ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.11 by root, Sat Mar 24 19:42:14 2007 UTC vs.
Revision 1.73 by root, Wed Mar 19 13:44:43 2008 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_RELAXED 0x00001000UL
		33
		34	#define F_MAXDEPTH 0xf8000000UL
		35	#define S_MAXDEPTH 27
		36	#define F_MAXSIZE 0x01f00000UL
		37	#define S_MAXSIZE 20
		38	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
		39
		40	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
		41	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
17		42
18	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	43	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
19	#define F_DEFAULT 0	44	#define F_DEFAULT (9UL << S_MAXDEPTH)
20		45
21	#define INIT_SIZE 32 // initial scalar size to be allocated	46	#define INIT_SIZE 32 // initial scalar size to be allocated
		47	#define INDENT_STEP 3 // spaces per indentation level
		48
		49	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
22		50
23	#define SB do {	51	#define SB do {
24	#define SE } while (0)	52	#define SE } while (0)
25		53
26	static HV *json_stash;	54	#if __GNUC__ >= 3
		55	# define expect(expr,value) __builtin_expect ((expr), (value))
		56	# define INLINE static inline
		57	#else
		58	# define expect(expr,value) (expr)
		59	# define INLINE static
		60	#endif
		61
		62	#define expect_false(expr) expect ((expr) != 0, 0)
		63	#define expect_true(expr) expect ((expr) != 0, 1)
		64
		65	#define IN_RANGE_INC(type,val,beg,end) \
		66	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		67	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		68
		69	#ifdef USE_ITHREADS
		70	# define JSON_SLOW 1
		71	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		72	#else
		73	# define JSON_SLOW 0
		74	# define JSON_STASH json_stash
		75	#endif
		76
		77	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		78	static SV *json_true, *json_false;
		79
		80	typedef struct {
		81	U32 flags;
		82	SV *cb_object;
		83	HV *cb_sk_object;
		84	} JSON;
		85
		86	/////////////////////////////////////////////////////////////////////////////
		87	// utility functions
		88
		89	INLINE void
		90	shrink (SV *sv)
		91	{
		92	sv_utf8_downgrade (sv, 1);
		93	if (SvLEN (sv) > SvCUR (sv) + 1)
		94	{
		95	#ifdef SvPV_shrink_to_cur
		96	SvPV_shrink_to_cur (sv);
		97	#elif defined (SvPV_renew)
		98	SvPV_renew (sv, SvCUR (sv) + 1);
		99	#endif
		100	}
		101	}
		102
		103	// decode an utf-8 character and return it, or (UV)-1 in
		104	// case of an error.
		105	// we special-case "safe" characters from U+80 .. U+7FF,
		106	// but use the very good perl function to parse anything else.
		107	// note that we never call this function for a ascii codepoints
		108	INLINE UV
		109	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		110	{
		111	if (expect_true (len >= 2
		112	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
		113	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
		114	{
		115	*clen = 2;
		116	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		117	}
		118	else
		119	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		120	}
		121
		122	// likewise for encoding, also never called for ascii codepoints
		123	// this function takes advantage of this fact, although current gccs
		124	// seem to optimise the check for >= 0x80 away anyways
		125	INLINE unsigned char *
		126	encode_utf8 (unsigned char *s, UV ch)
		127	{
		128	if (ch <= 0x7FF)
		129	{
		130	*s++ = (ch >> 6) | 0xc0;
		131	*s++ = (ch & 0x3f) | 0x80;
		132	}
		133	else
		134	s = uvuni_to_utf8_flags (s, ch, 0);
		135
		136	return s;
		137	}
		138
		139	/////////////////////////////////////////////////////////////////////////////
		140	// encoder
27		141
28	// structure used for encoding JSON	142	// structure used for encoding JSON
29	typedef struct	143	typedef struct
30	{	144	{
31	char *cur;	145	char *cur; // SvPVX (sv) + current output position
32	STRLEN len; // SvLEN (sv)
33	char *end; // SvEND (sv)	146	char *end; // SvEND (sv)
34	SV *sv;	147	SV *sv; // result scalar
35	UV flags;	148	JSON json;
36	int max_recurse;	149	U32 indent; // indentation level
37	int indent;	150	U32 maxdepth; // max. indentation/recursion level
		151	UV limit; // escape character values >= this value when encoding
38	} enc_t;	152	} enc_t;
39		153
40	// structure used for decoding JSON	154	INLINE void
41	typedef struct
42	{
43	char *cur;
44	char *end;
45	const char *err;
46	UV flags;
47	} dec_t;
48
49	static UV *
50	SvJSON (SV *sv)
51	{
52	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
53	croak ("object is not of type JSON::XS");
54
55	return &SvUVX (SvRV (sv));
56	}
57
58	static void
59	shrink (SV *sv)
60	{
61	sv_utf8_downgrade (sv, 1);
62	#ifdef SvPV_shrink_to_cur
63	SvPV_shrink_to_cur (sv);
64	#endif
65	}
66
67	/////////////////////////////////////////////////////////////////////////////
68
69	static void
70	need (enc_t *enc, STRLEN len)	155	need (enc_t *enc, STRLEN len)
71	{	156	{
72	if (enc->cur + len >= enc->end)	157	if (expect_false (enc->cur + len >= enc->end))
73	{	158	{
74	STRLEN cur = enc->cur - SvPVX (enc->sv);	159	STRLEN cur = enc->cur - SvPVX (enc->sv);
75	SvGROW (enc->sv, cur + len + 1);	160	SvGROW (enc->sv, cur + len + 1);
76	enc->cur = SvPVX (enc->sv) + cur;	161	enc->cur = SvPVX (enc->sv) + cur;
77	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	162	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
78	}	163	}
79	}	164	}
80		165
81	static void	166	INLINE void
82	encode_ch (enc_t *enc, char ch)	167	encode_ch (enc_t *enc, char ch)
83	{	168	{
84	need (enc, 1);	169	need (enc, 1);
85	*enc->cur++ = ch;	170	*enc->cur++ = ch;
86	}	171	}
…		…
94		179
95	while (str < end)	180	while (str < end)
96	{	181	{
97	unsigned char ch = *(unsigned char *)str;	182	unsigned char ch = *(unsigned char *)str;
98		183
99	if (ch >= 0x20 && ch < 0x80) // most common case	184	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
100	{	185	{
101	if (ch == '"') // but with slow exceptions	186	if (expect_false (ch == '"')) // but with slow exceptions
102	{	187	{
103	need (enc, len += 1);	188	need (enc, len += 1);
104	*enc->cur++ = '\\';	189	*enc->cur++ = '\\';
105	*enc->cur++ = '"';	190	*enc->cur++ = '"';
106	}	191	}
107	else if (ch == '\\')	192	else if (expect_false (ch == '\\'))
108	{	193	{
109	need (enc, len += 1);	194	need (enc, len += 1);
110	*enc->cur++ = '\\';	195	*enc->cur++ = '\\';
111	*enc->cur++ = '\\';	196	*enc->cur++ = '\\';
112	}	197	}
…		…
130	STRLEN clen;	215	STRLEN clen;
131	UV uch;	216	UV uch;
132		217
133	if (is_utf8)	218	if (is_utf8)
134	{	219	{
135	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	220	uch = decode_utf8 (str, end - str, &clen);
136	if (clen == (STRLEN)-1)	221	if (clen == (STRLEN)-1)
137	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	222	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
138	}	223	}
139	else	224	else
140	{	225	{
141	uch = ch;	226	uch = ch;
142	clen = 1;	227	clen = 1;
143	}	228	}
144		229
145	if (uch > 0x10FFFFUL)	230	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
146	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
147
148	if (uch < 0x80 || enc->flags & F_ASCII)
149	{	231	{
150	if (uch > 0xFFFFUL)	232	if (uch > 0xFFFFUL)
151	{	233	{
		234	if (uch > 0x10FFFFUL)
		235	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		236
152	need (enc, len += 11);	237	need (enc, len += 11);
153	sprintf (enc->cur, "\\u%04x\\u%04x",	238	sprintf (enc->cur, "\\u%04x\\u%04x",
154	(int)((uch - 0x10000) / 0x400 + 0xD800),	239	(int)((uch - 0x10000) / 0x400 + 0xD800),
155	(int)((uch - 0x10000) % 0x400 + 0xDC00));	240	(int)((uch - 0x10000) % 0x400 + 0xDC00));
156	enc->cur += 12;	241	enc->cur += 12;
…		…
167	*enc->cur++ = hexdigit [(uch >> 0) & 15];	252	*enc->cur++ = hexdigit [(uch >> 0) & 15];
168	}	253	}
169		254
170	str += clen;	255	str += clen;
171	}	256	}
		257	else if (enc->json.flags & F_LATIN1)
		258	{
		259	*enc->cur++ = uch;
		260	str += clen;
		261	}
172	else if (is_utf8)	262	else if (is_utf8)
173	{	263	{
174	need (enc, len += clen);	264	need (enc, len += clen);
175	do	265	do
176	{	266	{
…		…
178	}	268	}
179	while (--clen);	269	while (--clen);
180	}	270	}
181	else	271	else
182	{	272	{
183	need (enc, len += 10); // never more than 11 bytes needed	273	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
184	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	274	enc->cur = encode_utf8 (enc->cur, uch);
185	++str;	275	++str;
186	}	276	}
187	}	277	}
188	}	278	}
189	}	279	}
190		280
191	--len;	281	--len;
192	}	282	}
193	}	283	}
194		284
195	#define INDENT SB \	285	INLINE void
		286	encode_indent (enc_t *enc)
		287	{
196	if (enc->flags & F_INDENT) \	288	if (enc->json.flags & F_INDENT)
197	{ \	289	{
198	int i_; \	290	int spaces = enc->indent * INDENT_STEP;
199	need (enc, enc->indent); \	291
200	for (i_ = enc->indent * 3; i_--; )\	292	need (enc, spaces);
		293	memset (enc->cur, ' ', spaces);
		294	enc->cur += spaces;
		295	}
		296	}
		297
		298	INLINE void
		299	encode_space (enc_t *enc)
		300	{
		301	need (enc, 1);
		302	encode_ch (enc, ' ');
		303	}
		304
		305	INLINE void
		306	encode_nl (enc_t *enc)
		307	{
		308	if (enc->json.flags & F_INDENT)
		309	{
		310	need (enc, 1);
201	encode_ch (enc, ' '); \	311	encode_ch (enc, '\n');
202	} \	312	}
203	SE	313	}
204		314
205	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	315	INLINE void
206	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	316	encode_comma (enc_t *enc)
207	#define COMMA SB \	317	{
208	encode_ch (enc, ','); \	318	encode_ch (enc, ',');
		319
209	if (enc->flags & F_INDENT) \	320	if (enc->json.flags & F_INDENT)
210	NL; \	321	encode_nl (enc);
211	else if (enc->flags & F_SPACE_AFTER) \	322	else if (enc->json.flags & F_SPACE_AFTER)
212	SPACE; \	323	encode_space (enc);
213	SE	324	}
214		325
215	static void encode_sv (enc_t *enc, SV *sv);	326	static void encode_sv (enc_t *enc, SV *sv);
216		327
217	static void	328	static void
218	encode_av (enc_t *enc, AV *av)	329	encode_av (enc_t *enc, AV *av)
219	{	330	{
220	int i, len = av_len (av);	331	int i, len = av_len (av);
221		332
		333	if (enc->indent >= enc->maxdepth)
		334	croak ("data structure too deep (hit recursion limit)");
		335
222	encode_ch (enc, '['); NL;	336	encode_ch (enc, '[');
223	++enc->indent;	337
		338	if (len >= 0)
		339	{
		340	encode_nl (enc); ++enc->indent;
224		341
225	for (i = 0; i <= len; ++i)	342	for (i = 0; i <= len; ++i)
226	{	343	{
227	INDENT;	344	SV **svp = av_fetch (av, i, 0);
228	encode_sv (enc, *av_fetch (av, i, 0));
229		345
		346	encode_indent (enc);
		347
		348	if (svp)
		349	encode_sv (enc, *svp);
		350	else
		351	encode_str (enc, "null", 4, 0);
		352
230	if (i < len)	353	if (i < len)
231	COMMA;	354	encode_comma (enc);
232	}	355	}
233		356
234	NL;	357	encode_nl (enc); --enc->indent; encode_indent (enc);
235		358	}
236	--enc->indent;	359
237	INDENT; encode_ch (enc, ']');	360	encode_ch (enc, ']');
238	}	361	}
239		362
240	static void	363	static void
241	encode_he (enc_t *enc, HE *he)	364	encode_hk (enc_t *enc, HE *he)
242	{	365	{
243	encode_ch (enc, '"');	366	encode_ch (enc, '"');
244		367
245	if (HeKLEN (he) == HEf_SVKEY)	368	if (HeKLEN (he) == HEf_SVKEY)
246	{	369	{
…		…
256	else	379	else
257	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	380	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
258		381
259	encode_ch (enc, '"');	382	encode_ch (enc, '"');
260		383
261	if (enc->flags & F_SPACE_BEFORE) SPACE;	384	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
262	encode_ch (enc, ':');	385	encode_ch (enc, ':');
263	if (enc->flags & F_SPACE_AFTER ) SPACE;	386	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
264	encode_sv (enc, HeVAL (he));
265	}	387	}
266		388
267	// compare hash entries, used when all keys are bytestrings	389	// compare hash entries, used when all keys are bytestrings
268	static int	390	static int
269	he_cmp_fast (const void *a_, const void *b_)	391	he_cmp_fast (const void *a_, const void *b_)
…		…
274	HE *b = *(HE **)b_;	396	HE *b = *(HE **)b_;
275		397
276	STRLEN la = HeKLEN (a);	398	STRLEN la = HeKLEN (a);
277	STRLEN lb = HeKLEN (b);	399	STRLEN lb = HeKLEN (b);
278		400
279	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	401	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
280	cmp = la - lb;	402	cmp = lb - la;
281		403
282	return cmp;	404	return cmp;
283	}	405	}
284		406
285	// compare hash entries, used when some keys are sv's or utf-x	407	// compare hash entries, used when some keys are sv's or utf-x
286	static int	408	static int
287	he_cmp_slow (const void *a, const void *b)	409	he_cmp_slow (const void *a, const void *b)
288	{	410	{
289	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	411	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
290	}	412	}
291		413
292	static void	414	static void
293	encode_hv (enc_t *enc, HV *hv)	415	encode_hv (enc_t *enc, HV *hv)
294	{	416	{
		417	HE *he;
295	int count, i;	418	int count;
296		419
297	encode_ch (enc, '{'); NL; ++enc->indent;	420	if (enc->indent >= enc->maxdepth)
		421	croak ("data structure too deep (hit recursion limit)");
298		422
299	if ((count = hv_iterinit (hv)))	423	encode_ch (enc, '{');
300	{	424
301	// for canonical output we have to sort by keys first	425	// for canonical output we have to sort by keys first
302	// actually, this is mostly due to the stupid so-called	426	// actually, this is mostly due to the stupid so-called
303	// security workaround added somewhere in 5.8.x.	427	// security workaround added somewhere in 5.8.x.
304	// that randomises hash orderings	428	// that randomises hash orderings
305	if (enc->flags & F_CANONICAL)	429	if (enc->json.flags & F_CANONICAL)
		430	{
		431	int count = hv_iterinit (hv);
		432
		433	if (SvMAGICAL (hv))
306	{	434	{
307	HE *he, *hes [count];	435	// need to count by iterating. could improve by dynamically building the vector below
		436	// but I don't care for the speed of this special case.
		437	// note also that we will run into undefined behaviour when the two iterations
		438	// do not result in the same count, something I might care for in some later release.
		439
		440	count = 0;
		441	while (hv_iternext (hv))
		442	++count;
		443
		444	hv_iterinit (hv);
		445	}
		446
		447	if (count)
		448	{
308	int fast = 1;	449	int i, fast = 1;
		450	#if defined(__BORLANDC__) || defined(_MSC_VER)
		451	HE **hes = _alloca (count * sizeof (HE));
		452	#else
		453	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		454	#endif
309		455
310	i = 0;	456	i = 0;
311	while ((he = hv_iternext (hv)))	457	while ((he = hv_iternext (hv)))
312	{	458	{
313	hes [i++] = he;	459	hes [i++] = he;
…		…
335		481
336	FREETMPS;	482	FREETMPS;
337	LEAVE;	483	LEAVE;
338	}	484	}
339		485
340	for (i = 0; i < count; ++i)	486	encode_nl (enc); ++enc->indent;
		487
		488	while (count--)
341	{	489	{
342	INDENT;	490	encode_indent (enc);
		491	he = hes [count];
343	encode_he (enc, hes [i]);	492	encode_hk (enc, he);
		493	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
344		494
345	if (i < count - 1)	495	if (count)
346	COMMA;	496	encode_comma (enc);
347	}	497	}
348		498
		499	encode_nl (enc); --enc->indent; encode_indent (enc);
		500	}
		501	}
		502	else
		503	{
		504	if (hv_iterinit (hv) || SvMAGICAL (hv))
		505	if ((he = hv_iternext (hv)))
349	NL;	506	{
		507	encode_nl (enc); ++enc->indent;
		508
		509	for (;;)
		510	{
		511	encode_indent (enc);
		512	encode_hk (enc, he);
		513	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		514
		515	if (!(he = hv_iternext (hv)))
		516	break;
		517
		518	encode_comma (enc);
		519	}
		520
		521	encode_nl (enc); --enc->indent; encode_indent (enc);
		522	}
		523	}
		524
		525	encode_ch (enc, '}');
		526	}
		527
		528	// encode objects, arrays and special \0=false and \1=true values.
		529	static void
		530	encode_rv (enc_t *enc, SV *sv)
		531	{
		532	svtype svt;
		533
		534	SvGETMAGIC (sv);
		535	svt = SvTYPE (sv);
		536
		537	if (expect_false (SvOBJECT (sv)))
		538	{
		539	HV *stash = !JSON_SLOW || json_boolean_stash
		540	? json_boolean_stash
		541	: gv_stashpv ("JSON::XS::Boolean", 1);
		542
		543	if (SvSTASH (sv) == stash)
		544	{
		545	if (SvIV (sv))
		546	encode_str (enc, "true", 4, 0);
		547	else
		548	encode_str (enc, "false", 5, 0);
350	}	549	}
351	else	550	else
352	{	551	{
353	SV *sv;	552	#if 0
354	HE *he = hv_iternext (hv);	553	if (0 && sv_derived_from (rv, "JSON::Literal"))
355
356	for (;;)
357	{	554	{
358	INDENT;	555	// not yet
359	encode_he (enc, he);
360
361	if (!(he = hv_iternext (hv)))
362	break;
363
364	COMMA;
365	}	556	}
366		557	#endif
		558	if (enc->json.flags & F_CONV_BLESSED)
367	NL;	559	{
		560	// we re-bless the reference to get overload and other niceties right
		561	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
		562
		563	if (to_json)
		564	{
		565	dSP;
		566
		567	ENTER; SAVETMPS; PUSHMARK (SP);
		568	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		569
		570	// calling with G_SCALAR ensures that we always get a 1 return value
		571	PUTBACK;
		572	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		573	SPAGAIN;
		574
		575	// catch this surprisingly common error
		576	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		577	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		578
		579	sv = POPs;
		580	PUTBACK;
		581
		582	encode_sv (enc, sv);
		583
		584	FREETMPS; LEAVE;
		585	}
		586	else if (enc->json.flags & F_ALLOW_BLESSED)
		587	encode_str (enc, "null", 4, 0);
		588	else
		589	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		590	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		591	}
		592	else if (enc->json.flags & F_ALLOW_BLESSED)
		593	encode_str (enc, "null", 4, 0);
		594	else
		595	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		596	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
368	}	597	}
369	}	598	}
		599	else if (svt == SVt_PVHV)
		600	encode_hv (enc, (HV *)sv);
		601	else if (svt == SVt_PVAV)
		602	encode_av (enc, (AV *)sv);
		603	else if (svt < SVt_PVAV)
		604	{
		605	STRLEN len = 0;
		606	char *pv = svt ? SvPV (sv, len) : 0;
370		607
371	--enc->indent; INDENT; encode_ch (enc, '}');	608	if (len == 1 && *pv == '1')
		609	encode_str (enc, "true", 4, 0);
		610	else if (len == 1 && *pv == '0')
		611	encode_str (enc, "false", 5, 0);
		612	else
		613	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		614	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		615	}
		616	else
		617	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		618	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
372	}	619	}
373		620
374	static void	621	static void
375	encode_sv (enc_t *enc, SV *sv)	622	encode_sv (enc_t *enc, SV *sv)
376	{	623	{
…		…
384	encode_str (enc, str, len, SvUTF8 (sv));	631	encode_str (enc, str, len, SvUTF8 (sv));
385	encode_ch (enc, '"');	632	encode_ch (enc, '"');
386	}	633	}
387	else if (SvNOKp (sv))	634	else if (SvNOKp (sv))
388	{	635	{
		636	// trust that perl will do the right thing w.r.t. JSON syntax.
389	need (enc, NV_DIG + 32);	637	need (enc, NV_DIG + 32);
390	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	638	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
391	enc->cur += strlen (enc->cur);	639	enc->cur += strlen (enc->cur);
392	}	640	}
393	else if (SvIOKp (sv))	641	else if (SvIOKp (sv))
394	{	642	{
395	need (enc, 64);	643	// we assume we can always read an IV as a UV
		644	if (SvUV (sv) & ~(UV)0x7fff)
		645	{
		646	// large integer, use the (rather slow) snprintf way.
		647	need (enc, sizeof (UV) * 3);
396	enc->cur +=	648	enc->cur +=
397	SvIsUV(sv)	649	SvIsUV(sv)
398	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	650	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
399	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	651	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		652	}
		653	else
		654	{
		655	// optimise the "small number case"
		656	// code will likely be branchless and use only a single multiplication
		657	I32 i = SvIV (sv);
		658	U32 u;
		659	char digit, nz = 0;
		660
		661	need (enc, 6);
		662
		663	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		664	u = i < 0 ? -i : i;
		665
		666	// convert to 4.28 fixed-point representation
		667	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		668
		669	// now output digit by digit, each time masking out the integer part
		670	// and multiplying by 5 while moving the decimal point one to the right,
		671	// resulting in a net multiplication by 10.
		672	// we always write the digit to memory but conditionally increment
		673	// the pointer, to ease the usage of conditional move instructions.
		674	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
		675	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
		676	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
		677	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
		678	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		679	}
400	}	680	}
401	else if (SvROK (sv))	681	else if (SvROK (sv))
402	{	682	encode_rv (enc, SvRV (sv));
403	SV *rv = SvRV (sv);
404
405	if (!--enc->max_recurse)
406	croak ("data structure too deep (hit recursion limit)");
407
408	switch (SvTYPE (rv))
409	{
410	case SVt_PVAV: encode_av (enc, (AV *)rv); break;
411	case SVt_PVHV: encode_hv (enc, (HV *)rv); break;
412
413	default:
414	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
415	SvPV_nolen (sv));
416	}
417	}
418	else if (!SvOK (sv))	683	else if (!SvOK (sv))
419	encode_str (enc, "null", 4, 0);	684	encode_str (enc, "null", 4, 0);
420	else	685	else
421	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	686	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
422	SvPV_nolen (sv), SvFLAGS (sv));	687	SvPV_nolen (sv), SvFLAGS (sv));
423	}	688	}
424		689
425	static SV *	690	static SV *
426	encode_json (SV *scalar, UV flags)	691	encode_json (SV *scalar, JSON *json)
427	{	692	{
		693	enc_t enc;
		694
428	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	695	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
429	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	696	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
430		697
431	enc_t enc;	698	enc.json = *json;
432	enc.flags = flags;
433	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	699	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
434	enc.cur = SvPVX (enc.sv);	700	enc.cur = SvPVX (enc.sv);
435	enc.end = SvEND (enc.sv);	701	enc.end = SvEND (enc.sv);
436	enc.max_recurse = 0;
437	enc.indent = 0;	702	enc.indent = 0;
		703	enc.maxdepth = DEC_DEPTH (enc.json.flags);
		704	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		705	: enc.json.flags & F_LATIN1 ? 0x000100UL
		706	: 0x10FFFFUL;
438		707
439	SvPOK_only (enc.sv);	708	SvPOK_only (enc.sv);
440	encode_sv (&enc, scalar);	709	encode_sv (&enc, scalar);
441		710
		711	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		712	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		713
442	if (!(flags & (F_ASCII | F_UTF8)))	714	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
443	SvUTF8_on (enc.sv);	715	SvUTF8_on (enc.sv);
444		716
445	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
446
447	if (enc.flags & F_SHRINK)	717	if (enc.json.flags & F_SHRINK)
448	shrink (enc.sv);	718	shrink (enc.sv);
449		719
450	return enc.sv;	720	return enc.sv;
451	}	721	}
452		722
453	/////////////////////////////////////////////////////////////////////////////	723	/////////////////////////////////////////////////////////////////////////////
		724	// decoder
454		725
455	#define WS \	726	// structure used for decoding JSON
		727	typedef struct
		728	{
		729	char *cur; // current parser pointer
		730	char *end; // end of input string
		731	const char *err; // parse error, if != 0
		732	JSON json;
		733	U32 depth; // recursion depth
		734	U32 maxdepth; // recursion depth limit
		735	} dec_t;
		736
		737	INLINE void
		738	decode_comment (dec_t *dec)
		739	{
		740	// only '#'-style comments allowed a.t.m.
		741
		742	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		743	++dec->cur;
		744	}
		745
		746	INLINE void
		747	decode_ws (dec_t *dec)
		748	{
456	for (;;) \	749	for (;;)
457	{ \	750	{
458	char ch = *dec->cur; \	751	char ch = *dec->cur;
		752
459	if (ch > 0x20 \	753	if (ch > 0x20)
		754	{
		755	if (expect_false (ch == '#'))
		756	{
		757	if (dec->json.flags & F_RELAXED)
		758	decode_comment (dec);
		759	else
		760	break;
		761	}
		762	else
		763	break;
		764	}
460	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	765	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
461	break; \	766	break; // parse error, but let higher level handle it, gives better error messages
		767
462	++dec->cur; \	768	++dec->cur;
463	}	769	}
		770	}
464		771
465	#define ERR(reason) SB dec->err = reason; goto fail; SE	772	#define ERR(reason) SB dec->err = reason; goto fail; SE
		773
466	#define EXPECT_CH(ch) SB \	774	#define EXPECT_CH(ch) SB \
467	if (*dec->cur != ch) \	775	if (*dec->cur != ch) \
468	ERR (# ch " expected"); \	776	ERR (# ch " expected"); \
469	++dec->cur; \	777	++dec->cur; \
470	SE	778	SE
471		779
		780	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		781	#define DEC_DEC_DEPTH --dec->depth
		782
472	static SV *decode_sv (dec_t *dec);	783	static SV *decode_sv (dec_t *dec);
473		784
474	static signed char decode_hexdigit[256];	785	static signed char decode_hexdigit[256];
475		786
476	static UV	787	static UV
477	decode_4hex (dec_t *dec)	788	decode_4hex (dec_t *dec)
478	{	789	{
479	signed char d1, d2, d3, d4;	790	signed char d1, d2, d3, d4;
		791	unsigned char *cur = (unsigned char *)dec->cur;
480		792
481	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	793	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
482	if (d1 < 0) ERR ("four hexadecimal digits expected");	794	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
483	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	795	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
484	if (d2 < 0) ERR ("four hexadecimal digits expected");	796	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
485	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
486	if (d3 < 0) ERR ("four hexadecimal digits expected");
487	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
488	if (d4 < 0) ERR ("four hexadecimal digits expected");
489		797
490	dec->cur += 4;	798	dec->cur += 4;
491		799
492	return ((UV)d1) << 12	800	return ((UV)d1) << 12
493	| ((UV)d2) << 8	801	| ((UV)d2) << 8
…		…
496		804
497	fail:	805	fail:
498	return (UV)-1;	806	return (UV)-1;
499	}	807	}
500		808
501	#define APPEND_GROW(n) SB \
502	if (cur + (n) >= end) \
503	{ \
504	STRLEN ofs = cur - SvPVX (sv); \
505	SvGROW (sv, ofs + (n) + 1); \
506	cur = SvPVX (sv) + ofs; \
507	end = SvEND (sv); \
508	} \
509	SE
510
511	#define APPEND_CH(ch) SB \
512	APPEND_GROW (1); \
513	*cur++ = (ch); \
514	SE
515
516	static SV *	809	static SV *
517	decode_str (dec_t *dec)	810	decode_str (dec_t *dec)
518	{	811	{
519	SV *sv = NEWSV (0,2);	812	SV *sv = 0;
520	int utf8 = 0;	813	int utf8 = 0;
521	char *cur = SvPVX (sv);	814	char *dec_cur = dec->cur;
522	char *end = SvEND (sv);
523		815
524	for (;;)	816	do
525	{	817	{
526	unsigned char ch = *(unsigned char *)dec->cur;	818	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		819	char *cur = buf;
527		820
528	if (ch == '"')	821	do
529	break;
530	else if (ch == '\\')
531	{	822	{
532	switch (*++dec->cur)	823	unsigned char ch = *(unsigned char *)dec_cur++;
		824
		825	if (expect_false (ch == '"'))
533	{	826	{
534	case '\\':	827	--dec_cur;
535	case '/':	828	break;
536	case '"': APPEND_CH (*dec->cur++); break;	829	}
537		830	else if (expect_false (ch == '\\'))
538	case 'b': APPEND_CH ('\010'); ++dec->cur; break;	831	{
539	case 't': APPEND_CH ('\011'); ++dec->cur; break;	832	switch (*dec_cur)
540	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
541	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
542	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
543
544	case 'u':
545	{	833	{
546	UV lo, hi;	834	case '\\':
547	++dec->cur;	835	case '/':
		836	case '"': *cur++ = *dec_cur++; break;
548		837
549	hi = decode_4hex (dec);	838	case 'b': ++dec_cur; *cur++ = '\010'; break;
550	if (hi == (UV)-1)	839	case 't': ++dec_cur; *cur++ = '\011'; break;
551	goto fail;	840	case 'n': ++dec_cur; *cur++ = '\012'; break;
		841	case 'f': ++dec_cur; *cur++ = '\014'; break;
		842	case 'r': ++dec_cur; *cur++ = '\015'; break;
552		843
553	// possibly a surrogate pair	844	case 'u':
554	if (hi >= 0xd800 && hi < 0xdc00)
555	{	845	{
556	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	846	UV lo, hi;
557	ERR ("missing low surrogate character in surrogate pair");	847	++dec_cur;
558		848
559	dec->cur += 2;	849	dec->cur = dec_cur;
560
561	lo = decode_4hex (dec);	850	hi = decode_4hex (dec);
		851	dec_cur = dec->cur;
562	if (lo == (UV)-1)	852	if (hi == (UV)-1)
563	goto fail;	853	goto fail;
564		854
		855	// possibly a surrogate pair
		856	if (hi >= 0xd800)
		857	if (hi < 0xdc00)
		858	{
		859	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		860	ERR ("missing low surrogate character in surrogate pair");
		861
		862	dec_cur += 2;
		863
		864	dec->cur = dec_cur;
		865	lo = decode_4hex (dec);
		866	dec_cur = dec->cur;
		867	if (lo == (UV)-1)
		868	goto fail;
		869
565	if (lo < 0xdc00 || lo >= 0xe000)	870	if (lo < 0xdc00 || lo >= 0xe000)
566	ERR ("surrogate pair expected");	871	ERR ("surrogate pair expected");
567		872
568	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	873	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		874	}
		875	else if (hi < 0xe000)
		876	ERR ("missing high surrogate character in surrogate pair");
		877
		878	if (hi >= 0x80)
		879	{
		880	utf8 = 1;
		881
		882	cur = encode_utf8 (cur, hi);
		883	}
		884	else
		885	*cur++ = hi;
569	}	886	}
570	else if (hi >= 0xdc00 && hi < 0xe000)
571	ERR ("missing high surrogate character in surrogate pair");
572
573	if (hi >= 0x80)
574	{	887	break;
575	utf8 = 1;
576		888
577	APPEND_GROW (4); // at most 4 bytes for 21 bits
578	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
579	}
580	else	889	default:
581	APPEND_CH (hi);	890	--dec_cur;
		891	ERR ("illegal backslash escape sequence in string");
582	}	892	}
583	break;	893	}
		894	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
		895	*cur++ = ch;
		896	else if (ch >= 0x80)
		897	{
		898	STRLEN clen;
		899	UV uch;
584		900
585	default:
586	--dec->cur;	901	--dec_cur;
587	ERR ("illegal backslash escape sequence in string");	902
		903	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		904	if (clen == (STRLEN)-1)
		905	ERR ("malformed UTF-8 character in JSON string");
		906
		907	do
		908	*cur++ = *dec_cur++;
		909	while (--clen);
		910
		911	utf8 = 1;
		912	}
		913	else
		914	{
		915	--dec_cur;
		916
		917	if (!ch)
		918	ERR ("unexpected end of string while parsing JSON string");
		919	else
		920	ERR ("invalid character encountered while parsing JSON string");
588	}	921	}
589	}	922	}
590	else if (ch >= 0x20 && ch <= 0x7f)	923	while (cur < buf + SHORT_STRING_LEN);
591	APPEND_CH (*dec->cur++);	924
592	else if (ch >= 0x80)
593	{	925	{
594	STRLEN clen;	926	STRLEN len = cur - buf;
595	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
596	if (clen == (STRLEN)-1)
597	ERR ("malformed UTF-8 character in JSON string");
598		927
599	APPEND_GROW (clen);	928	if (sv)
600	do
601	{	929	{
602	*cur++ = *dec->cur++;	930	SvGROW (sv, SvCUR (sv) + len + 1);
		931	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		932	SvCUR_set (sv, SvCUR (sv) + len);
603	}	933	}
604	while (--clen);
605
606	utf8 = 1;
607	}
608	else if (dec->cur == dec->end)
609	ERR ("unexpected end of string while parsing json string");
610	else	934	else
611	ERR ("invalid character encountered");	935	sv = newSVpvn (buf, len);
612	}	936	}
		937	}
		938	while (*dec_cur != '"');
613		939
614	++dec->cur;	940	++dec_cur;
615		941
616	SvCUR_set (sv, cur - SvPVX (sv));	942	if (sv)
617		943	{
618	SvPOK_only (sv);	944	SvPOK_only (sv);
619	*SvEND (sv) = 0;	945	*SvEND (sv) = 0;
620		946
621	if (utf8)	947	if (utf8)
622	SvUTF8_on (sv);	948	SvUTF8_on (sv);
		949	}
		950	else
		951	sv = newSVpvn ("", 0);
623		952
624	if (dec->flags & F_SHRINK)	953	dec->cur = dec_cur;
625	shrink (sv);
626
627	return sv;	954	return sv;
628		955
629	fail:	956	fail:
630	SvREFCNT_dec (sv);	957	dec->cur = dec_cur;
631	return 0;	958	return 0;
632	}	959	}
633		960
634	static SV *	961	static SV *
635	decode_num (dec_t *dec)	962	decode_num (dec_t *dec)
…		…
693	is_nv = 1;	1020	is_nv = 1;
694	}	1021	}
695		1022
696	if (!is_nv)	1023	if (!is_nv)
697	{	1024	{
698	UV uv;	1025	int len = dec->cur - start;
699	int numtype = grok_number (start, dec->cur - start, &uv);	1026
700	if (numtype & IS_NUMBER_IN_UV)	1027	// special case the rather common 1..5-digit-int case
701	if (numtype & IS_NUMBER_NEG)	1028	if (*start == '-')
		1029	switch (len)
702	{	1030	{
703	if (uv < (UV)IV_MIN)	1031	case 2: return newSViv (-( start [1] - '0' * 1));
704	return newSViv (-(IV)uv);	1032	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		1033	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1034	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1035	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
705	}	1036	}
		1037	else
		1038	switch (len)
		1039	{
		1040	case 1: return newSViv ( start [0] - '0' * 1);
		1041	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1042	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1043	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1044	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1045	}
		1046
		1047	{
		1048	UV uv;
		1049	int numtype = grok_number (start, len, &uv);
		1050	if (numtype & IS_NUMBER_IN_UV)
		1051	if (numtype & IS_NUMBER_NEG)
		1052	{
		1053	if (uv < (UV)IV_MIN)
		1054	return newSViv (-(IV)uv);
		1055	}
706	else	1056	else
707	return newSVuv (uv);	1057	return newSVuv (uv);
708	}	1058	}
709		1059
		1060	len -= *start == '-' ? 1 : 0;
		1061
		1062	// does not fit into IV or UV, try NV
		1063	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1064	#if defined (LDBL_DIG)
		1065	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1066	#endif
		1067	)
		1068	// fits into NV without loss of precision
		1069	return newSVnv (Atof (start));
		1070
		1071	// everything else fails, convert it to a string
		1072	return newSVpvn (start, dec->cur - start);
		1073	}
		1074
		1075	// loss of precision here
710	return newSVnv (Atof (start));	1076	return newSVnv (Atof (start));
711		1077
712	fail:	1078	fail:
713	return 0;	1079	return 0;
714	}	1080	}
…		…
716	static SV *	1082	static SV *
717	decode_av (dec_t *dec)	1083	decode_av (dec_t *dec)
718	{	1084	{
719	AV *av = newAV ();	1085	AV *av = newAV ();
720		1086
721	WS;	1087	DEC_INC_DEPTH;
		1088	decode_ws (dec);
		1089
722	if (*dec->cur == ']')	1090	if (*dec->cur == ']')
723	++dec->cur;	1091	++dec->cur;
724	else	1092	else
725	for (;;)	1093	for (;;)
726	{	1094	{
…		…
730	if (!value)	1098	if (!value)
731	goto fail;	1099	goto fail;
732		1100
733	av_push (av, value);	1101	av_push (av, value);
734		1102
735	WS;	1103	decode_ws (dec);
736		1104
737	if (*dec->cur == ']')	1105	if (*dec->cur == ']')
738	{	1106	{
739	++dec->cur;	1107	++dec->cur;
740	break;	1108	break;
…		…
742		1110
743	if (*dec->cur != ',')	1111	if (*dec->cur != ',')
744	ERR (", or ] expected while parsing array");	1112	ERR (", or ] expected while parsing array");
745		1113
746	++dec->cur;	1114	++dec->cur;
		1115
		1116	decode_ws (dec);
		1117
		1118	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1119	{
		1120	++dec->cur;
		1121	break;
		1122	}
747	}	1123	}
748		1124
		1125	DEC_DEC_DEPTH;
749	return newRV_noinc ((SV *)av);	1126	return newRV_noinc ((SV *)av);
750		1127
751	fail:	1128	fail:
752	SvREFCNT_dec (av);	1129	SvREFCNT_dec (av);
		1130	DEC_DEC_DEPTH;
753	return 0;	1131	return 0;
754	}	1132	}
755		1133
756	static SV *	1134	static SV *
757	decode_hv (dec_t *dec)	1135	decode_hv (dec_t *dec)
758	{	1136	{
		1137	SV *sv;
759	HV *hv = newHV ();	1138	HV *hv = newHV ();
760		1139
761	WS;	1140	DEC_INC_DEPTH;
		1141	decode_ws (dec);
		1142
762	if (*dec->cur == '}')	1143	if (*dec->cur == '}')
763	++dec->cur;	1144	++dec->cur;
764	else	1145	else
765	for (;;)	1146	for (;;)
766	{	1147	{
767	SV *key, *value;
768
769	WS; EXPECT_CH ('"');	1148	EXPECT_CH ('"');
770		1149
771	key = decode_str (dec);	1150	// heuristic: assume that
772	if (!key)	1151	// a) decode_str + hv_store_ent are abysmally slow.
773	goto fail;	1152	// b) most hash keys are short, simple ascii text.
		1153	// => try to "fast-match" such strings to avoid
		1154	// the overhead of decode_str + hv_store_ent.
		1155	{
		1156	SV *value;
		1157	char *p = dec->cur;
		1158	char *e = p + 24; // only try up to 24 bytes
774		1159
775	WS; EXPECT_CH (':');	1160	for (;;)
776
777	value = decode_sv (dec);
778	if (!value)
779	{	1161	{
		1162	// the >= 0x80 is true on most architectures
		1163	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1164	{
		1165	// slow path, back up and use decode_str
		1166	SV *key = decode_str (dec);
		1167	if (!key)
		1168	goto fail;
		1169
		1170	decode_ws (dec); EXPECT_CH (':');
		1171
		1172	decode_ws (dec);
		1173	value = decode_sv (dec);
		1174	if (!value)
		1175	{
		1176	SvREFCNT_dec (key);
		1177	goto fail;
		1178	}
		1179
		1180	hv_store_ent (hv, key, value, 0);
780	SvREFCNT_dec (key);	1181	SvREFCNT_dec (key);
		1182
		1183	break;
		1184	}
		1185	else if (*p == '"')
		1186	{
		1187	// fast path, got a simple key
		1188	char *key = dec->cur;
		1189	int len = p - key;
		1190	dec->cur = p + 1;
		1191
		1192	decode_ws (dec); EXPECT_CH (':');
		1193
		1194	decode_ws (dec);
		1195	value = decode_sv (dec);
		1196	if (!value)
781	goto fail;	1197	goto fail;
		1198
		1199	hv_store (hv, key, len, value, 0);
		1200
		1201	break;
		1202	}
		1203
		1204	++p;
782	}	1205	}
783
784	//TODO: optimise
785	hv_store_ent (hv, key, value, 0);
786
787	WS;	1206	}
		1207
		1208	decode_ws (dec);
788		1209
789	if (*dec->cur == '}')	1210	if (*dec->cur == '}')
790	{	1211	{
791	++dec->cur;	1212	++dec->cur;
792	break;	1213	break;
…		…
794		1215
795	if (*dec->cur != ',')	1216	if (*dec->cur != ',')
796	ERR (", or } expected while parsing object/hash");	1217	ERR (", or } expected while parsing object/hash");
797		1218
798	++dec->cur;	1219	++dec->cur;
		1220
		1221	decode_ws (dec);
		1222
		1223	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1224	{
		1225	++dec->cur;
		1226	break;
		1227	}
799	}	1228	}
800		1229
		1230	DEC_DEC_DEPTH;
801	return newRV_noinc ((SV *)hv);	1231	sv = newRV_noinc ((SV *)hv);
		1232
		1233	// check filter callbacks
		1234	if (dec->json.flags & F_HOOK)
		1235	{
		1236	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1237	{
		1238	HE *cb, *he;
		1239
		1240	hv_iterinit (hv);
		1241	he = hv_iternext (hv);
		1242	hv_iterinit (hv);
		1243
		1244	// the next line creates a mortal sv each time its called.
		1245	// might want to optimise this for common cases.
		1246	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1247
		1248	if (cb)
		1249	{
		1250	dSP;
		1251	int count;
		1252
		1253	ENTER; SAVETMPS; PUSHMARK (SP);
		1254	XPUSHs (HeVAL (he));
		1255
		1256	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1257
		1258	if (count == 1)
		1259	{
		1260	sv = newSVsv (POPs);
		1261	FREETMPS; LEAVE;
		1262	return sv;
		1263	}
		1264
		1265	FREETMPS; LEAVE;
		1266	}
		1267	}
		1268
		1269	if (dec->json.cb_object)
		1270	{
		1271	dSP;
		1272	int count;
		1273
		1274	ENTER; SAVETMPS; PUSHMARK (SP);
		1275	XPUSHs (sv_2mortal (sv));
		1276
		1277	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1278
		1279	if (count == 1)
		1280	{
		1281	sv = newSVsv (POPs);
		1282	FREETMPS; LEAVE;
		1283	return sv;
		1284	}
		1285
		1286	SvREFCNT_inc (sv);
		1287	FREETMPS; LEAVE;
		1288	}
		1289	}
		1290
		1291	return sv;
802		1292
803	fail:	1293	fail:
804	SvREFCNT_dec (hv);	1294	SvREFCNT_dec (hv);
		1295	DEC_DEC_DEPTH;
805	return 0;	1296	return 0;
806	}	1297	}
807		1298
808	static SV *	1299	static SV *
809	decode_sv (dec_t *dec)	1300	decode_sv (dec_t *dec)
810	{	1301	{
811	WS;	1302	// the beauty of JSON: you need exactly one character lookahead
		1303	// to parse everything.
812	switch (*dec->cur)	1304	switch (*dec->cur)
813	{	1305	{
814	case '"': ++dec->cur; return decode_str (dec);	1306	case '"': ++dec->cur; return decode_str (dec);
815	case '[': ++dec->cur; return decode_av (dec);	1307	case '[': ++dec->cur; return decode_av (dec);
816	case '{': ++dec->cur; return decode_hv (dec);	1308	case '{': ++dec->cur; return decode_hv (dec);
817		1309
818	case '-':	1310	case '-':
819	case '0': case '1': case '2': case '3': case '4':	1311	case '0': case '1': case '2': case '3': case '4':
820	case '5': case '6': case '7': case '8': case '9':	1312	case '5': case '6': case '7': case '8': case '9':
821	return decode_num (dec);	1313	return decode_num (dec);
822		1314
823	case 't':	1315	case 't':
824	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1316	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
825	{	1317	{
826	dec->cur += 4;	1318	dec->cur += 4;
827	return newSViv (1);	1319	#if JSON_SLOW
		1320	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);
		1321	#endif
		1322	return SvREFCNT_inc (json_true);
828	}	1323	}
829	else	1324	else
830	ERR ("'true' expected");	1325	ERR ("'true' expected");
831		1326
832	break;	1327	break;
833		1328
834	case 'f':	1329	case 'f':
835	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1330	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
836	{	1331	{
837	dec->cur += 5;	1332	dec->cur += 5;
838	return newSViv (0);	1333	#if JSON_SLOW
		1334	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1335	#endif
		1336	return SvREFCNT_inc (json_false);
839	}	1337	}
840	else	1338	else
841	ERR ("'false' expected");	1339	ERR ("'false' expected");
842		1340
843	break;	1341	break;
…		…
852	ERR ("'null' expected");	1350	ERR ("'null' expected");
853		1351
854	break;	1352	break;
855		1353
856	default:	1354	default:
857	ERR ("malformed json string, neither array, object, number, string or atom");	1355	ERR ("malformed JSON string, neither array, object, number, string or atom");
858	break;	1356	break;
859	}	1357	}
860		1358
861	fail:	1359	fail:
862	return 0;	1360	return 0;
863	}	1361	}
864		1362
865	static SV *	1363	static SV *
866	decode_json (SV *string, UV flags)	1364	decode_json (SV *string, JSON *json, UV *offset_return)
867	{	1365	{
		1366	dec_t dec;
		1367	UV offset;
868	SV *sv;	1368	SV *sv;
869		1369
		1370	SvGETMAGIC (string);
		1371	SvUPGRADE (string, SVt_PV);
		1372
		1373	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1374	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1375	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1376
870	if (flags & F_UTF8)	1377	if (json->flags & F_UTF8)
871	sv_utf8_downgrade (string, 0);	1378	sv_utf8_downgrade (string, 0);
872	else	1379	else
873	sv_utf8_upgrade (string);	1380	sv_utf8_upgrade (string);
874		1381
875	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1382	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
876		1383
877	dec_t dec;	1384	dec.json = *json;
878	dec.flags = flags;
879	dec.cur = SvPVX (string);	1385	dec.cur = SvPVX (string);
880	dec.end = SvEND (string);	1386	dec.end = SvEND (string);
881	dec.err = 0;	1387	dec.err = 0;
		1388	dec.depth = 0;
		1389	dec.maxdepth = DEC_DEPTH (dec.json.flags);
882		1390
		1391	if (dec.json.cb_object || dec.json.cb_sk_object)
		1392	dec.json.flags |= F_HOOK;
		1393
		1394	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1395
		1396	decode_ws (&dec);
883	sv = decode_sv (&dec);	1397	sv = decode_sv (&dec);
884		1398
		1399	if (!(offset_return || !sv))
		1400	{
		1401	// check for trailing garbage
		1402	decode_ws (&dec);
		1403
		1404	if (*dec.cur)
		1405	{
		1406	dec.err = "garbage after JSON object";
		1407	SvREFCNT_dec (sv);
		1408	sv = 0;
		1409	}
		1410	}
		1411
		1412	if (offset_return || !sv)
		1413	{
		1414	offset = dec.json.flags & F_UTF8
		1415	? dec.cur - SvPVX (string)
		1416	: utf8_distance (dec.cur, SvPVX (string));
		1417
		1418	if (offset_return)
		1419	*offset_return = offset;
		1420	}
		1421
885	if (!sv)	1422	if (!sv)
886	{	1423	{
887	IV offset = dec.flags & F_UTF8
888	? dec.cur - SvPVX (string)
889	: utf8_distance (dec.cur, SvPVX (string));
890	SV *uni = sv_newmortal ();	1424	SV *uni = sv_newmortal ();
891		1425
892	// horrible hack to silence warning inside pv_uni_display	1426	// horrible hack to silence warning inside pv_uni_display
893	COP cop = *PL_curcop;	1427	COP cop = *PL_curcop;
894	cop.cop_warnings = pWARN_NONE;	1428	cop.cop_warnings = pWARN_NONE;
…		…
896	SAVEVPTR (PL_curcop);	1430	SAVEVPTR (PL_curcop);
897	PL_curcop = &cop;	1431	PL_curcop = &cop;
898	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1432	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
899	LEAVE;	1433	LEAVE;
900		1434
901	croak ("%s, at character offset %d (%s)",	1435	croak ("%s, at character offset %d [\"%s\"]",
902	dec.err,	1436	dec.err,
903	(int)offset,	1437	(int)offset,
904	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1438	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
905	}	1439	}
906		1440
907	sv = sv_2mortal (sv);	1441	sv = sv_2mortal (sv);
908		1442
909	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1443	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
910	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1444	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
911		1445
912	return sv;	1446	return sv;
913	}	1447	}
914		1448
		1449	/////////////////////////////////////////////////////////////////////////////
		1450	// XS interface functions
		1451
915	MODULE = JSON::XS PACKAGE = JSON::XS	1452	MODULE = JSON::XS PACKAGE = JSON::XS
916		1453
917	BOOT:	1454	BOOT:
918	{	1455	{
919	int i;	1456	int i;
920		1457
921	memset (decode_hexdigit, 0xff, 256);
922	for (i = 10; i--; )	1458	for (i = 0; i < 256; ++i)
923	decode_hexdigit ['0' + i] = i;	1459	decode_hexdigit [i] =
		1460	i >= '0' && i <= '9' ? i - '0'
		1461	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1462	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1463	: -1;
924		1464
925	for (i = 7; i--; )
926	{
927	decode_hexdigit ['a' + i] = 10 + i;
928	decode_hexdigit ['A' + i] = 10 + i;
929	}
930
931	json_stash = gv_stashpv ("JSON::XS", 1);	1465	json_stash = gv_stashpv ("JSON::XS" , 1);
		1466	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1467
		1468	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1469	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
932	}	1470	}
933		1471
934	PROTOTYPES: DISABLE	1472	PROTOTYPES: DISABLE
935		1473
936	SV *new (char *dummy)	1474	void CLONE (...)
937	CODE:	1475	CODE:
938	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1476	json_stash = 0;
		1477	json_boolean_stash = 0;
		1478
		1479	void new (char *klass)
		1480	PPCODE:
		1481	{
		1482	SV *pv = NEWSV (0, sizeof (JSON));
		1483	SvPOK_only (pv);
		1484	Zero (SvPVX (pv), 1, JSON);
		1485	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1486	XPUSHs (sv_2mortal (sv_bless (
		1487	newRV_noinc (pv),
		1488	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1489	)));
		1490	}
		1491
		1492	void ascii (JSON *self, int enable = 1)
		1493	ALIAS:
		1494	ascii = F_ASCII
		1495	latin1 = F_LATIN1
		1496	utf8 = F_UTF8
		1497	indent = F_INDENT
		1498	canonical = F_CANONICAL
		1499	space_before = F_SPACE_BEFORE
		1500	space_after = F_SPACE_AFTER
		1501	pretty = F_PRETTY
		1502	allow_nonref = F_ALLOW_NONREF
		1503	shrink = F_SHRINK
		1504	allow_blessed = F_ALLOW_BLESSED
		1505	convert_blessed = F_CONV_BLESSED
		1506	relaxed = F_RELAXED
		1507	PPCODE:
		1508	{
		1509	if (enable)
		1510	self->flags |= ix;
		1511	else
		1512	self->flags &= ~ix;
		1513
		1514	XPUSHs (ST (0));
		1515	}
		1516
		1517	void get_ascii (JSON *self)
		1518	ALIAS:
		1519	get_ascii = F_ASCII
		1520	get_latin1 = F_LATIN1
		1521	get_utf8 = F_UTF8
		1522	get_indent = F_INDENT
		1523	get_canonical = F_CANONICAL
		1524	get_space_before = F_SPACE_BEFORE
		1525	get_space_after = F_SPACE_AFTER
		1526	get_allow_nonref = F_ALLOW_NONREF
		1527	get_shrink = F_SHRINK
		1528	get_allow_blessed = F_ALLOW_BLESSED
		1529	get_convert_blessed = F_CONV_BLESSED
		1530	get_relaxed = F_RELAXED
		1531	PPCODE:
		1532	XPUSHs (boolSV (self->flags & ix));
		1533
		1534	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1535	PPCODE:
		1536	{
		1537	UV log2 = 0;
		1538
		1539	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1540
		1541	while ((1UL << log2) < max_depth)
		1542	++log2;
		1543
		1544	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1545
		1546	XPUSHs (ST (0));
		1547	}
		1548
		1549	U32 get_max_depth (JSON *self)
		1550	CODE:
		1551	RETVAL = DEC_DEPTH (self->flags);
939	OUTPUT:	1552	OUTPUT:
940	RETVAL	1553	RETVAL
941		1554
942	SV *ascii (SV *self, int enable = 1)	1555	void max_size (JSON *self, UV max_size = 0)
943	ALIAS:	1556	PPCODE:
944	ascii = F_ASCII	1557	{
945	utf8 = F_UTF8	1558	UV log2 = 0;
946	indent = F_INDENT	1559
947	canonical = F_CANONICAL	1560	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
948	space_before = F_SPACE_BEFORE	1561	if (max_size == 1) max_size = 2;
949	space_after = F_SPACE_AFTER	1562
950	pretty = F_PRETTY	1563	while ((1UL << log2) < max_size)
951	allow_nonref = F_ALLOW_NONREF	1564	++log2;
952	shrink = F_SHRINK	1565
		1566	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1567
		1568	XPUSHs (ST (0));
		1569	}
		1570
		1571	int get_max_size (JSON *self)
953	CODE:	1572	CODE:
954	{	1573	RETVAL = DEC_SIZE (self->flags);
955	UV *uv = SvJSON (self);
956	if (enable)
957	*uv |= ix;
958	else
959	*uv &= ~ix;
960
961	RETVAL = newSVsv (self);
962	}
963	OUTPUT:	1574	OUTPUT:
964	RETVAL	1575	RETVAL
965		1576
966	void encode (SV *self, SV *scalar)	1577	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
967	PPCODE:	1578	PPCODE:
968	XPUSHs (encode_json (scalar, *SvJSON (self)));	1579	{
		1580	SvREFCNT_dec (self->cb_object);
		1581	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
969		1582
970	void decode (SV *self, SV *jsonstr)	1583	XPUSHs (ST (0));
		1584	}
		1585
		1586	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
971	PPCODE:	1587	PPCODE:
		1588	{
		1589	if (!self->cb_sk_object)
		1590	self->cb_sk_object = newHV ();
		1591
		1592	if (SvOK (cb))
		1593	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1594	else
		1595	{
		1596	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1597
		1598	if (!HvKEYS (self->cb_sk_object))
		1599	{
		1600	SvREFCNT_dec (self->cb_sk_object);
		1601	self->cb_sk_object = 0;
		1602	}
		1603	}
		1604
		1605	XPUSHs (ST (0));
		1606	}
		1607
		1608	void encode (JSON *self, SV *scalar)
		1609	PPCODE:
		1610	XPUSHs (encode_json (scalar, self));
		1611
		1612	void decode (JSON *self, SV *jsonstr)
		1613	PPCODE:
972	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1614	XPUSHs (decode_json (jsonstr, self, 0));
		1615
		1616	void decode_prefix (JSON *self, SV *jsonstr)
		1617	PPCODE:
		1618	{
		1619	UV offset;
		1620	EXTEND (SP, 2);
		1621	PUSHs (decode_json (jsonstr, self, &offset));
		1622	PUSHs (sv_2mortal (newSVuv (offset)));
		1623	}
		1624
		1625	void DESTROY (JSON *self)
		1626	CODE:
		1627	SvREFCNT_dec (self->cb_sk_object);
		1628	SvREFCNT_dec (self->cb_object);
973		1629
974	PROTOTYPES: ENABLE	1630	PROTOTYPES: ENABLE
975		1631
976	void to_json (SV *scalar)	1632	void encode_json (SV *scalar)
977	PPCODE:	1633	PPCODE:
		1634	{
		1635	JSON json = { F_DEFAULT | F_UTF8 };
978	XPUSHs (encode_json (scalar, F_UTF8));	1636	XPUSHs (encode_json (scalar, &json));
		1637	}
979		1638
980	void from_json (SV *jsonstr)	1639	void decode_json (SV *jsonstr)
981	PPCODE:	1640	PPCODE:
		1641	{
		1642	JSON json = { F_DEFAULT | F_UTF8 };
982	XPUSHs (decode_json (jsonstr, F_UTF8));	1643	XPUSHs (decode_json (jsonstr, &json, 0));
		1644	}
983		1645

Diff Legend

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