ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.14 by root, Sat Mar 24 22:55:43 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
22	#define INDENT_STEP 3 // spaces per indentation level	47	#define INDENT_STEP 3 // spaces per indentation level
23		48
24	#define UTF8_MAX_LEN 11 // for perls UTF-X: max. number of octets per character
25	#define SHORT_STRING_LEN 512 // special-case strings of up to this size	49	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
26		50
27	#define SB do {	51	#define SB do {
28	#define SE } while (0)	52	#define SE } while (0)
29		53
		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
30	static HV *json_stash; // JSON::XS::	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;
31		85
32	/////////////////////////////////////////////////////////////////////////////	86	/////////////////////////////////////////////////////////////////////////////
33	// utility functions	87	// utility functions
34		88
35	static UV *	89	INLINE void
36	SvJSON (SV *sv)
37	{
38	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
39	croak ("object is not of type JSON::XS");
40
41	return &SvUVX (SvRV (sv));
42	}
43
44	static void
45	shrink (SV *sv)	90	shrink (SV *sv)
46	{	91	{
47	sv_utf8_downgrade (sv, 1);	92	sv_utf8_downgrade (sv, 1);
48	if (SvLEN (sv) > SvCUR (sv) + 1)	93	if (SvLEN (sv) > SvCUR (sv) + 1)
49	{	94	{
…		…
58	// decode an utf-8 character and return it, or (UV)-1 in	103	// decode an utf-8 character and return it, or (UV)-1 in
59	// case of an error.	104	// case of an error.
60	// we special-case "safe" characters from U+80 .. U+7FF,	105	// we special-case "safe" characters from U+80 .. U+7FF,
61	// but use the very good perl function to parse anything else.	106	// but use the very good perl function to parse anything else.
62	// note that we never call this function for a ascii codepoints	107	// note that we never call this function for a ascii codepoints
63	static UV	108	INLINE UV
64	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)	109	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
65	{	110	{
66	if (s[0] > 0xdf || s[0] < 0xc2)	111	if (expect_true (len >= 2
67	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	112	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
68	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	113	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
69	{	114	{
70	*clen = 2;	115	*clen = 2;
71	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);	116	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
72	}	117	}
73	else	118	else
74	return (UV)-1;	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;
75	}	137	}
76		138
77	/////////////////////////////////////////////////////////////////////////////	139	/////////////////////////////////////////////////////////////////////////////
78	// encoder	140	// encoder
79		141
…		…
81	typedef struct	143	typedef struct
82	{	144	{
83	char *cur; // SvPVX (sv) + current output position	145	char *cur; // SvPVX (sv) + current output position
84	char *end; // SvEND (sv)	146	char *end; // SvEND (sv)
85	SV *sv; // result scalar	147	SV *sv; // result scalar
86	UV flags; // F_*	148	JSON json;
87	int indent; // indentation level	149	U32 indent; // indentation level
88	int max_depth; // max. recursion level	150	U32 maxdepth; // max. indentation/recursion level
		151	UV limit; // escape character values >= this value when encoding
89	} enc_t;	152	} enc_t;
90		153
91	static void	154	INLINE void
92	need (enc_t *enc, STRLEN len)	155	need (enc_t *enc, STRLEN len)
93	{	156	{
94	if (enc->cur + len >= enc->end)	157	if (expect_false (enc->cur + len >= enc->end))
95	{	158	{
96	STRLEN cur = enc->cur - SvPVX (enc->sv);	159	STRLEN cur = enc->cur - SvPVX (enc->sv);
97	SvGROW (enc->sv, cur + len + 1);	160	SvGROW (enc->sv, cur + len + 1);
98	enc->cur = SvPVX (enc->sv) + cur;	161	enc->cur = SvPVX (enc->sv) + cur;
99	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	162	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
100	}	163	}
101	}	164	}
102		165
103	static void	166	INLINE void
104	encode_ch (enc_t *enc, char ch)	167	encode_ch (enc_t *enc, char ch)
105	{	168	{
106	need (enc, 1);	169	need (enc, 1);
107	*enc->cur++ = ch;	170	*enc->cur++ = ch;
108	}	171	}
…		…
116		179
117	while (str < end)	180	while (str < end)
118	{	181	{
119	unsigned char ch = *(unsigned char *)str;	182	unsigned char ch = *(unsigned char *)str;
120		183
121	if (ch >= 0x20 && ch < 0x80) // most common case	184	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
122	{	185	{
123	if (ch == '"') // but with slow exceptions	186	if (expect_false (ch == '"')) // but with slow exceptions
124	{	187	{
125	need (enc, len += 1);	188	need (enc, len += 1);
126	*enc->cur++ = '\\';	189	*enc->cur++ = '\\';
127	*enc->cur++ = '"';	190	*enc->cur++ = '"';
128	}	191	}
129	else if (ch == '\\')	192	else if (expect_false (ch == '\\'))
130	{	193	{
131	need (enc, len += 1);	194	need (enc, len += 1);
132	*enc->cur++ = '\\';	195	*enc->cur++ = '\\';
133	*enc->cur++ = '\\';	196	*enc->cur++ = '\\';
134	}	197	}
…		…
152	STRLEN clen;	215	STRLEN clen;
153	UV uch;	216	UV uch;
154		217
155	if (is_utf8)	218	if (is_utf8)
156	{	219	{
157	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
158	uch = decode_utf8 (str, end - str, &clen);	220	uch = decode_utf8 (str, end - str, &clen);
159	if (clen == (STRLEN)-1)	221	if (clen == (STRLEN)-1)
160	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);
161	}	223	}
162	else	224	else
163	{	225	{
164	uch = ch;	226	uch = ch;
165	clen = 1;	227	clen = 1;
166	}	228	}
167		229
168	if (uch > 0x10FFFFUL)	230	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
169	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
170
171	if (uch < 0x80 || enc->flags & F_ASCII)
172	{	231	{
173	if (uch > 0xFFFFUL)	232	if (uch > 0xFFFFUL)
174	{	233	{
		234	if (uch > 0x10FFFFUL)
		235	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		236
175	need (enc, len += 11);	237	need (enc, len += 11);
176	sprintf (enc->cur, "\\u%04x\\u%04x",	238	sprintf (enc->cur, "\\u%04x\\u%04x",
177	(int)((uch - 0x10000) / 0x400 + 0xD800),	239	(int)((uch - 0x10000) / 0x400 + 0xD800),
178	(int)((uch - 0x10000) % 0x400 + 0xDC00));	240	(int)((uch - 0x10000) % 0x400 + 0xDC00));
179	enc->cur += 12;	241	enc->cur += 12;
…		…
190	*enc->cur++ = hexdigit [(uch >> 0) & 15];	252	*enc->cur++ = hexdigit [(uch >> 0) & 15];
191	}	253	}
192		254
193	str += clen;	255	str += clen;
194	}	256	}
		257	else if (enc->json.flags & F_LATIN1)
		258	{
		259	*enc->cur++ = uch;
		260	str += clen;
		261	}
195	else if (is_utf8)	262	else if (is_utf8)
196	{	263	{
197	need (enc, len += clen);	264	need (enc, len += clen);
198	do	265	do
199	{	266	{
…		…
201	}	268	}
202	while (--clen);	269	while (--clen);
203	}	270	}
204	else	271	else
205	{	272	{
206	need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed	273	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
207	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	274	enc->cur = encode_utf8 (enc->cur, uch);
208	++str;	275	++str;
209	}	276	}
210	}	277	}
211	}	278	}
212	}	279	}
213		280
214	--len;	281	--len;
215	}	282	}
216	}	283	}
217		284
218	static void	285	INLINE void
219	encode_indent (enc_t *enc)	286	encode_indent (enc_t *enc)
220	{	287	{
221	if (enc->flags & F_INDENT)	288	if (enc->json.flags & F_INDENT)
222	{	289	{
223	int spaces = enc->indent * INDENT_STEP;	290	int spaces = enc->indent * INDENT_STEP;
224		291
225	need (enc, spaces);	292	need (enc, spaces);
226	memset (enc->cur, ' ', spaces);	293	memset (enc->cur, ' ', spaces);
227	enc->cur += spaces;	294	enc->cur += spaces;
228	}	295	}
229	}	296	}
230		297
231	static void	298	INLINE void
232	encode_space (enc_t *enc)	299	encode_space (enc_t *enc)
233	{	300	{
234	need (enc, 1);	301	need (enc, 1);
235	encode_ch (enc, ' ');	302	encode_ch (enc, ' ');
236	}	303	}
237		304
238	static void	305	INLINE void
239	encode_nl (enc_t *enc)	306	encode_nl (enc_t *enc)
240	{	307	{
241	if (enc->flags & F_INDENT)	308	if (enc->json.flags & F_INDENT)
242	{	309	{
243	need (enc, 1);	310	need (enc, 1);
244	encode_ch (enc, '\n');	311	encode_ch (enc, '\n');
245	}	312	}
246	}	313	}
247		314
248	static void	315	INLINE void
249	encode_comma (enc_t *enc)	316	encode_comma (enc_t *enc)
250	{	317	{
251	encode_ch (enc, ',');	318	encode_ch (enc, ',');
252		319
253	if (enc->flags & F_INDENT)	320	if (enc->json.flags & F_INDENT)
254	encode_nl (enc);	321	encode_nl (enc);
255	else if (enc->flags & F_SPACE_AFTER)	322	else if (enc->json.flags & F_SPACE_AFTER)
256	encode_space (enc);	323	encode_space (enc);
257	}	324	}
258		325
259	static void encode_sv (enc_t *enc, SV *sv);	326	static void encode_sv (enc_t *enc, SV *sv);
260		327
261	static void	328	static void
262	encode_av (enc_t *enc, AV *av)	329	encode_av (enc_t *enc, AV *av)
263	{	330	{
264	int i, len = av_len (av);	331	int i, len = av_len (av);
265		332
266	encode_ch (enc, '['); encode_nl (enc);	333	if (enc->indent >= enc->maxdepth)
267	++enc->indent;	334	croak ("data structure too deep (hit recursion limit)");
268		335
		336	encode_ch (enc, '[');
		337
		338	if (len >= 0)
		339	{
		340	encode_nl (enc); ++enc->indent;
		341
269	for (i = 0; i <= len; ++i)	342	for (i = 0; i <= len; ++i)
270	{	343	{
		344	SV **svp = av_fetch (av, i, 0);
		345
271	encode_indent (enc);	346	encode_indent (enc);
272	encode_sv (enc, *av_fetch (av, i, 0));
273		347
		348	if (svp)
		349	encode_sv (enc, *svp);
		350	else
		351	encode_str (enc, "null", 4, 0);
		352
274	if (i < len)	353	if (i < len)
275	encode_comma (enc);	354	encode_comma (enc);
276	}	355	}
277		356
		357	encode_nl (enc); --enc->indent; encode_indent (enc);
		358	}
		359
278	encode_nl (enc);	360	encode_ch (enc, ']');
279
280	--enc->indent;
281	encode_indent (enc); encode_ch (enc, ']');
282	}	361	}
283		362
284	static void	363	static void
285	encode_he (enc_t *enc, HE *he)	364	encode_hk (enc_t *enc, HE *he)
286	{	365	{
287	encode_ch (enc, '"');	366	encode_ch (enc, '"');
288		367
289	if (HeKLEN (he) == HEf_SVKEY)	368	if (HeKLEN (he) == HEf_SVKEY)
290	{	369	{
…		…
300	else	379	else
301	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	380	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
302		381
303	encode_ch (enc, '"');	382	encode_ch (enc, '"');
304		383
305	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	384	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
306	encode_ch (enc, ':');	385	encode_ch (enc, ':');
307	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	386	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
308	encode_sv (enc, HeVAL (he));
309	}	387	}
310		388
311	// compare hash entries, used when all keys are bytestrings	389	// compare hash entries, used when all keys are bytestrings
312	static int	390	static int
313	he_cmp_fast (const void *a_, const void *b_)	391	he_cmp_fast (const void *a_, const void *b_)
…		…
318	HE *b = *(HE **)b_;	396	HE *b = *(HE **)b_;
319		397
320	STRLEN la = HeKLEN (a);	398	STRLEN la = HeKLEN (a);
321	STRLEN lb = HeKLEN (b);	399	STRLEN lb = HeKLEN (b);
322		400
323	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	401	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
324	cmp = la - lb;	402	cmp = lb - la;
325		403
326	return cmp;	404	return cmp;
327	}	405	}
328		406
329	// 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
330	static int	408	static int
331	he_cmp_slow (const void *a, const void *b)	409	he_cmp_slow (const void *a, const void *b)
332	{	410	{
333	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	411	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
334	}	412	}
335		413
336	static void	414	static void
337	encode_hv (enc_t *enc, HV *hv)	415	encode_hv (enc_t *enc, HV *hv)
338	{	416	{
		417	HE *he;
339	int count, i;	418	int count;
340		419
341	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	420	if (enc->indent >= enc->maxdepth)
		421	croak ("data structure too deep (hit recursion limit)");
342		422
343	if ((count = hv_iterinit (hv)))	423	encode_ch (enc, '{');
344	{	424
345	// for canonical output we have to sort by keys first	425	// for canonical output we have to sort by keys first
346	// actually, this is mostly due to the stupid so-called	426	// actually, this is mostly due to the stupid so-called
347	// security workaround added somewhere in 5.8.x.	427	// security workaround added somewhere in 5.8.x.
348	// that randomises hash orderings	428	// that randomises hash orderings
349	if (enc->flags & F_CANONICAL)	429	if (enc->json.flags & F_CANONICAL)
		430	{
		431	int count = hv_iterinit (hv);
		432
		433	if (SvMAGICAL (hv))
350	{	434	{
		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	{
		449	int i, fast = 1;
		450	#if defined(__BORLANDC__) || defined(_MSC_VER)
		451	HE **hes = _alloca (count * sizeof (HE));
		452	#else
351	HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode	453	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
352	int fast = 1;	454	#endif
353		455
354	i = 0;	456	i = 0;
355	while ((he = hv_iternext (hv)))	457	while ((he = hv_iternext (hv)))
356	{	458	{
357	hes [i++] = he;	459	hes [i++] = he;
…		…
379		481
380	FREETMPS;	482	FREETMPS;
381	LEAVE;	483	LEAVE;
382	}	484	}
383		485
384	for (i = 0; i < count; ++i)	486	encode_nl (enc); ++enc->indent;
		487
		488	while (count--)
385	{	489	{
386	encode_indent (enc);	490	encode_indent (enc);
		491	he = hes [count];
387	encode_he (enc, hes [i]);	492	encode_hk (enc, he);
		493	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
388		494
389	if (i < count - 1)	495	if (count)
390	encode_comma (enc);	496	encode_comma (enc);
391	}	497	}
392		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)))
		506	{
		507	encode_nl (enc); ++enc->indent;
		508
		509	for (;;)
		510	{
393	encode_nl (enc);	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);
394	}	549	}
395	else	550	else
396	{	551	{
397	SV *sv;	552	#if 0
398	HE *he = hv_iternext (hv);	553	if (0 && sv_derived_from (rv, "JSON::Literal"))
399
400	for (;;)
401	{	554	{
402	encode_indent (enc);	555	// not yet
403	encode_he (enc, he);
404
405	if (!(he = hv_iternext (hv)))
406	break;
407
408	encode_comma (enc);
409	}	556	}
		557	#endif
		558	if (enc->json.flags & F_CONV_BLESSED)
		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);
410		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
411	encode_nl (enc);	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))));
412	}	597	}
413	}	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;
414		607
415	--enc->indent; encode_indent (enc); 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))));
416	}	619	}
417		620
418	static void	621	static void
419	encode_sv (enc_t *enc, SV *sv)	622	encode_sv (enc_t *enc, SV *sv)
420	{	623	{
…		…
428	encode_str (enc, str, len, SvUTF8 (sv));	631	encode_str (enc, str, len, SvUTF8 (sv));
429	encode_ch (enc, '"');	632	encode_ch (enc, '"');
430	}	633	}
431	else if (SvNOKp (sv))	634	else if (SvNOKp (sv))
432	{	635	{
		636	// trust that perl will do the right thing w.r.t. JSON syntax.
433	need (enc, NV_DIG + 32);	637	need (enc, NV_DIG + 32);
434	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	638	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
435	enc->cur += strlen (enc->cur);	639	enc->cur += strlen (enc->cur);
436	}	640	}
437	else if (SvIOKp (sv))	641	else if (SvIOKp (sv))
438	{	642	{
439	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);
440	enc->cur +=	648	enc->cur +=
441	SvIsUV(sv)	649	SvIsUV(sv)
442	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	650	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
443	: 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	}
444	}	680	}
445	else if (SvROK (sv))	681	else if (SvROK (sv))
446	{	682	encode_rv (enc, SvRV (sv));
447	SV *rv = SvRV (sv);
448
449	if (enc->indent >= enc->max_depth)
450	croak ("data structure too deep (hit recursion limit)");
451
452	switch (SvTYPE (rv))
453	{
454	case SVt_PVAV: encode_av (enc, (AV *)rv); break;
455	case SVt_PVHV: encode_hv (enc, (HV *)rv); break;
456
457	default:
458	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
459	SvPV_nolen (sv));
460	}
461	}
462	else if (!SvOK (sv))	683	else if (!SvOK (sv))
463	encode_str (enc, "null", 4, 0);	684	encode_str (enc, "null", 4, 0);
464	else	685	else
465	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",
466	SvPV_nolen (sv), SvFLAGS (sv));	687	SvPV_nolen (sv), SvFLAGS (sv));
467	}	688	}
468		689
469	static SV *	690	static SV *
470	encode_json (SV *scalar, UV flags)	691	encode_json (SV *scalar, JSON *json)
471	{	692	{
		693	enc_t enc;
		694
472	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	695	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
473	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)");
474		697
475	enc_t enc;	698	enc.json = *json;
476	enc.flags = flags;
477	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	699	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
478	enc.cur = SvPVX (enc.sv);	700	enc.cur = SvPVX (enc.sv);
479	enc.end = SvEND (enc.sv);	701	enc.end = SvEND (enc.sv);
480	enc.indent = 0;	702	enc.indent = 0;
481	enc.max_depth = 0x7fffffffUL;	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;
482		707
483	SvPOK_only (enc.sv);	708	SvPOK_only (enc.sv);
484	encode_sv (&enc, scalar);	709	encode_sv (&enc, scalar);
485		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
486	if (!(flags & (F_ASCII | F_UTF8)))	714	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
487	SvUTF8_on (enc.sv);	715	SvUTF8_on (enc.sv);
488		716
489	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
490
491	if (enc.flags & F_SHRINK)	717	if (enc.json.flags & F_SHRINK)
492	shrink (enc.sv);	718	shrink (enc.sv);
493		719
494	return enc.sv;	720	return enc.sv;
495	}	721	}
496		722
…		…
501	typedef struct	727	typedef struct
502	{	728	{
503	char *cur; // current parser pointer	729	char *cur; // current parser pointer
504	char *end; // end of input string	730	char *end; // end of input string
505	const char *err; // parse error, if != 0	731	const char *err; // parse error, if != 0
506	UV flags; // F_*	732	JSON json;
		733	U32 depth; // recursion depth
		734	U32 maxdepth; // recursion depth limit
507	} dec_t;	735	} dec_t;
508		736
509	static void	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
510	decode_ws (dec_t *dec)	747	decode_ws (dec_t *dec)
511	{	748	{
512	for (;;)	749	for (;;)
513	{	750	{
514	char ch = *dec->cur;	751	char ch = *dec->cur;
515		752
516	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	}
517	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	765	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
518	break;	766	break; // parse error, but let higher level handle it, gives better error messages
519		767
520	++dec->cur;	768	++dec->cur;
521	}	769	}
522	}	770	}
523		771
524	#define ERR(reason) SB dec->err = reason; goto fail; SE	772	#define ERR(reason) SB dec->err = reason; goto fail; SE
		773
525	#define EXPECT_CH(ch) SB \	774	#define EXPECT_CH(ch) SB \
526	if (*dec->cur != ch) \	775	if (*dec->cur != ch) \
527	ERR (# ch " expected"); \	776	ERR (# ch " expected"); \
528	++dec->cur; \	777	++dec->cur; \
529	SE	778	SE
530		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
531	static SV *decode_sv (dec_t *dec);	783	static SV *decode_sv (dec_t *dec);
532		784
533	static signed char decode_hexdigit[256];	785	static signed char decode_hexdigit[256];
534		786
535	static UV	787	static UV
536	decode_4hex (dec_t *dec)	788	decode_4hex (dec_t *dec)
537	{	789	{
538	signed char d1, d2, d3, d4;	790	signed char d1, d2, d3, d4;
539	unsigned char *cur = (unsigned char *)dec->cur;	791	unsigned char *cur = (unsigned char *)dec->cur;
540		792
541	d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected");	793	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
542	d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected");	794	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
543	d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected");	795	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
544	d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected");	796	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
545		797
546	dec->cur += 4;	798	dec->cur += 4;
547		799
548	return ((UV)d1) << 12	800	return ((UV)d1) << 12
549	| ((UV)d2) << 8	801	| ((UV)d2) << 8
…		…
557	static SV *	809	static SV *
558	decode_str (dec_t *dec)	810	decode_str (dec_t *dec)
559	{	811	{
560	SV *sv = 0;	812	SV *sv = 0;
561	int utf8 = 0;	813	int utf8 = 0;
		814	char *dec_cur = dec->cur;
562		815
563	do	816	do
564	{	817	{
565	char buf [SHORT_STRING_LEN + UTF8_MAX_LEN];	818	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
566	char *cur = buf;	819	char *cur = buf;
567		820
568	do	821	do
569	{	822	{
570	unsigned char ch = *(unsigned char *)dec->cur++;	823	unsigned char ch = *(unsigned char *)dec_cur++;
571		824
572	if (ch == '"')	825	if (expect_false (ch == '"'))
573	{	826	{
574	--dec->cur;	827	--dec_cur;
575	break;	828	break;
576	}	829	}
577	else if (ch == '\\')	830	else if (expect_false (ch == '\\'))
578	{	831	{
579	switch (*dec->cur)	832	switch (*dec_cur)
580	{	833	{
581	case '\\':	834	case '\\':
582	case '/':	835	case '/':
583	case '"': *cur++ = *dec->cur++; break;	836	case '"': *cur++ = *dec_cur++; break;
584		837
585	case 'b': ++dec->cur; *cur++ = '\010'; break;	838	case 'b': ++dec_cur; *cur++ = '\010'; break;
586	case 't': ++dec->cur; *cur++ = '\011'; break;	839	case 't': ++dec_cur; *cur++ = '\011'; break;
587	case 'n': ++dec->cur; *cur++ = '\012'; break;	840	case 'n': ++dec_cur; *cur++ = '\012'; break;
588	case 'f': ++dec->cur; *cur++ = '\014'; break;	841	case 'f': ++dec_cur; *cur++ = '\014'; break;
589	case 'r': ++dec->cur; *cur++ = '\015'; break;	842	case 'r': ++dec_cur; *cur++ = '\015'; break;
590		843
591	case 'u':	844	case 'u':
592	{	845	{
593	UV lo, hi;	846	UV lo, hi;
594	++dec->cur;	847	++dec_cur;
595		848
		849	dec->cur = dec_cur;
596	hi = decode_4hex (dec);	850	hi = decode_4hex (dec);
		851	dec_cur = dec->cur;
597	if (hi == (UV)-1)	852	if (hi == (UV)-1)
598	goto fail;	853	goto fail;
599		854
600	// possibly a surrogate pair	855	// possibly a surrogate pair
601	if (hi >= 0xd800)	856	if (hi >= 0xd800)
602	if (hi < 0xdc00)	857	if (hi < 0xdc00)
603	{	858	{
604	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	859	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
605	ERR ("missing low surrogate character in surrogate pair");	860	ERR ("missing low surrogate character in surrogate pair");
606		861
607	dec->cur += 2;	862	dec_cur += 2;
608		863
		864	dec->cur = dec_cur;
609	lo = decode_4hex (dec);	865	lo = decode_4hex (dec);
		866	dec_cur = dec->cur;
610	if (lo == (UV)-1)	867	if (lo == (UV)-1)
611	goto fail;	868	goto fail;
612		869
613	if (lo < 0xdc00 || lo >= 0xe000)	870	if (lo < 0xdc00 || lo >= 0xe000)
614	ERR ("surrogate pair expected");	871	ERR ("surrogate pair expected");
…		…
620		877
621	if (hi >= 0x80)	878	if (hi >= 0x80)
622	{	879	{
623	utf8 = 1;	880	utf8 = 1;
624		881
625	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	882	cur = encode_utf8 (cur, hi);
626	}	883	}
627	else	884	else
628	*cur++ = hi;	885	*cur++ = hi;
629	}	886	}
630	break;	887	break;
631		888
632	default:	889	default:
633	--dec->cur;	890	--dec_cur;
634	ERR ("illegal backslash escape sequence in string");	891	ERR ("illegal backslash escape sequence in string");
635	}	892	}
636	}	893	}
637	else if (ch >= 0x20 && ch <= 0x7f)	894	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
638	*cur++ = ch;	895	*cur++ = ch;
639	else if (ch >= 0x80)	896	else if (ch >= 0x80)
640	{	897	{
641	--dec->cur;
642
643	STRLEN clen;	898	STRLEN clen;
		899	UV uch;
		900
		901	--dec_cur;
		902
644	UV uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	903	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
645	if (clen == (STRLEN)-1)	904	if (clen == (STRLEN)-1)
646	ERR ("malformed UTF-8 character in JSON string");	905	ERR ("malformed UTF-8 character in JSON string");
647		906
648	do	907	do
649	{
650	*cur++ = *dec->cur++;	908	*cur++ = *dec_cur++;
651	}
652	while (--clen);	909	while (--clen);
653		910
654	utf8 = 1;	911	utf8 = 1;
655	}	912	}
656	else if (!ch)
657	ERR ("unexpected end of string while parsing json string");
658	else	913	else
		914	{
		915	--dec_cur;
		916
		917	if (!ch)
		918	ERR ("unexpected end of string while parsing JSON string");
		919	else
659	ERR ("invalid character encountered");	920	ERR ("invalid character encountered while parsing JSON string");
660		921	}
661	}	922	}
662	while (cur < buf + SHORT_STRING_LEN);	923	while (cur < buf + SHORT_STRING_LEN);
663		924
		925	{
664	STRLEN len = cur - buf;	926	STRLEN len = cur - buf;
665		927
666	if (sv)	928	if (sv)
667	{	929	{
668	SvGROW (sv, SvCUR (sv) + len + 1);	930	SvGROW (sv, SvCUR (sv) + len + 1);
669	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	931	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
670	SvCUR_set (sv, SvCUR (sv) + len);	932	SvCUR_set (sv, SvCUR (sv) + len);
671	}	933	}
672	else	934	else
673	sv = newSVpvn (buf, len);	935	sv = newSVpvn (buf, len);
674	}	936	}
		937	}
675	while (*dec->cur != '"');	938	while (*dec_cur != '"');
676		939
677	++dec->cur;	940	++dec_cur;
678		941
679	if (sv)	942	if (sv)
680	{	943	{
681	SvPOK_only (sv);	944	SvPOK_only (sv);
682	*SvEND (sv) = 0;	945	*SvEND (sv) = 0;
…		…
685	SvUTF8_on (sv);	948	SvUTF8_on (sv);
686	}	949	}
687	else	950	else
688	sv = newSVpvn ("", 0);	951	sv = newSVpvn ("", 0);
689		952
		953	dec->cur = dec_cur;
690	return sv;	954	return sv;
691		955
692	fail:	956	fail:
		957	dec->cur = dec_cur;
693	return 0;	958	return 0;
694	}	959	}
695		960
696	static SV *	961	static SV *
697	decode_num (dec_t *dec)	962	decode_num (dec_t *dec)
…		…
755	is_nv = 1;	1020	is_nv = 1;
756	}	1021	}
757		1022
758	if (!is_nv)	1023	if (!is_nv)
759	{	1024	{
760	UV uv;	1025	int len = dec->cur - start;
761	int numtype = grok_number (start, dec->cur - start, &uv);	1026
762	if (numtype & IS_NUMBER_IN_UV)	1027	// special case the rather common 1..5-digit-int case
763	if (numtype & IS_NUMBER_NEG)	1028	if (*start == '-')
		1029	switch (len)
764	{	1030	{
765	if (uv < (UV)IV_MIN)	1031	case 2: return newSViv (-( start [1] - '0' * 1));
766	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));
767	}	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	}
768	else	1056	else
769	return newSVuv (uv);	1057	return newSVuv (uv);
770	}	1058	}
771		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
772	return newSVnv (Atof (start));	1076	return newSVnv (Atof (start));
773		1077
774	fail:	1078	fail:
775	return 0;	1079	return 0;
776	}	1080	}
…		…
778	static SV *	1082	static SV *
779	decode_av (dec_t *dec)	1083	decode_av (dec_t *dec)
780	{	1084	{
781	AV *av = newAV ();	1085	AV *av = newAV ();
782		1086
		1087	DEC_INC_DEPTH;
783	decode_ws (dec);	1088	decode_ws (dec);
		1089
784	if (*dec->cur == ']')	1090	if (*dec->cur == ']')
785	++dec->cur;	1091	++dec->cur;
786	else	1092	else
787	for (;;)	1093	for (;;)
788	{	1094	{
…		…
804		1110
805	if (*dec->cur != ',')	1111	if (*dec->cur != ',')
806	ERR (", or ] expected while parsing array");	1112	ERR (", or ] expected while parsing array");
807		1113
808	++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	}
809	}	1123	}
810		1124
		1125	DEC_DEC_DEPTH;
811	return newRV_noinc ((SV *)av);	1126	return newRV_noinc ((SV *)av);
812		1127
813	fail:	1128	fail:
814	SvREFCNT_dec (av);	1129	SvREFCNT_dec (av);
		1130	DEC_DEC_DEPTH;
815	return 0;	1131	return 0;
816	}	1132	}
817		1133
818	static SV *	1134	static SV *
819	decode_hv (dec_t *dec)	1135	decode_hv (dec_t *dec)
820	{	1136	{
		1137	SV *sv;
821	HV *hv = newHV ();	1138	HV *hv = newHV ();
822		1139
		1140	DEC_INC_DEPTH;
823	decode_ws (dec);	1141	decode_ws (dec);
		1142
824	if (*dec->cur == '}')	1143	if (*dec->cur == '}')
825	++dec->cur;	1144	++dec->cur;
826	else	1145	else
827	for (;;)	1146	for (;;)
828	{	1147	{
		1148	EXPECT_CH ('"');
		1149
		1150	// heuristic: assume that
		1151	// a) decode_str + hv_store_ent are abysmally slow.
		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	{
829	SV *key, *value;	1156	SV *value;
		1157	char *p = dec->cur;
		1158	char *e = p + 24; // only try up to 24 bytes
830		1159
831	decode_ws (dec); EXPECT_CH ('"');	1160	for (;;)
832
833	key = decode_str (dec);
834	if (!key)
835	goto fail;
836
837	decode_ws (dec); EXPECT_CH (':');
838
839	value = decode_sv (dec);
840	if (!value)
841	{	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);
842	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)
843	goto fail;	1197	goto fail;
		1198
		1199	hv_store (hv, key, len, value, 0);
		1200
		1201	break;
		1202	}
		1203
		1204	++p;
844	}	1205	}
845		1206	}
846	//TODO: optimise
847	hv_store_ent (hv, key, value, 0);
848		1207
849	decode_ws (dec);	1208	decode_ws (dec);
850		1209
851	if (*dec->cur == '}')	1210	if (*dec->cur == '}')
852	{	1211	{
…		…
856		1215
857	if (*dec->cur != ',')	1216	if (*dec->cur != ',')
858	ERR (", or } expected while parsing object/hash");	1217	ERR (", or } expected while parsing object/hash");
859		1218
860	++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	}
861	}	1228	}
862		1229
		1230	DEC_DEC_DEPTH;
863	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;
864		1292
865	fail:	1293	fail:
866	SvREFCNT_dec (hv);	1294	SvREFCNT_dec (hv);
		1295	DEC_DEC_DEPTH;
867	return 0;	1296	return 0;
868	}	1297	}
869		1298
870	static SV *	1299	static SV *
871	decode_sv (dec_t *dec)	1300	decode_sv (dec_t *dec)
872	{	1301	{
873	decode_ws (dec);	1302	// the beauty of JSON: you need exactly one character lookahead
		1303	// to parse everything.
874	switch (*dec->cur)	1304	switch (*dec->cur)
875	{	1305	{
876	case '"': ++dec->cur; return decode_str (dec);	1306	case '"': ++dec->cur; return decode_str (dec);
877	case '[': ++dec->cur; return decode_av (dec);	1307	case '[': ++dec->cur; return decode_av (dec);
878	case '{': ++dec->cur; return decode_hv (dec);	1308	case '{': ++dec->cur; return decode_hv (dec);
879		1309
880	case '-':	1310	case '-':
881	case '0': case '1': case '2': case '3': case '4':	1311	case '0': case '1': case '2': case '3': case '4':
882	case '5': case '6': case '7': case '8': case '9':	1312	case '5': case '6': case '7': case '8': case '9':
883	return decode_num (dec);	1313	return decode_num (dec);
884		1314
885	case 't':	1315	case 't':
886	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1316	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
887	{	1317	{
888	dec->cur += 4;	1318	dec->cur += 4;
889	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);
890	}	1323	}
891	else	1324	else
892	ERR ("'true' expected");	1325	ERR ("'true' expected");
893		1326
894	break;	1327	break;
895		1328
896	case 'f':	1329	case 'f':
897	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1330	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
898	{	1331	{
899	dec->cur += 5;	1332	dec->cur += 5;
900	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);
901	}	1337	}
902	else	1338	else
903	ERR ("'false' expected");	1339	ERR ("'false' expected");
904		1340
905	break;	1341	break;
…		…
914	ERR ("'null' expected");	1350	ERR ("'null' expected");
915		1351
916	break;	1352	break;
917		1353
918	default:	1354	default:
919	ERR ("malformed json string, neither array, object, number, string or atom");	1355	ERR ("malformed JSON string, neither array, object, number, string or atom");
920	break;	1356	break;
921	}	1357	}
922		1358
923	fail:	1359	fail:
924	return 0;	1360	return 0;
925	}	1361	}
926		1362
927	static SV *	1363	static SV *
928	decode_json (SV *string, UV flags)	1364	decode_json (SV *string, JSON *json, UV *offset_return)
929	{	1365	{
		1366	dec_t dec;
		1367	UV offset;
930	SV *sv;	1368	SV *sv;
931		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
932	if (flags & F_UTF8)	1377	if (json->flags & F_UTF8)
933	sv_utf8_downgrade (string, 0);	1378	sv_utf8_downgrade (string, 0);
934	else	1379	else
935	sv_utf8_upgrade (string);	1380	sv_utf8_upgrade (string);
936		1381
937	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1382	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
938		1383
939	dec_t dec;	1384	dec.json = *json;
940	dec.flags = flags;
941	dec.cur = SvPVX (string);	1385	dec.cur = SvPVX (string);
942	dec.end = SvEND (string);	1386	dec.end = SvEND (string);
943	dec.err = 0;	1387	dec.err = 0;
		1388	dec.depth = 0;
		1389	dec.maxdepth = DEC_DEPTH (dec.json.flags);
944		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);
945	sv = decode_sv (&dec);	1397	sv = decode_sv (&dec);
946		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
947	if (!sv)	1422	if (!sv)
948	{	1423	{
949	IV offset = dec.flags & F_UTF8
950	? dec.cur - SvPVX (string)
951	: utf8_distance (dec.cur, SvPVX (string));
952	SV *uni = sv_newmortal ();	1424	SV *uni = sv_newmortal ();
953		1425
954	// horrible hack to silence warning inside pv_uni_display	1426	// horrible hack to silence warning inside pv_uni_display
955	COP cop = *PL_curcop;	1427	COP cop = *PL_curcop;
956	cop.cop_warnings = pWARN_NONE;	1428	cop.cop_warnings = pWARN_NONE;
…		…
958	SAVEVPTR (PL_curcop);	1430	SAVEVPTR (PL_curcop);
959	PL_curcop = &cop;	1431	PL_curcop = &cop;
960	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);
961	LEAVE;	1433	LEAVE;
962		1434
963	croak ("%s, at character offset %d (%s)",	1435	croak ("%s, at character offset %d [\"%s\"]",
964	dec.err,	1436	dec.err,
965	(int)offset,	1437	(int)offset,
966	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1438	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
967	}	1439	}
968		1440
969	sv = sv_2mortal (sv);	1441	sv = sv_2mortal (sv);
970		1442
971	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1443	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
972	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)");
973		1445
974	return sv;	1446	return sv;
975	}	1447	}
976		1448
…		…
981		1453
982	BOOT:	1454	BOOT:
983	{	1455	{
984	int i;	1456	int i;
985		1457
986	memset (decode_hexdigit, 0xff, 256);
987	for (i = 10; i--; )	1458	for (i = 0; i < 256; ++i)
988	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;
989		1464
990	for (i = 7; i--; )
991	{
992	decode_hexdigit ['a' + i] = 10 + i;
993	decode_hexdigit ['A' + i] = 10 + i;
994	}
995
996	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);
997	}	1470	}
998		1471
999	PROTOTYPES: DISABLE	1472	PROTOTYPES: DISABLE
1000		1473
1001	SV *new (char *dummy)	1474	void CLONE (...)
1002	CODE:	1475	CODE:
1003	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);
1004	OUTPUT:	1552	OUTPUT:
1005	RETVAL	1553	RETVAL
1006		1554
1007	SV *ascii (SV *self, int enable = 1)	1555	void max_size (JSON *self, UV max_size = 0)
1008	ALIAS:	1556	PPCODE:
1009	ascii = F_ASCII	1557	{
1010	utf8 = F_UTF8	1558	UV log2 = 0;
1011	indent = F_INDENT	1559
1012	canonical = F_CANONICAL	1560	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1013	space_before = F_SPACE_BEFORE	1561	if (max_size == 1) max_size = 2;
1014	space_after = F_SPACE_AFTER	1562
1015	pretty = F_PRETTY	1563	while ((1UL << log2) < max_size)
1016	allow_nonref = F_ALLOW_NONREF	1564	++log2;
1017	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)
1018	CODE:	1572	CODE:
1019	{	1573	RETVAL = DEC_SIZE (self->flags);
1020	UV *uv = SvJSON (self);
1021	if (enable)
1022	*uv |= ix;
1023	else
1024	*uv &= ~ix;
1025
1026	RETVAL = newSVsv (self);
1027	}
1028	OUTPUT:	1574	OUTPUT:
1029	RETVAL	1575	RETVAL
1030		1576
1031	void encode (SV *self, SV *scalar)	1577	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1032	PPCODE:	1578	PPCODE:
1033	XPUSHs (encode_json (scalar, *SvJSON (self)));	1579	{
		1580	SvREFCNT_dec (self->cb_object);
		1581	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
1034		1582
1035	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)
1036	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:
1037	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);
1038		1629
1039	PROTOTYPES: ENABLE	1630	PROTOTYPES: ENABLE
1040		1631
1041	void to_json (SV *scalar)	1632	void encode_json (SV *scalar)
1042	PPCODE:	1633	PPCODE:
		1634	{
		1635	JSON json = { F_DEFAULT | F_UTF8 };
1043	XPUSHs (encode_json (scalar, F_UTF8));	1636	XPUSHs (encode_json (scalar, &json));
		1637	}
1044		1638
1045	void from_json (SV *jsonstr)	1639	void decode_json (SV *jsonstr)
1046	PPCODE:	1640	PPCODE:
		1641	{
		1642	JSON json = { F_DEFAULT | F_UTF8 };
1047	XPUSHs (decode_json (jsonstr, F_UTF8));	1643	XPUSHs (decode_json (jsonstr, &json, 0));
		1644	}
1048		1645

Diff Legend

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