[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 1.60 by root, Mon Aug 13 16:19:13 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
		63	#ifdef USE_ITHREADS
		64	# define JSON_SLOW 1
		65	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		66	#else
		67	# define JSON_SLOW 0
		68	# define JSON_STASH json_stash
		69	#endif
		70
34	static HV *json_stash; // JSON::XS::	71	static HV json_stash, json_boolean_stash; // JSON::XS::
		72	static SV json_true, json_false;
		73
		74	typedef struct {
		75	U32 flags;
		76	SV *cb_object;
		77	HV *cb_sk_object;
		78	} JSON;
35		79
36	/////////////////////////////////////////////////////////////////////////////	80	/////////////////////////////////////////////////////////////////////////////
37	// utility functions	81	// utility functions
38		82
39	static UV *	83	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)	84	shrink (SV *sv)
50	{	85	{
51	sv_utf8_downgrade (sv, 1);	86	sv_utf8_downgrade (sv, 1);
52	if (SvLEN (sv) > SvCUR (sv) + 1)	87	if (SvLEN (sv) > SvCUR (sv) + 1)
53	{	88	{
62	// decode an utf-8 character and return it, or (UV)-1 in	97	// decode an utf-8 character and return it, or (UV)-1 in
63	// case of an error.	98	// case of an error.
64	// we special-case "safe" characters from U+80 .. U+7FF,	99	// we special-case "safe" characters from U+80 .. U+7FF,
65	// but use the very good perl function to parse anything else.	100	// but use the very good perl function to parse anything else.
66	// note that we never call this function for a ascii codepoints	101	// note that we never call this function for a ascii codepoints
67	static UV	102	inline UV
68	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	103	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
69	{	104	{
70	if (s[0] > 0xdf \|\| s[0] < 0xc2)	105	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))
71	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	106	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
72	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	107	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
73	{	108	{
74	*clen = 2;	109	*clen = 2;
75	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);	110	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
76	}	111	}
77	else	112	else
		113	{
		114	*clen = (STRLEN)-1;
78	return (UV)-1;	115	return (UV)-1;
		116	}
79	}	117	}
80		118
81	/////////////////////////////////////////////////////////////////////////////	119	/////////////////////////////////////////////////////////////////////////////
82	// encoder	120	// encoder
83		121
…		…
85	typedef struct	123	typedef struct
86	{	124	{
87	char *cur; // SvPVX (sv) + current output position	125	char *cur; // SvPVX (sv) + current output position
88	char *end; // SvEND (sv)	126	char *end; // SvEND (sv)
89	SV *sv; // result scalar	127	SV *sv; // result scalar
90	UV flags; // F_*	128	JSON json;
91	int indent; // indentation level	129	U32 indent; // indentation level
92	int max_depth; // max. recursion level	130	U32 maxdepth; // max. indentation/recursion level
93	} enc_t;	131	} enc_t;
94		132
95	static void	133	inline void
96	need (enc_t *enc, STRLEN len)	134	need (enc_t *enc, STRLEN len)
97	{	135	{
98	if (enc->cur + len >= enc->end)	136	if (expect_false (enc->cur + len >= enc->end))
99	{	137	{
100	STRLEN cur = enc->cur - SvPVX (enc->sv);	138	STRLEN cur = enc->cur - SvPVX (enc->sv);
101	SvGROW (enc->sv, cur + len + 1);	139	SvGROW (enc->sv, cur + len + 1);
102	enc->cur = SvPVX (enc->sv) + cur;	140	enc->cur = SvPVX (enc->sv) + cur;
103	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	141	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
104	}	142	}
105	}	143	}
106		144
107	static void	145	inline void
108	encode_ch (enc_t *enc, char ch)	146	encode_ch (enc_t *enc, char ch)
109	{	147	{
110	need (enc, 1);	148	need (enc, 1);
111	*enc->cur++ = ch;	149	*enc->cur++ = ch;
112	}	150	}
…		…
120		158
121	while (str < end)	159	while (str < end)
122	{	160	{
123	unsigned char ch = (unsigned char )str;	161	unsigned char ch = (unsigned char )str;
124		162
125	if (ch >= 0x20 && ch < 0x80) // most common case	163	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
126	{	164	{
127	if (ch == '"') // but with slow exceptions	165	if (expect_false (ch == '"')) // but with slow exceptions
128	{	166	{
129	need (enc, len += 1);	167	need (enc, len += 1);
130	*enc->cur++ = '\\';	168	*enc->cur++ = '\\';
131	*enc->cur++ = '"';	169	*enc->cur++ = '"';
132	}	170	}
133	else if (ch == '\\')	171	else if (expect_false (ch == '\\'))
134	{	172	{
135	need (enc, len += 1);	173	need (enc, len += 1);
136	*enc->cur++ = '\\';	174	*enc->cur++ = '\\';
137	*enc->cur++ = '\\';	175	*enc->cur++ = '\\';
138	}	176	}
…		…
156	STRLEN clen;	194	STRLEN clen;
157	UV uch;	195	UV uch;
158		196
159	if (is_utf8)	197	if (is_utf8)
160	{	198	{
161	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
162	uch = decode_utf8 (str, end - str, &clen);	199	uch = decode_utf8 (str, end - str, &clen);
163	if (clen == (STRLEN)-1)	200	if (clen == (STRLEN)-1)
164	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	201	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
165	}	202	}
166	else	203	else
…		…
170	}	207	}
171		208
172	if (uch > 0x10FFFFUL)	209	if (uch > 0x10FFFFUL)
173	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);	210	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
174		211
175	if (uch < 0x80 \|\| enc->flags & F_ASCII)	212	if (uch < 0x80 \|\| enc->json.flags & F_ASCII \|\| (enc->json.flags & F_LATIN1 && uch > 0xFF))
176	{	213	{
177	if (uch > 0xFFFFUL)	214	if (uch > 0xFFFFUL)
178	{	215	{
179	need (enc, len += 11);	216	need (enc, len += 11);
180	sprintf (enc->cur, "\\u%04x\\u%04x",	217	sprintf (enc->cur, "\\u%04x\\u%04x",
…		…
194	*enc->cur++ = hexdigit [(uch >> 0) & 15];	231	*enc->cur++ = hexdigit [(uch >> 0) & 15];
195	}	232	}
196		233
197	str += clen;	234	str += clen;
198	}	235	}
		236	else if (enc->json.flags & F_LATIN1)
		237	{
		238	*enc->cur++ = uch;
		239	str += clen;
		240	}
199	else if (is_utf8)	241	else if (is_utf8)
200	{	242	{
201	need (enc, len += clen);	243	need (enc, len += clen);
202	do	244	do
203	{	245	{
…		…
205	}	247	}
206	while (--clen);	248	while (--clen);
207	}	249	}
208	else	250	else
209	{	251	{
210	need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed	252	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
211	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	253	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
212	++str;	254	++str;
213	}	255	}
214	}	256	}
215	}	257	}
…		…
217		259
218	--len;	260	--len;
219	}	261	}
220	}	262	}
221		263
222	static void	264	inline void
223	encode_indent (enc_t *enc)	265	encode_indent (enc_t *enc)
224	{	266	{
225	if (enc->flags & F_INDENT)	267	if (enc->json.flags & F_INDENT)
226	{	268	{
227	int spaces = enc->indent * INDENT_STEP;	269	int spaces = enc->indent * INDENT_STEP;
228		270
229	need (enc, spaces);	271	need (enc, spaces);
230	memset (enc->cur, ' ', spaces);	272	memset (enc->cur, ' ', spaces);
231	enc->cur += spaces;	273	enc->cur += spaces;
232	}	274	}
233	}	275	}
234		276
235	static void	277	inline void
236	encode_space (enc_t *enc)	278	encode_space (enc_t *enc)
237	{	279	{
238	need (enc, 1);	280	need (enc, 1);
239	encode_ch (enc, ' ');	281	encode_ch (enc, ' ');
240	}	282	}
241		283
242	static void	284	inline void
243	encode_nl (enc_t *enc)	285	encode_nl (enc_t *enc)
244	{	286	{
245	if (enc->flags & F_INDENT)	287	if (enc->json.flags & F_INDENT)
246	{	288	{
247	need (enc, 1);	289	need (enc, 1);
248	encode_ch (enc, '\n');	290	encode_ch (enc, '\n');
249	}	291	}
250	}	292	}
251		293
252	static void	294	inline void
253	encode_comma (enc_t *enc)	295	encode_comma (enc_t *enc)
254	{	296	{
255	encode_ch (enc, ',');	297	encode_ch (enc, ',');
256		298
257	if (enc->flags & F_INDENT)	299	if (enc->json.flags & F_INDENT)
258	encode_nl (enc);	300	encode_nl (enc);
259	else if (enc->flags & F_SPACE_AFTER)	301	else if (enc->json.flags & F_SPACE_AFTER)
260	encode_space (enc);	302	encode_space (enc);
261	}	303	}
262		304
263	static void encode_sv (enc_t enc, SV sv);	305	static void encode_sv (enc_t enc, SV sv);
264		306
265	static void	307	static void
266	encode_av (enc_t enc, AV av)	308	encode_av (enc_t enc, AV av)
267	{	309	{
268	int i, len = av_len (av);	310	int i, len = av_len (av);
269		311
		312	if (enc->indent >= enc->maxdepth)
		313	croak ("data structure too deep (hit recursion limit)");
		314
270	encode_ch (enc, '['); encode_nl (enc);	315	encode_ch (enc, '['); encode_nl (enc);
271	++enc->indent;	316	++enc->indent;
272		317
273	for (i = 0; i <= len; ++i)	318	for (i = 0; i <= len; ++i)
274	{	319	{
		320	SV **svp = av_fetch (av, i, 0);
		321
275	encode_indent (enc);	322	encode_indent (enc);
276	encode_sv (enc, *av_fetch (av, i, 0));	323
		324	if (svp)
		325	encode_sv (enc, *svp);
		326	else
		327	encode_str (enc, "null", 4, 0);
277		328
278	if (i < len)	329	if (i < len)
279	encode_comma (enc);	330	encode_comma (enc);
280	}	331	}
281		332
…		…
304	else	355	else
305	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	356	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
306		357
307	encode_ch (enc, '"');	358	encode_ch (enc, '"');
308		359
309	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	360	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
310	encode_ch (enc, ':');	361	encode_ch (enc, ':');
311	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	362	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
312	encode_sv (enc, HeVAL (he));	363	encode_sv (enc, HeVAL (he));
313	}	364	}
314		365
315	// compare hash entries, used when all keys are bytestrings	366	// compare hash entries, used when all keys are bytestrings
316	static int	367	static int
…		…
339		390
340	static void	391	static void
341	encode_hv (enc_t enc, HV hv)	392	encode_hv (enc_t enc, HV hv)
342	{	393	{
343	int count, i;	394	int count, i;
		395
		396	if (enc->indent >= enc->maxdepth)
		397	croak ("data structure too deep (hit recursion limit)");
344		398
345	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	399	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
346		400
347	if ((count = hv_iterinit (hv)))	401	if ((count = hv_iterinit (hv)))
348	{	402	{
349	// for canonical output we have to sort by keys first	403	// for canonical output we have to sort by keys first
350	// actually, this is mostly due to the stupid so-called	404	// actually, this is mostly due to the stupid so-called
351	// security workaround added somewhere in 5.8.x.	405	// security workaround added somewhere in 5.8.x.
352	// that randomises hash orderings	406	// that randomises hash orderings
353	if (enc->flags & F_CANONICAL)	407	if (enc->json.flags & F_CANONICAL)
354	{	408	{
355	HE he, hes [count]; // if your compiler dies here, you need to enable C99 mode
356	int fast = 1;	409	int fast = 1;
		410	HE *he;
		411	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
		412	HE *hes = _alloca (count sizeof (HE));
		413	#else
		414	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		415	#endif
357		416
358	i = 0;	417	i = 0;
359	while ((he = hv_iternext (hv)))	418	while ((he = hv_iternext (hv)))
360	{	419	{
361	hes [i++] = he;	420	hes [i++] = he;
…		…
396		455
397	encode_nl (enc);	456	encode_nl (enc);
398	}	457	}
399	else	458	else
400	{	459	{
401	SV *sv;
402	HE *he = hv_iternext (hv);	460	HE *he = hv_iternext (hv);
403		461
404	for (;;)	462	for (;;)
405	{	463	{
406	encode_indent (enc);	464	encode_indent (enc);
…		…
415	encode_nl (enc);	473	encode_nl (enc);
416	}	474	}
417	}	475	}
418		476
419	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	477	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
		478	}
		479
		480	// encode objects, arrays and special \0=false and \1=true values.
		481	static void
		482	encode_rv (enc_t enc, SV sv)
		483	{
		484	svtype svt;
		485
		486	SvGETMAGIC (sv);
		487	svt = SvTYPE (sv);
		488
		489	if (expect_false (SvOBJECT (sv)))
		490	{
		491	HV *stash = !JSON_SLOW \|\| json_boolean_stash
		492	? json_boolean_stash
		493	: gv_stashpv ("JSON::XS::Boolean", 1);
		494
		495	if (SvSTASH (sv) == stash)
		496	{
		497	if (SvIV (sv))
		498	encode_str (enc, "true", 4, 0);
		499	else
		500	encode_str (enc, "false", 5, 0);
		501	}
		502	else
		503	{
		504	#if 0
		505	if (0 && sv_derived_from (rv, "JSON::Literal"))
		506	{
		507	// not yet
		508	}
		509	#endif
		510	if (enc->json.flags & F_CONV_BLESSED)
		511	{
		512	// we re-bless the reference to get overload and other niceties right
		513	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
		514
		515	if (to_json)
		516	{
		517	dSP;
		518
		519	ENTER; SAVETMPS; PUSHMARK (SP);
		520	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		521
		522	// calling with G_SCALAR ensures that we always get a 1 return value
		523	PUTBACK;
		524	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		525	SPAGAIN;
		526
		527	// catch this surprisingly common error
		528	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		529	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		530
		531	sv = POPs;
		532	PUTBACK;
		533
		534	encode_sv (enc, sv);
		535
		536	FREETMPS; LEAVE;
		537	}
		538	else if (enc->json.flags & F_ALLOW_BLESSED)
		539	encode_str (enc, "null", 4, 0);
		540	else
		541	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		542	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		543	}
		544	else if (enc->json.flags & F_ALLOW_BLESSED)
		545	encode_str (enc, "null", 4, 0);
		546	else
		547	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		548	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		549	}
		550	}
		551	else if (svt == SVt_PVHV)
		552	encode_hv (enc, (HV *)sv);
		553	else if (svt == SVt_PVAV)
		554	encode_av (enc, (AV *)sv);
		555	else if (svt < SVt_PVAV)
		556	{
		557	STRLEN len = 0;
		558	char *pv = svt ? SvPV (sv, len) : 0;
		559
		560	if (len == 1 && *pv == '1')
		561	encode_str (enc, "true", 4, 0);
		562	else if (len == 1 && *pv == '0')
		563	encode_str (enc, "false", 5, 0);
		564	else
		565	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		566	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		567	}
		568	else
		569	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		570	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
420	}	571	}
421		572
422	static void	573	static void
423	encode_sv (enc_t enc, SV sv)	574	encode_sv (enc_t enc, SV sv)
424	{	575	{
…		…
432	encode_str (enc, str, len, SvUTF8 (sv));	583	encode_str (enc, str, len, SvUTF8 (sv));
433	encode_ch (enc, '"');	584	encode_ch (enc, '"');
434	}	585	}
435	else if (SvNOKp (sv))	586	else if (SvNOKp (sv))
436	{	587	{
		588	// trust that perl will do the right thing w.r.t. JSON syntax.
437	need (enc, NV_DIG + 32);	589	need (enc, NV_DIG + 32);
438	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	590	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
439	enc->cur += strlen (enc->cur);	591	enc->cur += strlen (enc->cur);
440	}	592	}
441	else if (SvIOKp (sv))	593	else if (SvIOKp (sv))
442	{	594	{
443	need (enc, 64);	595	// we assume we can always read an IV as a UV
		596	if (SvUV (sv) & ~(UV)0x7fff)
		597	{
		598	// large integer, use the (rather slow) snprintf way.
		599	need (enc, sizeof (UV) * 3);
444	enc->cur +=	600	enc->cur +=
445	SvIsUV(sv)	601	SvIsUV(sv)
446	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	602	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
447	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	603	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		604	}
		605	else
		606	{
		607	// optimise the "small number case"
		608	// code will likely be branchless and use only a single multiplication
		609	I32 i = SvIV (sv);
		610	U32 u;
		611	char digit, nz = 0;
		612
		613	need (enc, 6);
		614
		615	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		616	u = i < 0 ? -i : i;
		617
		618	// convert to 4.28 fixed-point representation
		619	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		620
		621	// now output digit by digit, each time masking out the integer part
		622	// and multiplying by 5 while moving the decimal point one to the right,
		623	// resulting in a net multiplication by 10.
		624	// we always write the digit to memory but conditionally increment
		625	// the pointer, to ease the usage of conditional move instructions.
		626	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffff) 5;
		627	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffff) 5;
		628	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffff) 5;
		629	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffff) 5;
		630	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		631	}
448	}	632	}
449	else if (SvROK (sv))	633	else if (SvROK (sv))
450	{	634	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))	635	else if (!SvOK (sv))
467	encode_str (enc, "null", 4, 0);	636	encode_str (enc, "null", 4, 0);
468	else	637	else
469	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	638	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
470	SvPV_nolen (sv), SvFLAGS (sv));	639	SvPV_nolen (sv), SvFLAGS (sv));
471	}	640	}
472		641
473	static SV *	642	static SV *
474	encode_json (SV *scalar, UV flags)	643	encode_json (SV scalar, JSON json)
475	{	644	{
		645	enc_t enc;
		646
476	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	647	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
477	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	648	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
478		649
479	enc_t enc;	650	enc.json = *json;
480	enc.flags = flags;
481	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	651	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
482	enc.cur = SvPVX (enc.sv);	652	enc.cur = SvPVX (enc.sv);
483	enc.end = SvEND (enc.sv);	653	enc.end = SvEND (enc.sv);
484	enc.indent = 0;	654	enc.indent = 0;
485	enc.max_depth = 0x7fffffffUL;	655	enc.maxdepth = DEC_DEPTH (enc.json.flags);
486		656
487	SvPOK_only (enc.sv);	657	SvPOK_only (enc.sv);
488	encode_sv (&enc, scalar);	658	encode_sv (&enc, scalar);
489		659
		660	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		661	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		662
490	if (!(flags & (F_ASCII \| F_UTF8)))	663	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))
491	SvUTF8_on (enc.sv);	664	SvUTF8_on (enc.sv);
492		665
493	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
494
495	if (enc.flags & F_SHRINK)	666	if (enc.json.flags & F_SHRINK)
496	shrink (enc.sv);	667	shrink (enc.sv);
497		668
498	return enc.sv;	669	return enc.sv;
499	}	670	}
500		671
…		…
505	typedef struct	676	typedef struct
506	{	677	{
507	char *cur; // current parser pointer	678	char *cur; // current parser pointer
508	char *end; // end of input string	679	char *end; // end of input string
509	const char *err; // parse error, if != 0	680	const char *err; // parse error, if != 0
510	UV flags; // F_*	681	JSON json;
		682	U32 depth; // recursion depth
		683	U32 maxdepth; // recursion depth limit
511	} dec_t;	684	} dec_t;
512		685
513	static void	686	inline void
514	decode_ws (dec_t *dec)	687	decode_ws (dec_t *dec)
515	{	688	{
516	for (;;)	689	for (;;)
517	{	690	{
518	char ch = *dec->cur;	691	char ch = *dec->cur;
…		…
524	++dec->cur;	697	++dec->cur;
525	}	698	}
526	}	699	}
527		700
528	#define ERR(reason) SB dec->err = reason; goto fail; SE	701	#define ERR(reason) SB dec->err = reason; goto fail; SE
		702
529	#define EXPECT_CH(ch) SB \	703	#define EXPECT_CH(ch) SB \
530	if (*dec->cur != ch) \	704	if (*dec->cur != ch) \
531	ERR (# ch " expected"); \	705	ERR (# ch " expected"); \
532	++dec->cur; \	706	++dec->cur; \
533	SE	707	SE
534		708
		709	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		710	#define DEC_DEC_DEPTH --dec->depth
		711
535	static SV decode_sv (dec_t dec);	712	static SV decode_sv (dec_t dec);
536		713
537	static signed char decode_hexdigit[256];	714	static signed char decode_hexdigit[256];
538		715
539	static UV	716	static UV
540	decode_4hex (dec_t *dec)	717	decode_4hex (dec_t *dec)
541	{	718	{
542	signed char d1, d2, d3, d4;	719	signed char d1, d2, d3, d4;
543	unsigned char cur = (unsigned char )dec->cur;	720	unsigned char cur = (unsigned char )dec->cur;
544		721
545	d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected");	722	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");	723	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");	724	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");	725	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
549		726
550	dec->cur += 4;	727	dec->cur += 4;
551		728
552	return ((UV)d1) << 12	729	return ((UV)d1) << 12
553	\| ((UV)d2) << 8	730	\| ((UV)d2) << 8
…		…
561	static SV *	738	static SV *
562	decode_str (dec_t *dec)	739	decode_str (dec_t *dec)
563	{	740	{
564	SV *sv = 0;	741	SV *sv = 0;
565	int utf8 = 0;	742	int utf8 = 0;
		743	char *dec_cur = dec->cur;
566		744
567	do	745	do
568	{	746	{
569	char buf [SHORT_STRING_LEN + UTF8_MAX_LEN];	747	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
570	char *cur = buf;	748	char *cur = buf;
571		749
572	do	750	do
573	{	751	{
574	unsigned char ch = (unsigned char )dec->cur++;	752	unsigned char ch = (unsigned char )dec_cur++;
575		753
576	if (ch == '"')	754	if (expect_false (ch == '"'))
577	{	755	{
578	--dec->cur;	756	--dec_cur;
579	break;	757	break;
580	}	758	}
581	else if (ch == '\\')	759	else if (expect_false (ch == '\\'))
582	{	760	{
583	switch (*dec->cur)	761	switch (*dec_cur)
584	{	762	{
585	case '\\':	763	case '\\':
586	case '/':	764	case '/':
587	case '"': cur++ = dec->cur++; break;	765	case '"': cur++ = dec_cur++; break;
588		766
589	case 'b': ++dec->cur; *cur++ = '\010'; break;	767	case 'b': ++dec_cur; *cur++ = '\010'; break;
590	case 't': ++dec->cur; *cur++ = '\011'; break;	768	case 't': ++dec_cur; *cur++ = '\011'; break;
591	case 'n': ++dec->cur; *cur++ = '\012'; break;	769	case 'n': ++dec_cur; *cur++ = '\012'; break;
592	case 'f': ++dec->cur; *cur++ = '\014'; break;	770	case 'f': ++dec_cur; *cur++ = '\014'; break;
593	case 'r': ++dec->cur; *cur++ = '\015'; break;	771	case 'r': ++dec_cur; *cur++ = '\015'; break;
594		772
595	case 'u':	773	case 'u':
596	{	774	{
597	UV lo, hi;	775	UV lo, hi;
598	++dec->cur;	776	++dec_cur;
599		777
		778	dec->cur = dec_cur;
600	hi = decode_4hex (dec);	779	hi = decode_4hex (dec);
		780	dec_cur = dec->cur;
601	if (hi == (UV)-1)	781	if (hi == (UV)-1)
602	goto fail;	782	goto fail;
603		783
604	// possibly a surrogate pair	784	// possibly a surrogate pair
605	if (hi >= 0xd800)	785	if (hi >= 0xd800)
606	if (hi < 0xdc00)	786	if (hi < 0xdc00)
607	{	787	{
608	if (dec->cur [0] != '\\' \|\| dec->cur [1] != 'u')	788	if (dec_cur [0] != '\\' \|\| dec_cur [1] != 'u')
609	ERR ("missing low surrogate character in surrogate pair");	789	ERR ("missing low surrogate character in surrogate pair");
610		790
611	dec->cur += 2;	791	dec_cur += 2;
612		792
		793	dec->cur = dec_cur;
613	lo = decode_4hex (dec);	794	lo = decode_4hex (dec);
		795	dec_cur = dec->cur;
614	if (lo == (UV)-1)	796	if (lo == (UV)-1)
615	goto fail;	797	goto fail;
616		798
617	if (lo < 0xdc00 \|\| lo >= 0xe000)	799	if (lo < 0xdc00 \|\| lo >= 0xe000)
618	ERR ("surrogate pair expected");	800	ERR ("surrogate pair expected");
…		…
632	*cur++ = hi;	814	*cur++ = hi;
633	}	815	}
634	break;	816	break;
635		817
636	default:	818	default:
637	--dec->cur;	819	--dec_cur;
638	ERR ("illegal backslash escape sequence in string");	820	ERR ("illegal backslash escape sequence in string");
639	}	821	}
640	}	822	}
641	else if (ch >= 0x20 && ch <= 0x7f)	823	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
642	*cur++ = ch;	824	*cur++ = ch;
643	else if (ch >= 0x80)	825	else if (ch >= 0x80)
644	{	826	{
645	--dec->cur;
646
647	STRLEN clen;	827	STRLEN clen;
		828	UV uch;
		829
		830	--dec_cur;
		831
648	UV uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	832	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
649	if (clen == (STRLEN)-1)	833	if (clen == (STRLEN)-1)
650	ERR ("malformed UTF-8 character in JSON string");	834	ERR ("malformed UTF-8 character in JSON string");
651		835
652	do	836	do
653	{
654	cur++ = dec->cur++;	837	cur++ = dec_cur++;
655	}
656	while (--clen);	838	while (--clen);
657		839
658	utf8 = 1;	840	utf8 = 1;
659	}	841	}
660	else if (!ch)
661	ERR ("unexpected end of string while parsing json string");
662	else	842	else
		843	{
		844	--dec_cur;
		845
		846	if (!ch)
		847	ERR ("unexpected end of string while parsing JSON string");
		848	else
663	ERR ("invalid character encountered");	849	ERR ("invalid character encountered while parsing JSON string");
664		850	}
665	}	851	}
666	while (cur < buf + SHORT_STRING_LEN);	852	while (cur < buf + SHORT_STRING_LEN);
667		853
		854	{
668	STRLEN len = cur - buf;	855	STRLEN len = cur - buf;
669		856
670	if (sv)	857	if (sv)
671	{	858	{
672	SvGROW (sv, SvCUR (sv) + len + 1);	859	SvGROW (sv, SvCUR (sv) + len + 1);
673	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	860	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
674	SvCUR_set (sv, SvCUR (sv) + len);	861	SvCUR_set (sv, SvCUR (sv) + len);
675	}	862	}
676	else	863	else
677	sv = newSVpvn (buf, len);	864	sv = newSVpvn (buf, len);
678	}	865	}
		866	}
679	while (*dec->cur != '"');	867	while (*dec_cur != '"');
680		868
681	++dec->cur;	869	++dec_cur;
682		870
683	if (sv)	871	if (sv)
684	{	872	{
685	SvPOK_only (sv);	873	SvPOK_only (sv);
686	*SvEND (sv) = 0;	874	*SvEND (sv) = 0;
…		…
689	SvUTF8_on (sv);	877	SvUTF8_on (sv);
690	}	878	}
691	else	879	else
692	sv = newSVpvn ("", 0);	880	sv = newSVpvn ("", 0);
693		881
		882	dec->cur = dec_cur;
694	return sv;	883	return sv;
695		884
696	fail:	885	fail:
		886	dec->cur = dec_cur;
697	return 0;	887	return 0;
698	}	888	}
699		889
700	static SV *	890	static SV *
701	decode_num (dec_t *dec)	891	decode_num (dec_t *dec)
…		…
759	is_nv = 1;	949	is_nv = 1;
760	}	950	}
761		951
762	if (!is_nv)	952	if (!is_nv)
763	{	953	{
764	UV uv;	954	int len = dec->cur - start;
765	int numtype = grok_number (start, dec->cur - start, &uv);	955
766	if (numtype & IS_NUMBER_IN_UV)	956	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
767	if (numtype & IS_NUMBER_NEG)	957	if (*start == '-')
		958	switch (len)
768	{	959	{
769	if (uv < (UV)IV_MIN)	960	case 2: return newSViv (-( start [1] - '0' * 1));
770	return newSViv (-(IV)uv);	961	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		962	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		963	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
771	}	964	}
		965	else
		966	switch (len)
		967	{
		968	case 1: return newSViv ( start [0] - '0' * 1);
		969	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		970	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		971	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		972	}
		973
		974	{
		975	UV uv;
		976	int numtype = grok_number (start, len, &uv);
		977	if (numtype & IS_NUMBER_IN_UV)
		978	if (numtype & IS_NUMBER_NEG)
		979	{
		980	if (uv < (UV)IV_MIN)
		981	return newSViv (-(IV)uv);
		982	}
772	else	983	else
773	return newSVuv (uv);	984	return newSVuv (uv);
774	}	985	}
775		986
		987	len -= *start == '-' ? 1 : 0;
		988
		989	// does not fit into IV or UV, try NV
		990	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		991	#if defined (LDBL_DIG)
		992	\|\| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		993	#endif
		994	)
		995	// fits into NV without loss of precision
		996	return newSVnv (Atof (start));
		997
		998	// everything else fails, convert it to a string
		999	return newSVpvn (start, dec->cur - start);
		1000	}
		1001
		1002	// loss of precision here
776	return newSVnv (Atof (start));	1003	return newSVnv (Atof (start));
777		1004
778	fail:	1005	fail:
779	return 0;	1006	return 0;
780	}	1007	}
…		…
782	static SV *	1009	static SV *
783	decode_av (dec_t *dec)	1010	decode_av (dec_t *dec)
784	{	1011	{
785	AV *av = newAV ();	1012	AV *av = newAV ();
786		1013
		1014	DEC_INC_DEPTH;
787	decode_ws (dec);	1015	decode_ws (dec);
		1016
788	if (*dec->cur == ']')	1017	if (*dec->cur == ']')
789	++dec->cur;	1018	++dec->cur;
790	else	1019	else
791	for (;;)	1020	for (;;)
792	{	1021	{
…		…
810	ERR (", or ] expected while parsing array");	1039	ERR (", or ] expected while parsing array");
811		1040
812	++dec->cur;	1041	++dec->cur;
813	}	1042	}
814		1043
		1044	DEC_DEC_DEPTH;
815	return newRV_noinc ((SV *)av);	1045	return newRV_noinc ((SV *)av);
816		1046
817	fail:	1047	fail:
818	SvREFCNT_dec (av);	1048	SvREFCNT_dec (av);
		1049	DEC_DEC_DEPTH;
819	return 0;	1050	return 0;
820	}	1051	}
821		1052
822	static SV *	1053	static SV *
823	decode_hv (dec_t *dec)	1054	decode_hv (dec_t *dec)
824	{	1055	{
		1056	SV *sv;
825	HV *hv = newHV ();	1057	HV *hv = newHV ();
826		1058
		1059	DEC_INC_DEPTH;
827	decode_ws (dec);	1060	decode_ws (dec);
		1061
828	if (*dec->cur == '}')	1062	if (*dec->cur == '}')
829	++dec->cur;	1063	++dec->cur;
830	else	1064	else
831	for (;;)	1065	for (;;)
832	{	1066	{
833	SV key, value;
834
835	decode_ws (dec); EXPECT_CH ('"');	1067	decode_ws (dec); EXPECT_CH ('"');
836		1068
837	key = decode_str (dec);	1069	// heuristic: assume that
838	if (!key)	1070	// a) decode_str + hv_store_ent are abysmally slow.
839	goto fail;	1071	// b) most hash keys are short, simple ascii text.
		1072	// => try to "fast-match" such strings to avoid
		1073	// the overhead of decode_str + hv_store_ent.
		1074	{
		1075	SV *value;
		1076	char *p = dec->cur;
		1077	char *e = p + 24; // only try up to 24 bytes
840		1078
841	decode_ws (dec); EXPECT_CH (':');	1079	for (;;)
842
843	value = decode_sv (dec);
844	if (!value)
845	{	1080	{
		1081	// the >= 0x80 is true on most architectures
		1082	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
		1083	{
		1084	// slow path, back up and use decode_str
		1085	SV *key = decode_str (dec);
		1086	if (!key)
		1087	goto fail;
		1088
		1089	decode_ws (dec); EXPECT_CH (':');
		1090
		1091	value = decode_sv (dec);
		1092	if (!value)
		1093	{
		1094	SvREFCNT_dec (key);
		1095	goto fail;
		1096	}
		1097
		1098	hv_store_ent (hv, key, value, 0);
846	SvREFCNT_dec (key);	1099	SvREFCNT_dec (key);
		1100
		1101	break;
		1102	}
		1103	else if (*p == '"')
		1104	{
		1105	// fast path, got a simple key
		1106	char *key = dec->cur;
		1107	int len = p - key;
		1108	dec->cur = p + 1;
		1109
		1110	decode_ws (dec); EXPECT_CH (':');
		1111
		1112	value = decode_sv (dec);
		1113	if (!value)
847	goto fail;	1114	goto fail;
		1115
		1116	hv_store (hv, key, len, value, 0);
		1117
		1118	break;
		1119	}
		1120
		1121	++p;
848	}	1122	}
849		1123	}
850	//TODO: optimise
851	hv_store_ent (hv, key, value, 0);
852		1124
853	decode_ws (dec);	1125	decode_ws (dec);
854		1126
855	if (*dec->cur == '}')	1127	if (*dec->cur == '}')
856	{	1128	{
…		…
862	ERR (", or } expected while parsing object/hash");	1134	ERR (", or } expected while parsing object/hash");
863		1135
864	++dec->cur;	1136	++dec->cur;
865	}	1137	}
866		1138
		1139	DEC_DEC_DEPTH;
867	return newRV_noinc ((SV *)hv);	1140	sv = newRV_noinc ((SV *)hv);
		1141
		1142	// check filter callbacks
		1143	if (dec->json.flags & F_HOOK)
		1144	{
		1145	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1146	{
		1147	HE cb, he;
		1148
		1149	hv_iterinit (hv);
		1150	he = hv_iternext (hv);
		1151	hv_iterinit (hv);
		1152
		1153	// the next line creates a mortal sv each time its called.
		1154	// might want to optimise this for common cases.
		1155	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1156
		1157	if (cb)
		1158	{
		1159	dSP;
		1160	int count;
		1161
		1162	ENTER; SAVETMPS; PUSHMARK (SP);
		1163	XPUSHs (HeVAL (he));
		1164
		1165	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1166
		1167	if (count == 1)
		1168	{
		1169	sv = newSVsv (POPs);
		1170	FREETMPS; LEAVE;
		1171	return sv;
		1172	}
		1173
		1174	FREETMPS; LEAVE;
		1175	}
		1176	}
		1177
		1178	if (dec->json.cb_object)
		1179	{
		1180	dSP;
		1181	int count;
		1182
		1183	ENTER; SAVETMPS; PUSHMARK (SP);
		1184	XPUSHs (sv_2mortal (sv));
		1185
		1186	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1187
		1188	if (count == 1)
		1189	{
		1190	sv = newSVsv (POPs);
		1191	FREETMPS; LEAVE;
		1192	return sv;
		1193	}
		1194
		1195	SvREFCNT_inc (sv);
		1196	FREETMPS; LEAVE;
		1197	}
		1198	}
		1199
		1200	return sv;
868		1201
869	fail:	1202	fail:
870	SvREFCNT_dec (hv);	1203	SvREFCNT_dec (hv);
		1204	DEC_DEC_DEPTH;
871	return 0;	1205	return 0;
872	}	1206	}
873		1207
874	static SV *	1208	static SV *
875	decode_sv (dec_t *dec)	1209	decode_sv (dec_t *dec)
876	{	1210	{
877	decode_ws (dec);	1211	decode_ws (dec);
		1212
		1213	// the beauty of JSON: you need exactly one character lookahead
		1214	// to parse anything.
878	switch (*dec->cur)	1215	switch (*dec->cur)
879	{	1216	{
880	case '"': ++dec->cur; return decode_str (dec);	1217	case '"': ++dec->cur; return decode_str (dec);
881	case '[': ++dec->cur; return decode_av (dec);	1218	case '[': ++dec->cur; return decode_av (dec);
882	case '{': ++dec->cur; return decode_hv (dec);	1219	case '{': ++dec->cur; return decode_hv (dec);
…		…
888		1225
889	case 't':	1226	case 't':
890	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1227	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
891	{	1228	{
892	dec->cur += 4;	1229	dec->cur += 4;
893	return newSViv (1);	1230	#if JSON_SLOW
		1231	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);
		1232	#endif
		1233	return SvREFCNT_inc (json_true);
894	}	1234	}
895	else	1235	else
896	ERR ("'true' expected");	1236	ERR ("'true' expected");
897		1237
898	break;	1238	break;
899		1239
900	case 'f':	1240	case 'f':
901	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1241	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
902	{	1242	{
903	dec->cur += 5;	1243	dec->cur += 5;
904	return newSViv (0);	1244	#if JSON_SLOW
		1245	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1246	#endif
		1247	return SvREFCNT_inc (json_false);
905	}	1248	}
906	else	1249	else
907	ERR ("'false' expected");	1250	ERR ("'false' expected");
908		1251
909	break;	1252	break;
…		…
918	ERR ("'null' expected");	1261	ERR ("'null' expected");
919		1262
920	break;	1263	break;
921		1264
922	default:	1265	default:
923	ERR ("malformed json string, neither array, object, number, string or atom");	1266	ERR ("malformed JSON string, neither array, object, number, string or atom");
924	break;	1267	break;
925	}	1268	}
926		1269
927	fail:	1270	fail:
928	return 0;	1271	return 0;
929	}	1272	}
930		1273
931	static SV *	1274	static SV *
932	decode_json (SV *string, UV flags)	1275	decode_json (SV string, JSON json, UV *offset_return)
933	{	1276	{
		1277	dec_t dec;
		1278	UV offset;
934	SV *sv;	1279	SV *sv;
935		1280
		1281	SvGETMAGIC (string);
		1282	SvUPGRADE (string, SVt_PV);
		1283
		1284	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1285	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1286	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1287
936	if (flags & F_UTF8)	1288	if (json->flags & F_UTF8)
937	sv_utf8_downgrade (string, 0);	1289	sv_utf8_downgrade (string, 0);
938	else	1290	else
939	sv_utf8_upgrade (string);	1291	sv_utf8_upgrade (string);
940		1292
941	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1293	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
942		1294
943	dec_t dec;	1295	dec.json = *json;
944	dec.flags = flags;
945	dec.cur = SvPVX (string);	1296	dec.cur = SvPVX (string);
946	dec.end = SvEND (string);	1297	dec.end = SvEND (string);
947	dec.err = 0;	1298	dec.err = 0;
		1299	dec.depth = 0;
		1300	dec.maxdepth = DEC_DEPTH (dec.json.flags);
948		1301
		1302	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
		1303	dec.json.flags \|= F_HOOK;
		1304
		1305	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
949	sv = decode_sv (&dec);	1306	sv = decode_sv (&dec);
950		1307
		1308	if (!(offset_return \|\| !sv))
		1309	{
		1310	// check for trailing garbage
		1311	decode_ws (&dec);
		1312
		1313	if (*dec.cur)
		1314	{
		1315	dec.err = "garbage after JSON object";
		1316	SvREFCNT_dec (sv);
		1317	sv = 0;
		1318	}
		1319	}
		1320
		1321	if (offset_return \|\| !sv)
		1322	{
		1323	offset = dec.json.flags & F_UTF8
		1324	? dec.cur - SvPVX (string)
		1325	: utf8_distance (dec.cur, SvPVX (string));
		1326
		1327	if (offset_return)
		1328	*offset_return = offset;
		1329	}
		1330
951	if (!sv)	1331	if (!sv)
952	{	1332	{
953	IV offset = dec.flags & F_UTF8
954	? dec.cur - SvPVX (string)
955	: utf8_distance (dec.cur, SvPVX (string));
956	SV *uni = sv_newmortal ();	1333	SV *uni = sv_newmortal ();
957		1334
958	// horrible hack to silence warning inside pv_uni_display	1335	// horrible hack to silence warning inside pv_uni_display
959	COP cop = *PL_curcop;	1336	COP cop = *PL_curcop;
960	cop.cop_warnings = pWARN_NONE;	1337	cop.cop_warnings = pWARN_NONE;
…		…
962	SAVEVPTR (PL_curcop);	1339	SAVEVPTR (PL_curcop);
963	PL_curcop = &cop;	1340	PL_curcop = &cop;
964	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1341	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
965	LEAVE;	1342	LEAVE;
966		1343
967	croak ("%s, at character offset %d (%s)",	1344	croak ("%s, at character offset %d [\"%s\"]",
968	dec.err,	1345	dec.err,
969	(int)offset,	1346	(int)offset,
970	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1347	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
971	}	1348	}
972		1349
973	sv = sv_2mortal (sv);	1350	sv = sv_2mortal (sv);
974		1351
975	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1352	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)");	1353	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
977		1354
978	return sv;	1355	return sv;
979	}	1356	}
980		1357
…		…
985		1362
986	BOOT:	1363	BOOT:
987	{	1364	{
988	int i;	1365	int i;
989		1366
990	memset (decode_hexdigit, 0xff, 256);
991	for (i = 10; i--; )	1367	for (i = 0; i < 256; ++i)
992	decode_hexdigit ['0' + i] = i;	1368	decode_hexdigit [i] =
		1369	i >= '0' && i <= '9' ? i - '0'
		1370	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1371	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1372	: -1;
993		1373
994	for (i = 7; i--; )	1374	json_stash = gv_stashpv ("JSON::XS" , 1);
		1375	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1376
		1377	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1378	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1379	}
		1380
		1381	PROTOTYPES: DISABLE
		1382
		1383	void CLONE (...)
		1384	CODE:
		1385	json_stash = 0;
		1386	json_boolean_stash = 0;
		1387
		1388	void new (char *klass)
		1389	PPCODE:
		1390	{
		1391	SV *pv = NEWSV (0, sizeof (JSON));
		1392	SvPOK_only (pv);
		1393	Zero (SvPVX (pv), 1, JSON);
		1394	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1395	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH)));
		1396	}
		1397
		1398	void ascii (JSON *self, int enable = 1)
		1399	ALIAS:
		1400	ascii = F_ASCII
		1401	latin1 = F_LATIN1
		1402	utf8 = F_UTF8
		1403	indent = F_INDENT
		1404	canonical = F_CANONICAL
		1405	space_before = F_SPACE_BEFORE
		1406	space_after = F_SPACE_AFTER
		1407	pretty = F_PRETTY
		1408	allow_nonref = F_ALLOW_NONREF
		1409	shrink = F_SHRINK
		1410	allow_blessed = F_ALLOW_BLESSED
		1411	convert_blessed = F_CONV_BLESSED
		1412	PPCODE:
		1413	{
		1414	if (enable)
		1415	self->flags \|= ix;
		1416	else
		1417	self->flags &= ~ix;
		1418
		1419	XPUSHs (ST (0));
		1420	}
		1421
		1422	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1423	PPCODE:
		1424	{
		1425	UV log2 = 0;
		1426
		1427	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1428
		1429	while ((1UL << log2) < max_depth)
		1430	++log2;
		1431
		1432	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
		1433
		1434	XPUSHs (ST (0));
		1435	}
		1436
		1437	void max_size (JSON *self, UV max_size = 0)
		1438	PPCODE:
		1439	{
		1440	UV log2 = 0;
		1441
		1442	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1443	if (max_size == 1) max_size = 2;
		1444
		1445	while ((1UL << log2) < max_size)
		1446	++log2;
		1447
		1448	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
		1449
		1450	XPUSHs (ST (0));
		1451	}
		1452
		1453	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
		1454	PPCODE:
		1455	{
		1456	SvREFCNT_dec (self->cb_object);
		1457	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1458
		1459	XPUSHs (ST (0));
		1460	}
		1461
		1462	void filter_json_single_key_object (JSON self, SV key, SV *cb = &PL_sv_undef)
		1463	PPCODE:
		1464	{
		1465	if (!self->cb_sk_object)
		1466	self->cb_sk_object = newHV ();
		1467
		1468	if (SvOK (cb))
		1469	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1470	else
995	{	1471	{
996	decode_hexdigit ['a' + i] = 10 + i;	1472	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
997	decode_hexdigit ['A' + i] = 10 + i;	1473
		1474	if (!HvKEYS (self->cb_sk_object))
		1475	{
		1476	SvREFCNT_dec (self->cb_sk_object);
		1477	self->cb_sk_object = 0;
		1478	}
998	}	1479	}
999		1480
1000	json_stash = gv_stashpv ("JSON::XS", 1);	1481	XPUSHs (ST (0));
1001	}	1482	}
1002		1483
1003	PROTOTYPES: DISABLE	1484	void encode (JSON self, SV scalar)
		1485	PPCODE:
		1486	XPUSHs (encode_json (scalar, self));
1004		1487
1005	SV new (char dummy)	1488	void decode (JSON self, SV jsonstr)
		1489	PPCODE:
		1490	XPUSHs (decode_json (jsonstr, self, 0));
		1491
		1492	void decode_prefix (JSON self, SV jsonstr)
		1493	PPCODE:
		1494	{
		1495	UV offset;
		1496	EXTEND (SP, 2);
		1497	PUSHs (decode_json (jsonstr, self, &offset));
		1498	PUSHs (sv_2mortal (newSVuv (offset)));
		1499	}
		1500
		1501	void DESTROY (JSON *self)
1006	CODE:	1502	CODE:
1007	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1503	SvREFCNT_dec (self->cb_sk_object);
1008	OUTPUT:	1504	SvREFCNT_dec (self->cb_object);
1009	RETVAL
1010		1505
1011	SV ascii (SV self, int enable = 1)	1506	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		1507
1030	RETVAL = newSVsv (self);	1508	void to_json (SV *scalar)
1031	}
1032	OUTPUT:
1033	RETVAL
1034
1035	void encode (SV self, SV scalar)
1036	PPCODE:	1509	PPCODE:
		1510	{
		1511	JSON json = { F_DEFAULT \| F_UTF8 };
1037	XPUSHs (encode_json (scalar, *SvJSON (self)));	1512	XPUSHs (encode_json (scalar, &json));
		1513	}
1038		1514
1039	void decode (SV self, SV jsonstr)	1515	void from_json (SV *jsonstr)
1040	PPCODE:	1516	PPCODE:
1041	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1517	{
1042		1518	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));	1519	XPUSHs (decode_json (jsonstr, &json, 0));
		1520	}
1056		1521

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 1.60 by root, Mon Aug 13 16:19:13 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 1.60 by root, Mon Aug 13 16:19:13 2007 UTC