar 25 02:37:00 2007 UTC vs.
Revision 1.56 by root, Thu Jul 26 11:33:35 2007 UTC