ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.18 by root, Sun Mar 25 21:19:13 2007 UTC vs.
Revision 1.108 by root, Tue Aug 17 23:27:36 2010 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 <limits.h>
		10	#include <float.h>
8		11
		12	#if defined(__BORLANDC__) || defined(_MSC_VER)
		13	# define snprintf _snprintf // C compilers have this in stdio.h
		14	#endif
		15
		16	// some old perls do not have this, try to make it work, no
		17	// guarantees, though. if it breaks, you get to keep the pieces.
		18	#ifndef UTF8_MAXBYTES
		19	# define UTF8_MAXBYTES 13
		20	#endif
		21
		22	// three extra for rounding, sign, and end of string
		23	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
		24
9	#define F_ASCII 0x00000001UL	25	#define F_ASCII 0x00000001UL
		26	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002UL	27	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004UL	28	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008UL	29	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010UL	30	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020UL	31	#define F_SPACE_AFTER 0x00000040UL
15	#define F_ALLOW_NONREF 0x00000080UL	32	#define F_ALLOW_NONREF 0x00000100UL
16	#define F_SHRINK 0x00000100UL	33	#define F_SHRINK 0x00000200UL
		34	#define F_ALLOW_BLESSED 0x00000400UL
		35	#define F_CONV_BLESSED 0x00000800UL
17	#define F_MAXDEPTH 0xf8000000UL	36	#define F_RELAXED 0x00001000UL
18	#define S_MAXDEPTH 27	37	#define F_ALLOW_UNKNOWN 0x00002000UL
19		38	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
20	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
21
22	// F_SELFCONVERT? <=> to_json/toJson
23	// F_BLESSED? <=> { $__class__$ => }
24		39
25	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	40	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
26	#define F_DEFAULT (13UL << S_MAXDEPTH)
27		41
28	#define INIT_SIZE 32 // initial scalar size to be allocated	42	#define INIT_SIZE 32 // initial scalar size to be allocated
29	#define INDENT_STEP 3 // spaces per indentation level	43	#define INDENT_STEP 3 // spaces per indentation level
30		44
31	#define UTF8_MAX_LEN 11 // for perls UTF-X: max. number of octets per character
32	#define SHORT_STRING_LEN 512 // special-case strings of up to this size	45	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
		46
		47	#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
33		48
34	#define SB do {	49	#define SB do {
35	#define SE } while (0)	50	#define SE } while (0)
36		51
		52	#if __GNUC__ >= 3
		53	# define expect(expr,value) __builtin_expect ((expr), (value))
		54	# define INLINE static inline
		55	#else
		56	# define expect(expr,value) (expr)
		57	# define INLINE static
		58	#endif
		59
		60	#define expect_false(expr) expect ((expr) != 0, 0)
		61	#define expect_true(expr) expect ((expr) != 0, 1)
		62
		63	#define IN_RANGE_INC(type,val,beg,end) \
		64	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		65	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		66
		67	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
		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
37	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	enum {
		81	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		82	INCR_M_STR, // inside string
		83	INCR_M_BS, // inside backslash
		84	INCR_M_C0, // inside comment in initial whitespace sequence
		85	INCR_M_C1, // inside comment in other places
		86	INCR_M_JSON // outside anything, count nesting
		87	};
		88
		89	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		90
		91	typedef struct {
		92	U32 flags;
		93	U32 max_depth;
		94	STRLEN max_size;
		95
		96	SV *cb_object;
		97	HV *cb_sk_object;
		98
		99	// for the incremental parser
		100	SV *incr_text; // the source text so far
		101	STRLEN incr_pos; // the current offset into the text
		102	int incr_nest; // {[]}-nesting level
		103	unsigned char incr_mode;
		104	} JSON;
		105
		106	INLINE void
		107	json_init (JSON *json)
		108	{
		109	Zero (json, 1, JSON);
		110	json->max_depth = 512;
		111	}
38		112
39	/////////////////////////////////////////////////////////////////////////////	113	/////////////////////////////////////////////////////////////////////////////
40	// utility functions	114	// utility functions
41		115
42	static UV *	116	INLINE SV *
43	SvJSON (SV *sv)	117	get_bool (const char *name)
44	{	118	{
45	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))	119	SV *sv = get_sv (name, 1);
46	croak ("object is not of type JSON::XS");
47		120
48	return &SvUVX (SvRV (sv));	121	SvREADONLY_on (sv);
49	}	122	SvREADONLY_on (SvRV (sv));
50		123
51	static void	124	return sv;
		125	}
		126
		127	INLINE void
52	shrink (SV *sv)	128	shrink (SV *sv)
53	{	129	{
54	sv_utf8_downgrade (sv, 1);	130	sv_utf8_downgrade (sv, 1);
		131
55	if (SvLEN (sv) > SvCUR (sv) + 1)	132	if (SvLEN (sv) > SvCUR (sv) + 1)
56	{	133	{
57	#ifdef SvPV_shrink_to_cur	134	#ifdef SvPV_shrink_to_cur
58	SvPV_shrink_to_cur (sv);	135	SvPV_shrink_to_cur (sv);
59	#elif defined (SvPV_renew)	136	#elif defined (SvPV_renew)
…		…
65	// decode an utf-8 character and return it, or (UV)-1 in	142	// decode an utf-8 character and return it, or (UV)-1 in
66	// case of an error.	143	// case of an error.
67	// we special-case "safe" characters from U+80 .. U+7FF,	144	// we special-case "safe" characters from U+80 .. U+7FF,
68	// but use the very good perl function to parse anything else.	145	// but use the very good perl function to parse anything else.
69	// note that we never call this function for a ascii codepoints	146	// note that we never call this function for a ascii codepoints
70	static UV	147	INLINE UV
71	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)	148	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
72	{	149	{
73	if (s[0] > 0xdf || s[0] < 0xc2)	150	if (expect_true (len >= 2
74	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	151	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
75	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	152	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
76	{	153	{
77	*clen = 2;	154	*clen = 2;
78	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);	155	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
79	}	156	}
80	else	157	else
81	return (UV)-1;	158	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
82	}	159	}
83		160
		161	// likewise for encoding, also never called for ascii codepoints
		162	// this function takes advantage of this fact, although current gccs
		163	// seem to optimise the check for >= 0x80 away anyways
		164	INLINE unsigned char *
		165	encode_utf8 (unsigned char *s, UV ch)
		166	{
		167	if (expect_false (ch < 0x000080))
		168	*s++ = ch;
		169	else if (expect_true (ch < 0x000800))
		170	*s++ = 0xc0 | ( ch >> 6),
		171	*s++ = 0x80 | ( ch & 0x3f);
		172	else if ( ch < 0x010000)
		173	*s++ = 0xe0 | ( ch >> 12),
		174	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		175	*s++ = 0x80 | ( ch & 0x3f);
		176	else if ( ch < 0x110000)
		177	*s++ = 0xf0 | ( ch >> 18),
		178	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		179	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		180	*s++ = 0x80 | ( ch & 0x3f);
		181
		182	return s;
		183	}
		184
		185	// convert offset pointer to character index, sv must be string
		186	static STRLEN
		187	ptr_to_index (SV *sv, char *offset)
		188	{
		189	return SvUTF8 (sv)
		190	? utf8_distance (offset, SvPVX (sv))
		191	: offset - SvPVX (sv);
		192	}
		193
		194	/////////////////////////////////////////////////////////////////////////////
		195	// fp hell
		196
		197	// scan a group of digits, and a trailing exponent
		198	static void
		199	json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
		200	{
		201	UV uaccum = 0;
		202	int eaccum = 0;
		203
		204	// if we recurse too deep, skip all remaining digits
		205	// to avoid a stack overflow attack
		206	if (expect_false (--maxdepth <= 0))
		207	while (((U8)*s - '0') < 10)
		208	++s;
		209
		210	for (;;)
		211	{
		212	U8 dig = (U8)*s - '0';
		213
		214	if (expect_false (dig >= 10))
		215	{
		216	if (dig == (U8)((U8)'.' - (U8)'0'))
		217	{
		218	++s;
		219	json_atof_scan1 (s, accum, expo, 1, maxdepth);
		220	}
		221	else if ((dig | ' ') == 'e' - '0')
		222	{
		223	int exp2 = 0;
		224	int neg = 0;
		225
		226	++s;
		227
		228	if (*s == '-')
		229	{
		230	++s;
		231	neg = 1;
		232	}
		233	else if (*s == '+')
		234	++s;
		235
		236	while ((dig = (U8)*s - '0') < 10)
		237	exp2 = exp2 * 10 + *s++ - '0';
		238
		239	*expo += neg ? -exp2 : exp2;
		240	}
		241
		242	break;
		243	}
		244
		245	++s;
		246
		247	uaccum = uaccum * 10 + dig;
		248	++eaccum;
		249
		250	// if we have too many digits, then recurse for more
		251	// we actually do this for rather few digits
		252	if (uaccum >= (UV_MAX - 9) / 10)
		253	{
		254	if (postdp) *expo -= eaccum;
		255	json_atof_scan1 (s, accum, expo, postdp, maxdepth);
		256	if (postdp) *expo += eaccum;
		257
		258	break;
		259	}
		260	}
		261
		262	// this relies greatly on the quality of the pow ()
		263	// implementation of the platform, but a good
		264	// implementation is hard to beat.
		265	if (postdp) *expo -= eaccum;
		266	*accum += uaccum * Perl_pow (10., *expo);
		267	*expo += eaccum;
		268	}
		269
		270	static NV
		271	json_atof (const char *s)
		272	{
		273	NV accum = 0.;
		274	int expo = 0;
		275	int neg = 0;
		276
		277	if (*s == '-')
		278	{
		279	++s;
		280	neg = 1;
		281	}
		282
		283	// a recursion depth of ten gives us >>500 bits
		284	json_atof_scan1 (s, &accum, &expo, 0, 10);
		285
		286	return neg ? -accum : accum;
		287	}
84	/////////////////////////////////////////////////////////////////////////////	288	/////////////////////////////////////////////////////////////////////////////
85	// encoder	289	// encoder
86		290
87	// structure used for encoding JSON	291	// structure used for encoding JSON
88	typedef struct	292	typedef struct
89	{	293	{
90	char *cur; // SvPVX (sv) + current output position	294	char *cur; // SvPVX (sv) + current output position
91	char *end; // SvEND (sv)	295	char *end; // SvEND (sv)
92	SV *sv; // result scalar	296	SV *sv; // result scalar
93	U32 flags; // F_*	297	JSON json;
94	U32 indent; // indentation level	298	U32 indent; // indentation level
95	U32 maxdepth; // max. indentation/recursion level	299	UV limit; // escape character values >= this value when encoding
96	} enc_t;	300	} enc_t;
97		301
98	static void	302	INLINE void
99	need (enc_t *enc, STRLEN len)	303	need (enc_t *enc, STRLEN len)
100	{	304	{
101	if (enc->cur + len >= enc->end)	305	if (expect_false (enc->cur + len >= enc->end))
102	{	306	{
103	STRLEN cur = enc->cur - SvPVX (enc->sv);	307	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
104	SvGROW (enc->sv, cur + len + 1);	308	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
105	enc->cur = SvPVX (enc->sv) + cur;	309	enc->cur = SvPVX (enc->sv) + cur;
106	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	310	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
107	}	311	}
108	}	312	}
109		313
110	static void	314	INLINE void
111	encode_ch (enc_t *enc, char ch)	315	encode_ch (enc_t *enc, char ch)
112	{	316	{
113	need (enc, 1);	317	need (enc, 1);
114	*enc->cur++ = ch;	318	*enc->cur++ = ch;
115	}	319	}
…		…
123		327
124	while (str < end)	328	while (str < end)
125	{	329	{
126	unsigned char ch = *(unsigned char *)str;	330	unsigned char ch = *(unsigned char *)str;
127		331
128	if (ch >= 0x20 && ch < 0x80) // most common case	332	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
129	{	333	{
130	if (ch == '"') // but with slow exceptions	334	if (expect_false (ch == '"')) // but with slow exceptions
131	{	335	{
132	need (enc, len += 1);	336	need (enc, len += 1);
133	*enc->cur++ = '\\';	337	*enc->cur++ = '\\';
134	*enc->cur++ = '"';	338	*enc->cur++ = '"';
135	}	339	}
136	else if (ch == '\\')	340	else if (expect_false (ch == '\\'))
137	{	341	{
138	need (enc, len += 1);	342	need (enc, len += 1);
139	*enc->cur++ = '\\';	343	*enc->cur++ = '\\';
140	*enc->cur++ = '\\';	344	*enc->cur++ = '\\';
141	}	345	}
…		…
159	STRLEN clen;	363	STRLEN clen;
160	UV uch;	364	UV uch;
161		365
162	if (is_utf8)	366	if (is_utf8)
163	{	367	{
164	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
165	uch = decode_utf8 (str, end - str, &clen);	368	uch = decode_utf8 (str, end - str, &clen);
166	if (clen == (STRLEN)-1)	369	if (clen == (STRLEN)-1)
167	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	370	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
168	}	371	}
169	else	372	else
170	{	373	{
171	uch = ch;	374	uch = ch;
172	clen = 1;	375	clen = 1;
173	}	376	}
174		377
175	if (uch > 0x10FFFFUL)	378	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
176	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
177
178	if (uch < 0x80 || enc->flags & F_ASCII)
179	{	379	{
180	if (uch > 0xFFFFUL)	380	if (uch >= 0x10000UL)
181	{	381	{
		382	if (uch >= 0x110000UL)
		383	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		384
182	need (enc, len += 11);	385	need (enc, len += 11);
183	sprintf (enc->cur, "\\u%04x\\u%04x",	386	sprintf (enc->cur, "\\u%04x\\u%04x",
184	(int)((uch - 0x10000) / 0x400 + 0xD800),	387	(int)((uch - 0x10000) / 0x400 + 0xD800),
185	(int)((uch - 0x10000) % 0x400 + 0xDC00));	388	(int)((uch - 0x10000) % 0x400 + 0xDC00));
186	enc->cur += 12;	389	enc->cur += 12;
187	}	390	}
188	else	391	else
189	{	392	{
190	static char hexdigit [16] = "0123456789abcdef";
191	need (enc, len += 5);	393	need (enc, len += 5);
192	*enc->cur++ = '\\';	394	*enc->cur++ = '\\';
193	*enc->cur++ = 'u';	395	*enc->cur++ = 'u';
194	*enc->cur++ = hexdigit [ uch >> 12 ];	396	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
195	*enc->cur++ = hexdigit [(uch >> 8) & 15];	397	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
196	*enc->cur++ = hexdigit [(uch >> 4) & 15];	398	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
197	*enc->cur++ = hexdigit [(uch >> 0) & 15];	399	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
198	}	400	}
199		401
		402	str += clen;
		403	}
		404	else if (enc->json.flags & F_LATIN1)
		405	{
		406	*enc->cur++ = uch;
200	str += clen;	407	str += clen;
201	}	408	}
202	else if (is_utf8)	409	else if (is_utf8)
203	{	410	{
204	need (enc, len += clen);	411	need (enc, len += clen);
…		…
208	}	415	}
209	while (--clen);	416	while (--clen);
210	}	417	}
211	else	418	else
212	{	419	{
213	need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed	420	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
214	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	421	enc->cur = encode_utf8 (enc->cur, uch);
215	++str;	422	++str;
216	}	423	}
217	}	424	}
218	}	425	}
219	}	426	}
220		427
221	--len;	428	--len;
222	}	429	}
223	}	430	}
224		431
225	static void	432	INLINE void
226	encode_indent (enc_t *enc)	433	encode_indent (enc_t *enc)
227	{	434	{
228	if (enc->flags & F_INDENT)	435	if (enc->json.flags & F_INDENT)
229	{	436	{
230	int spaces = enc->indent * INDENT_STEP;	437	int spaces = enc->indent * INDENT_STEP;
231		438
232	need (enc, spaces);	439	need (enc, spaces);
233	memset (enc->cur, ' ', spaces);	440	memset (enc->cur, ' ', spaces);
234	enc->cur += spaces;	441	enc->cur += spaces;
235	}	442	}
236	}	443	}
237		444
238	static void	445	INLINE void
239	encode_space (enc_t *enc)	446	encode_space (enc_t *enc)
240	{	447	{
241	need (enc, 1);	448	need (enc, 1);
242	encode_ch (enc, ' ');	449	encode_ch (enc, ' ');
243	}	450	}
244		451
245	static void	452	INLINE void
246	encode_nl (enc_t *enc)	453	encode_nl (enc_t *enc)
247	{	454	{
248	if (enc->flags & F_INDENT)	455	if (enc->json.flags & F_INDENT)
249	{	456	{
250	need (enc, 1);	457	need (enc, 1);
251	encode_ch (enc, '\n');	458	encode_ch (enc, '\n');
252	}	459	}
253	}	460	}
254		461
255	static void	462	INLINE void
256	encode_comma (enc_t *enc)	463	encode_comma (enc_t *enc)
257	{	464	{
258	encode_ch (enc, ',');	465	encode_ch (enc, ',');
259		466
260	if (enc->flags & F_INDENT)	467	if (enc->json.flags & F_INDENT)
261	encode_nl (enc);	468	encode_nl (enc);
262	else if (enc->flags & F_SPACE_AFTER)	469	else if (enc->json.flags & F_SPACE_AFTER)
263	encode_space (enc);	470	encode_space (enc);
264	}	471	}
265		472
266	static void encode_sv (enc_t *enc, SV *sv);	473	static void encode_sv (enc_t *enc, SV *sv);
267		474
268	static void	475	static void
269	encode_av (enc_t *enc, AV *av)	476	encode_av (enc_t *enc, AV *av)
270	{	477	{
271	int i, len = av_len (av);	478	int i, len = av_len (av);
272		479
273	encode_ch (enc, '['); encode_nl (enc);	480	if (enc->indent >= enc->json.max_depth)
274	++enc->indent;	481	croak (ERR_NESTING_EXCEEDED);
275		482
		483	encode_ch (enc, '[');
		484
		485	if (len >= 0)
		486	{
		487	encode_nl (enc); ++enc->indent;
		488
276	for (i = 0; i <= len; ++i)	489	for (i = 0; i <= len; ++i)
277	{	490	{
		491	SV **svp = av_fetch (av, i, 0);
		492
278	encode_indent (enc);	493	encode_indent (enc);
279	encode_sv (enc, *av_fetch (av, i, 0));
280		494
		495	if (svp)
		496	encode_sv (enc, *svp);
		497	else
		498	encode_str (enc, "null", 4, 0);
		499
281	if (i < len)	500	if (i < len)
282	encode_comma (enc);	501	encode_comma (enc);
283	}	502	}
284		503
		504	encode_nl (enc); --enc->indent; encode_indent (enc);
		505	}
		506
285	encode_nl (enc);	507	encode_ch (enc, ']');
286
287	--enc->indent;
288	encode_indent (enc); encode_ch (enc, ']');
289	}	508	}
290		509
291	static void	510	static void
292	encode_he (enc_t *enc, HE *he)	511	encode_hk (enc_t *enc, HE *he)
293	{	512	{
294	encode_ch (enc, '"');	513	encode_ch (enc, '"');
295		514
296	if (HeKLEN (he) == HEf_SVKEY)	515	if (HeKLEN (he) == HEf_SVKEY)
297	{	516	{
…		…
307	else	526	else
308	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	527	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
309		528
310	encode_ch (enc, '"');	529	encode_ch (enc, '"');
311		530
312	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	531	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
313	encode_ch (enc, ':');	532	encode_ch (enc, ':');
314	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	533	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
315	encode_sv (enc, HeVAL (he));
316	}	534	}
317		535
318	// compare hash entries, used when all keys are bytestrings	536	// compare hash entries, used when all keys are bytestrings
319	static int	537	static int
320	he_cmp_fast (const void *a_, const void *b_)	538	he_cmp_fast (const void *a_, const void *b_)
…		…
325	HE *b = *(HE **)b_;	543	HE *b = *(HE **)b_;
326		544
327	STRLEN la = HeKLEN (a);	545	STRLEN la = HeKLEN (a);
328	STRLEN lb = HeKLEN (b);	546	STRLEN lb = HeKLEN (b);
329		547
330	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	548	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
331	cmp = la - lb;	549	cmp = lb - la;
332		550
333	return cmp;	551	return cmp;
334	}	552	}
335		553
336	// compare hash entries, used when some keys are sv's or utf-x	554	// compare hash entries, used when some keys are sv's or utf-x
337	static int	555	static int
338	he_cmp_slow (const void *a, const void *b)	556	he_cmp_slow (const void *a, const void *b)
339	{	557	{
340	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	558	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
341	}	559	}
342		560
343	static void	561	static void
344	encode_hv (enc_t *enc, HV *hv)	562	encode_hv (enc_t *enc, HV *hv)
345	{	563	{
346	int count, i;	564	HE *he;
347		565
348	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	566	if (enc->indent >= enc->json.max_depth)
		567	croak (ERR_NESTING_EXCEEDED);
349		568
350	if ((count = hv_iterinit (hv)))	569	encode_ch (enc, '{');
351	{	570
352	// for canonical output we have to sort by keys first	571	// for canonical output we have to sort by keys first
353	// actually, this is mostly due to the stupid so-called	572	// actually, this is mostly due to the stupid so-called
354	// security workaround added somewhere in 5.8.x.	573	// security workaround added somewhere in 5.8.x
355	// that randomises hash orderings	574	// that randomises hash orderings
356	if (enc->flags & F_CANONICAL)	575	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
		576	{
		577	int count = hv_iterinit (hv);
		578
		579	if (SvMAGICAL (hv))
357	{	580	{
		581	// need to count by iterating. could improve by dynamically building the vector below
		582	// but I don't care for the speed of this special case.
		583	// note also that we will run into undefined behaviour when the two iterations
		584	// do not result in the same count, something I might care for in some later release.
		585
		586	count = 0;
		587	while (hv_iternext (hv))
		588	++count;
		589
		590	hv_iterinit (hv);
		591	}
		592
		593	if (count)
		594	{
		595	int i, fast = 1;
		596	#if defined(__BORLANDC__) || defined(_MSC_VER)
		597	HE **hes = _alloca (count * sizeof (HE));
		598	#else
358	HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode	599	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
359	int fast = 1;	600	#endif
360		601
361	i = 0;	602	i = 0;
362	while ((he = hv_iternext (hv)))	603	while ((he = hv_iternext (hv)))
363	{	604	{
364	hes [i++] = he;	605	hes [i++] = he;
…		…
386		627
387	FREETMPS;	628	FREETMPS;
388	LEAVE;	629	LEAVE;
389	}	630	}
390		631
391	for (i = 0; i < count; ++i)	632	encode_nl (enc); ++enc->indent;
		633
		634	while (count--)
392	{	635	{
393	encode_indent (enc);	636	encode_indent (enc);
		637	he = hes [count];
394	encode_he (enc, hes [i]);	638	encode_hk (enc, he);
		639	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
395		640
396	if (i < count - 1)	641	if (count)
397	encode_comma (enc);	642	encode_comma (enc);
398	}	643	}
399		644
		645	encode_nl (enc); --enc->indent; encode_indent (enc);
		646	}
		647	}
		648	else
		649	{
		650	if (hv_iterinit (hv) || SvMAGICAL (hv))
		651	if ((he = hv_iternext (hv)))
		652	{
		653	encode_nl (enc); ++enc->indent;
		654
		655	for (;;)
		656	{
400	encode_nl (enc);	657	encode_indent (enc);
		658	encode_hk (enc, he);
		659	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		660
		661	if (!(he = hv_iternext (hv)))
		662	break;
		663
		664	encode_comma (enc);
		665	}
		666
		667	encode_nl (enc); --enc->indent; encode_indent (enc);
		668	}
		669	}
		670
		671	encode_ch (enc, '}');
		672	}
		673
		674	// encode objects, arrays and special \0=false and \1=true values.
		675	static void
		676	encode_rv (enc_t *enc, SV *sv)
		677	{
		678	svtype svt;
		679
		680	SvGETMAGIC (sv);
		681	svt = SvTYPE (sv);
		682
		683	if (expect_false (SvOBJECT (sv)))
		684	{
		685	HV *stash = !JSON_SLOW || json_boolean_stash
		686	? json_boolean_stash
		687	: gv_stashpv ("JSON::XS::Boolean", 1);
		688
		689	if (SvSTASH (sv) == stash)
		690	{
		691	if (SvIV (sv))
		692	encode_str (enc, "true", 4, 0);
		693	else
		694	encode_str (enc, "false", 5, 0);
401	}	695	}
402	else	696	else
403	{	697	{
404	SV *sv;	698	#if 0
405	HE *he = hv_iternext (hv);	699	if (0 && sv_derived_from (rv, "JSON::Literal"))
406
407	for (;;)
408	{
409	encode_indent (enc);
410	encode_he (enc, he);
411
412	if (!(he = hv_iternext (hv)))
413	break;
414
415	encode_comma (enc);
416	}	700	{
		701	// not yet
		702	}
		703	#endif
		704	if (enc->json.flags & F_CONV_BLESSED)
		705	{
		706	// we re-bless the reference to get overload and other niceties right
		707	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
417		708
		709	if (to_json)
		710	{
		711	dSP;
		712
		713	ENTER; SAVETMPS; PUSHMARK (SP);
		714	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		715
		716	// calling with G_SCALAR ensures that we always get a 1 return value
		717	PUTBACK;
		718	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		719	SPAGAIN;
		720
		721	// catch this surprisingly common error
		722	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		723	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		724
		725	sv = POPs;
		726	PUTBACK;
		727
418	encode_nl (enc);	728	encode_sv (enc, sv);
		729
		730	FREETMPS; LEAVE;
		731	}
		732	else if (enc->json.flags & F_ALLOW_BLESSED)
		733	encode_str (enc, "null", 4, 0);
		734	else
		735	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		736	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		737	}
		738	else if (enc->json.flags & F_ALLOW_BLESSED)
		739	encode_str (enc, "null", 4, 0);
		740	else
		741	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		742	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
419	}	743	}
420	}	744	}
		745	else if (svt == SVt_PVHV)
		746	encode_hv (enc, (HV *)sv);
		747	else if (svt == SVt_PVAV)
		748	encode_av (enc, (AV *)sv);
		749	else if (svt < SVt_PVAV)
		750	{
		751	STRLEN len = 0;
		752	char *pv = svt ? SvPV (sv, len) : 0;
421		753
422	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	754	if (len == 1 && *pv == '1')
		755	encode_str (enc, "true", 4, 0);
		756	else if (len == 1 && *pv == '0')
		757	encode_str (enc, "false", 5, 0);
		758	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		759	encode_str (enc, "null", 4, 0);
		760	else
		761	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		762	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		763	}
		764	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		765	encode_str (enc, "null", 4, 0);
		766	else
		767	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		768	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
423	}	769	}
424		770
425	static void	771	static void
426	encode_sv (enc_t *enc, SV *sv)	772	encode_sv (enc_t *enc, SV *sv)
427	{	773	{
…		…
435	encode_str (enc, str, len, SvUTF8 (sv));	781	encode_str (enc, str, len, SvUTF8 (sv));
436	encode_ch (enc, '"');	782	encode_ch (enc, '"');
437	}	783	}
438	else if (SvNOKp (sv))	784	else if (SvNOKp (sv))
439	{	785	{
		786	// trust that perl will do the right thing w.r.t. JSON syntax.
440	need (enc, NV_DIG + 32);	787	need (enc, NV_DIG + 32);
441	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	788	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
442	enc->cur += strlen (enc->cur);	789	enc->cur += strlen (enc->cur);
443	}	790	}
444	else if (SvIOKp (sv))	791	else if (SvIOKp (sv))
445	{	792	{
		793	// we assume we can always read an IV as a UV and vice versa
		794	// we assume two's complement
		795	// we assume no aliasing issues in the union
		796	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		797	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		798	{
		799	// optimise the "small number case"
		800	// code will likely be branchless and use only a single multiplication
		801	// works for numbers up to 59074
		802	I32 i = SvIVX (sv);
		803	U32 u;
		804	char digit, nz = 0;
		805
446	need (enc, 64);	806	need (enc, 6);
		807
		808	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		809	u = i < 0 ? -i : i;
		810
		811	// convert to 4.28 fixed-point representation
		812	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		813
		814	// now output digit by digit, each time masking out the integer part
		815	// and multiplying by 5 while moving the decimal point one to the right,
		816	// resulting in a net multiplication by 10.
		817	// we always write the digit to memory but conditionally increment
		818	// the pointer, to enable the use of conditional move instructions.
		819	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		820	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		821	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		822	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		823	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		824	}
		825	else
		826	{
		827	// large integer, use the (rather slow) snprintf way.
		828	need (enc, IVUV_MAXCHARS);
447	enc->cur +=	829	enc->cur +=
448	SvIsUV(sv)	830	SvIsUV(sv)
449	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	831	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
450	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	832	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		833	}
451	}	834	}
452	else if (SvROK (sv))	835	else if (SvROK (sv))
453	{	836	encode_rv (enc, SvRV (sv));
454	SV *rv = SvRV (sv);	837	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
455
456	if (enc->indent >= enc->maxdepth)
457	croak ("data structure too deep (hit recursion limit)");
458
459	switch (SvTYPE (rv))
460	{
461	case SVt_PVAV: encode_av (enc, (AV *)rv); break;
462	case SVt_PVHV: encode_hv (enc, (HV *)rv); break;
463
464	default:
465	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
466	SvPV_nolen (sv));
467	}
468	}
469	else if (!SvOK (sv))
470	encode_str (enc, "null", 4, 0);	838	encode_str (enc, "null", 4, 0);
471	else	839	else
472	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	840	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
473	SvPV_nolen (sv), SvFLAGS (sv));	841	SvPV_nolen (sv), SvFLAGS (sv));
474	}	842	}
475		843
476	static SV *	844	static SV *
477	encode_json (SV *scalar, U32 flags)	845	encode_json (SV *scalar, JSON *json)
478	{	846	{
		847	enc_t enc;
		848
479	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	849	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
480	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	850	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
481		851
482	enc_t enc;	852	enc.json = *json;
483	enc.flags = flags;
484	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	853	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
485	enc.cur = SvPVX (enc.sv);	854	enc.cur = SvPVX (enc.sv);
486	enc.end = SvEND (enc.sv);	855	enc.end = SvEND (enc.sv);
487	enc.indent = 0;	856	enc.indent = 0;
488	enc.maxdepth = DEC_DEPTH (flags);	857	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		858	: enc.json.flags & F_LATIN1 ? 0x000100UL
		859	: 0x110000UL;
489		860
490	SvPOK_only (enc.sv);	861	SvPOK_only (enc.sv);
491	encode_sv (&enc, scalar);	862	encode_sv (&enc, scalar);
		863	encode_nl (&enc);
492		864
		865	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		866	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		867
493	if (!(flags & (F_ASCII | F_UTF8)))	868	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
494	SvUTF8_on (enc.sv);	869	SvUTF8_on (enc.sv);
495		870
496	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
497
498	if (enc.flags & F_SHRINK)	871	if (enc.json.flags & F_SHRINK)
499	shrink (enc.sv);	872	shrink (enc.sv);
500		873
501	return enc.sv;	874	return enc.sv;
502	}	875	}
503		876
…		…
508	typedef struct	881	typedef struct
509	{	882	{
510	char *cur; // current parser pointer	883	char *cur; // current parser pointer
511	char *end; // end of input string	884	char *end; // end of input string
512	const char *err; // parse error, if != 0	885	const char *err; // parse error, if != 0
513	U32 flags; // F_*	886	JSON json;
514	U32 depth; // recursion depth	887	U32 depth; // recursion depth
515	U32 maxdepth; // recursion depth limit	888	U32 maxdepth; // recursion depth limit
516	} dec_t;	889	} dec_t;
517		890
518	static void	891	INLINE void
		892	decode_comment (dec_t *dec)
		893	{
		894	// only '#'-style comments allowed a.t.m.
		895
		896	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		897	++dec->cur;
		898	}
		899
		900	INLINE void
519	decode_ws (dec_t *dec)	901	decode_ws (dec_t *dec)
520	{	902	{
521	for (;;)	903	for (;;)
522	{	904	{
523	char ch = *dec->cur;	905	char ch = *dec->cur;
524		906
525	if (ch > 0x20	907	if (ch > 0x20)
		908	{
		909	if (expect_false (ch == '#'))
		910	{
		911	if (dec->json.flags & F_RELAXED)
		912	decode_comment (dec);
		913	else
		914	break;
		915	}
		916	else
		917	break;
		918	}
526	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	919	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
527	break;	920	break; // parse error, but let higher level handle it, gives better error messages
528		921
529	++dec->cur;	922	++dec->cur;
530	}	923	}
531	}	924	}
532		925
…		…
536	if (*dec->cur != ch) \	929	if (*dec->cur != ch) \
537	ERR (# ch " expected"); \	930	ERR (# ch " expected"); \
538	++dec->cur; \	931	++dec->cur; \
539	SE	932	SE
540		933
541	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	934	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
542	#define DEC_DEC_DEPTH --dec->depth	935	#define DEC_DEC_DEPTH --dec->depth
543		936
544	static SV *decode_sv (dec_t *dec);	937	static SV *decode_sv (dec_t *dec);
545		938
546	static signed char decode_hexdigit[256];	939	static signed char decode_hexdigit[256];
…		…
549	decode_4hex (dec_t *dec)	942	decode_4hex (dec_t *dec)
550	{	943	{
551	signed char d1, d2, d3, d4;	944	signed char d1, d2, d3, d4;
552	unsigned char *cur = (unsigned char *)dec->cur;	945	unsigned char *cur = (unsigned char *)dec->cur;
553		946
554	d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected");	947	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
555	d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected");	948	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
556	d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected");	949	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
557	d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected");	950	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
558		951
559	dec->cur += 4;	952	dec->cur += 4;
560		953
561	return ((UV)d1) << 12	954	return ((UV)d1) << 12
562	| ((UV)d2) << 8	955	| ((UV)d2) << 8
…		…
570	static SV *	963	static SV *
571	decode_str (dec_t *dec)	964	decode_str (dec_t *dec)
572	{	965	{
573	SV *sv = 0;	966	SV *sv = 0;
574	int utf8 = 0;	967	int utf8 = 0;
		968	char *dec_cur = dec->cur;
575		969
576	do	970	do
577	{	971	{
578	char buf [SHORT_STRING_LEN + UTF8_MAX_LEN];	972	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
579	char *cur = buf;	973	char *cur = buf;
580		974
581	do	975	do
582	{	976	{
583	unsigned char ch = *(unsigned char *)dec->cur++;	977	unsigned char ch = *(unsigned char *)dec_cur++;
584		978
585	if (ch == '"')	979	if (expect_false (ch == '"'))
586	{	980	{
587	--dec->cur;	981	--dec_cur;
588	break;	982	break;
589	}	983	}
590	else if (ch == '\\')	984	else if (expect_false (ch == '\\'))
591	{	985	{
592	switch (*dec->cur)	986	switch (*dec_cur)
593	{	987	{
594	case '\\':	988	case '\\':
595	case '/':	989	case '/':
596	case '"': *cur++ = *dec->cur++; break;	990	case '"': *cur++ = *dec_cur++; break;
597		991
598	case 'b': ++dec->cur; *cur++ = '\010'; break;	992	case 'b': ++dec_cur; *cur++ = '\010'; break;
599	case 't': ++dec->cur; *cur++ = '\011'; break;	993	case 't': ++dec_cur; *cur++ = '\011'; break;
600	case 'n': ++dec->cur; *cur++ = '\012'; break;	994	case 'n': ++dec_cur; *cur++ = '\012'; break;
601	case 'f': ++dec->cur; *cur++ = '\014'; break;	995	case 'f': ++dec_cur; *cur++ = '\014'; break;
602	case 'r': ++dec->cur; *cur++ = '\015'; break;	996	case 'r': ++dec_cur; *cur++ = '\015'; break;
603		997
604	case 'u':	998	case 'u':
605	{	999	{
606	UV lo, hi;	1000	UV lo, hi;
607	++dec->cur;	1001	++dec_cur;
608		1002
		1003	dec->cur = dec_cur;
609	hi = decode_4hex (dec);	1004	hi = decode_4hex (dec);
		1005	dec_cur = dec->cur;
610	if (hi == (UV)-1)	1006	if (hi == (UV)-1)
611	goto fail;	1007	goto fail;
612		1008
613	// possibly a surrogate pair	1009	// possibly a surrogate pair
614	if (hi >= 0xd800)	1010	if (hi >= 0xd800)
615	if (hi < 0xdc00)	1011	if (hi < 0xdc00)
616	{	1012	{
617	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	1013	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
618	ERR ("missing low surrogate character in surrogate pair");	1014	ERR ("missing low surrogate character in surrogate pair");
619		1015
620	dec->cur += 2;	1016	dec_cur += 2;
621		1017
		1018	dec->cur = dec_cur;
622	lo = decode_4hex (dec);	1019	lo = decode_4hex (dec);
		1020	dec_cur = dec->cur;
623	if (lo == (UV)-1)	1021	if (lo == (UV)-1)
624	goto fail;	1022	goto fail;
625		1023
626	if (lo < 0xdc00 || lo >= 0xe000)	1024	if (lo < 0xdc00 || lo >= 0xe000)
627	ERR ("surrogate pair expected");	1025	ERR ("surrogate pair expected");
…		…
633		1031
634	if (hi >= 0x80)	1032	if (hi >= 0x80)
635	{	1033	{
636	utf8 = 1;	1034	utf8 = 1;
637		1035
638	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	1036	cur = encode_utf8 (cur, hi);
639	}	1037	}
640	else	1038	else
641	*cur++ = hi;	1039	*cur++ = hi;
642	}	1040	}
643	break;	1041	break;
644		1042
645	default:	1043	default:
646	--dec->cur;	1044	--dec_cur;
647	ERR ("illegal backslash escape sequence in string");	1045	ERR ("illegal backslash escape sequence in string");
648	}	1046	}
649	}	1047	}
650	else if (ch >= 0x20 && ch <= 0x7f)	1048	else if (expect_true (ch >= 0x20 && ch < 0x80))
651	*cur++ = ch;	1049	*cur++ = ch;
652	else if (ch >= 0x80)	1050	else if (ch >= 0x80)
653	{	1051	{
654	--dec->cur;
655
656	STRLEN clen;	1052	STRLEN clen;
		1053
		1054	--dec_cur;
		1055
657	UV uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	1056	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
658	if (clen == (STRLEN)-1)	1057	if (clen == (STRLEN)-1)
659	ERR ("malformed UTF-8 character in JSON string");	1058	ERR ("malformed UTF-8 character in JSON string");
660		1059
661	do	1060	do
662	{
663	*cur++ = *dec->cur++;	1061	*cur++ = *dec_cur++;
664	}
665	while (--clen);	1062	while (--clen);
666		1063
667	utf8 = 1;	1064	utf8 = 1;
668	}	1065	}
669	else if (!ch)
670	ERR ("unexpected end of string while parsing json string");
671	else	1066	else
		1067	{
		1068	--dec_cur;
		1069
		1070	if (!ch)
		1071	ERR ("unexpected end of string while parsing JSON string");
		1072	else
672	ERR ("invalid character encountered");	1073	ERR ("invalid character encountered while parsing JSON string");
673		1074	}
674	}	1075	}
675	while (cur < buf + SHORT_STRING_LEN);	1076	while (cur < buf + SHORT_STRING_LEN);
676		1077
		1078	{
677	STRLEN len = cur - buf;	1079	STRLEN len = cur - buf;
678		1080
679	if (sv)	1081	if (sv)
680	{	1082	{
681	SvGROW (sv, SvCUR (sv) + len + 1);	1083	STRLEN cur = SvCUR (sv);
		1084
		1085	if (SvLEN (sv) <= cur + len)
		1086	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1087
682	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	1088	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
683	SvCUR_set (sv, SvCUR (sv) + len);	1089	SvCUR_set (sv, SvCUR (sv) + len);
684	}	1090	}
685	else	1091	else
686	sv = newSVpvn (buf, len);	1092	sv = newSVpvn (buf, len);
687	}	1093	}
		1094	}
688	while (*dec->cur != '"');	1095	while (*dec_cur != '"');
689		1096
690	++dec->cur;	1097	++dec_cur;
691		1098
692	if (sv)	1099	if (sv)
693	{	1100	{
694	SvPOK_only (sv);	1101	SvPOK_only (sv);
695	*SvEND (sv) = 0;	1102	*SvEND (sv) = 0;
…		…
698	SvUTF8_on (sv);	1105	SvUTF8_on (sv);
699	}	1106	}
700	else	1107	else
701	sv = newSVpvn ("", 0);	1108	sv = newSVpvn ("", 0);
702		1109
		1110	dec->cur = dec_cur;
703	return sv;	1111	return sv;
704		1112
705	fail:	1113	fail:
		1114	dec->cur = dec_cur;
706	return 0;	1115	return 0;
707	}	1116	}
708		1117
709	static SV *	1118	static SV *
710	decode_num (dec_t *dec)	1119	decode_num (dec_t *dec)
…		…
768	is_nv = 1;	1177	is_nv = 1;
769	}	1178	}
770		1179
771	if (!is_nv)	1180	if (!is_nv)
772	{	1181	{
773	UV uv;	1182	int len = dec->cur - start;
774	int numtype = grok_number (start, dec->cur - start, &uv);	1183
775	if (numtype & IS_NUMBER_IN_UV)	1184	// special case the rather common 1..5-digit-int case
776	if (numtype & IS_NUMBER_NEG)	1185	if (*start == '-')
		1186	switch (len)
777	{	1187	{
778	if (uv < (UV)IV_MIN)	1188	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
779	return newSViv (-(IV)uv);	1189	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
		1190	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1191	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1192	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
780	}	1193	}
		1194	else
		1195	switch (len)
		1196	{
		1197	case 1: return newSViv ( start [0] - '0' * 1);
		1198	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1199	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1200	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1201	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1202	}
		1203
		1204	{
		1205	UV uv;
		1206	int numtype = grok_number (start, len, &uv);
		1207	if (numtype & IS_NUMBER_IN_UV)
		1208	if (numtype & IS_NUMBER_NEG)
		1209	{
		1210	if (uv < (UV)IV_MIN)
		1211	return newSViv (-(IV)uv);
		1212	}
781	else	1213	else
782	return newSVuv (uv);	1214	return newSVuv (uv);
783	}	1215	}
784		1216
		1217	len -= *start == '-' ? 1 : 0;
		1218
		1219	// does not fit into IV or UV, try NV
		1220	if (len <= NV_DIG)
		1221	// fits into NV without loss of precision
		1222	return newSVnv (json_atof (start));
		1223
		1224	// everything else fails, convert it to a string
		1225	return newSVpvn (start, dec->cur - start);
		1226	}
		1227
		1228	// loss of precision here
785	return newSVnv (Atof (start));	1229	return newSVnv (json_atof (start));
786		1230
787	fail:	1231	fail:
788	return 0;	1232	return 0;
789	}	1233	}
790		1234
…		…
819		1263
820	if (*dec->cur != ',')	1264	if (*dec->cur != ',')
821	ERR (", or ] expected while parsing array");	1265	ERR (", or ] expected while parsing array");
822		1266
823	++dec->cur;	1267	++dec->cur;
		1268
		1269	decode_ws (dec);
		1270
		1271	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1272	{
		1273	++dec->cur;
		1274	break;
		1275	}
824	}	1276	}
825		1277
826	DEC_DEC_DEPTH;	1278	DEC_DEC_DEPTH;
827	return newRV_noinc ((SV *)av);	1279	return newRV_noinc ((SV *)av);
828		1280
…		…
833	}	1285	}
834		1286
835	static SV *	1287	static SV *
836	decode_hv (dec_t *dec)	1288	decode_hv (dec_t *dec)
837	{	1289	{
		1290	SV *sv;
838	HV *hv = newHV ();	1291	HV *hv = newHV ();
839		1292
840	DEC_INC_DEPTH;	1293	DEC_INC_DEPTH;
841	decode_ws (dec);	1294	decode_ws (dec);
842		1295
843	if (*dec->cur == '}')	1296	if (*dec->cur == '}')
844	++dec->cur;	1297	++dec->cur;
845	else	1298	else
846	for (;;)	1299	for (;;)
847	{	1300	{
		1301	EXPECT_CH ('"');
		1302
		1303	// heuristic: assume that
		1304	// a) decode_str + hv_store_ent are abysmally slow.
		1305	// b) most hash keys are short, simple ascii text.
		1306	// => try to "fast-match" such strings to avoid
		1307	// the overhead of decode_str + hv_store_ent.
		1308	{
848	SV *key, *value;	1309	SV *value;
		1310	char *p = dec->cur;
		1311	char *e = p + 24; // only try up to 24 bytes
849		1312
850	decode_ws (dec); EXPECT_CH ('"');	1313	for (;;)
851
852	key = decode_str (dec);
853	if (!key)
854	goto fail;
855
856	decode_ws (dec); EXPECT_CH (':');
857
858	value = decode_sv (dec);
859	if (!value)
860	{	1314	{
		1315	// the >= 0x80 is false on most architectures
		1316	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1317	{
		1318	// slow path, back up and use decode_str
		1319	SV *key = decode_str (dec);
		1320	if (!key)
		1321	goto fail;
		1322
		1323	decode_ws (dec); EXPECT_CH (':');
		1324
		1325	decode_ws (dec);
		1326	value = decode_sv (dec);
		1327	if (!value)
		1328	{
		1329	SvREFCNT_dec (key);
		1330	goto fail;
		1331	}
		1332
		1333	hv_store_ent (hv, key, value, 0);
861	SvREFCNT_dec (key);	1334	SvREFCNT_dec (key);
		1335
		1336	break;
		1337	}
		1338	else if (*p == '"')
		1339	{
		1340	// fast path, got a simple key
		1341	char *key = dec->cur;
		1342	int len = p - key;
		1343	dec->cur = p + 1;
		1344
		1345	decode_ws (dec); EXPECT_CH (':');
		1346
		1347	decode_ws (dec);
		1348	value = decode_sv (dec);
		1349	if (!value)
862	goto fail;	1350	goto fail;
		1351
		1352	hv_store (hv, key, len, value, 0);
		1353
		1354	break;
		1355	}
		1356
		1357	++p;
863	}	1358	}
864		1359	}
865	hv_store_ent (hv, key, value, 0);
866	SvREFCNT_dec (key);
867		1360
868	decode_ws (dec);	1361	decode_ws (dec);
869		1362
870	if (*dec->cur == '}')	1363	if (*dec->cur == '}')
871	{	1364	{
…		…
875		1368
876	if (*dec->cur != ',')	1369	if (*dec->cur != ',')
877	ERR (", or } expected while parsing object/hash");	1370	ERR (", or } expected while parsing object/hash");
878		1371
879	++dec->cur;	1372	++dec->cur;
		1373
		1374	decode_ws (dec);
		1375
		1376	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1377	{
		1378	++dec->cur;
		1379	break;
		1380	}
880	}	1381	}
881		1382
882	DEC_DEC_DEPTH;	1383	DEC_DEC_DEPTH;
883	return newRV_noinc ((SV *)hv);	1384	sv = newRV_noinc ((SV *)hv);
		1385
		1386	// check filter callbacks
		1387	if (dec->json.flags & F_HOOK)
		1388	{
		1389	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1390	{
		1391	HE *cb, *he;
		1392
		1393	hv_iterinit (hv);
		1394	he = hv_iternext (hv);
		1395	hv_iterinit (hv);
		1396
		1397	// the next line creates a mortal sv each time its called.
		1398	// might want to optimise this for common cases.
		1399	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1400
		1401	if (cb)
		1402	{
		1403	dSP;
		1404	int count;
		1405
		1406	ENTER; SAVETMPS; PUSHMARK (SP);
		1407	XPUSHs (HeVAL (he));
		1408	sv_2mortal (sv);
		1409
		1410	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1411
		1412	if (count == 1)
		1413	{
		1414	sv = newSVsv (POPs);
		1415	FREETMPS; LEAVE;
		1416	return sv;
		1417	}
		1418
		1419	SvREFCNT_inc (sv);
		1420	FREETMPS; LEAVE;
		1421	}
		1422	}
		1423
		1424	if (dec->json.cb_object)
		1425	{
		1426	dSP;
		1427	int count;
		1428
		1429	ENTER; SAVETMPS; PUSHMARK (SP);
		1430	XPUSHs (sv_2mortal (sv));
		1431
		1432	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1433
		1434	if (count == 1)
		1435	{
		1436	sv = newSVsv (POPs);
		1437	FREETMPS; LEAVE;
		1438	return sv;
		1439	}
		1440
		1441	SvREFCNT_inc (sv);
		1442	FREETMPS; LEAVE;
		1443	}
		1444	}
		1445
		1446	return sv;
884		1447
885	fail:	1448	fail:
886	SvREFCNT_dec (hv);	1449	SvREFCNT_dec (hv);
887	DEC_DEC_DEPTH;	1450	DEC_DEC_DEPTH;
888	return 0;	1451	return 0;
889	}	1452	}
890		1453
891	static SV *	1454	static SV *
892	decode_sv (dec_t *dec)	1455	decode_sv (dec_t *dec)
893	{	1456	{
894	decode_ws (dec);	1457	// the beauty of JSON: you need exactly one character lookahead
		1458	// to parse everything.
895	switch (*dec->cur)	1459	switch (*dec->cur)
896	{	1460	{
897	case '"': ++dec->cur; return decode_str (dec);	1461	case '"': ++dec->cur; return decode_str (dec);
898	case '[': ++dec->cur; return decode_av (dec);	1462	case '[': ++dec->cur; return decode_av (dec);
899	case '{': ++dec->cur; return decode_hv (dec);	1463	case '{': ++dec->cur; return decode_hv (dec);
900		1464
901	case '-':	1465	case '-':
902	case '0': case '1': case '2': case '3': case '4':	1466	case '0': case '1': case '2': case '3': case '4':
903	case '5': case '6': case '7': case '8': case '9':	1467	case '5': case '6': case '7': case '8': case '9':
904	return decode_num (dec);	1468	return decode_num (dec);
905		1469
906	case 't':	1470	case 't':
907	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1471	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
908	{	1472	{
909	dec->cur += 4;	1473	dec->cur += 4;
		1474	#if JSON_SLOW
		1475	json_true = get_bool ("JSON::XS::true");
		1476	#endif
910	return newSViv (1);	1477	return newSVsv (json_true);
911	}	1478	}
912	else	1479	else
913	ERR ("'true' expected");	1480	ERR ("'true' expected");
914		1481
915	break;	1482	break;
916		1483
917	case 'f':	1484	case 'f':
918	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1485	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
919	{	1486	{
920	dec->cur += 5;	1487	dec->cur += 5;
		1488	#if JSON_SLOW
		1489	json_false = get_bool ("JSON::XS::false");
		1490	#endif
921	return newSViv (0);	1491	return newSVsv (json_false);
922	}	1492	}
923	else	1493	else
924	ERR ("'false' expected");	1494	ERR ("'false' expected");
925		1495
926	break;	1496	break;
…		…
935	ERR ("'null' expected");	1505	ERR ("'null' expected");
936		1506
937	break;	1507	break;
938		1508
939	default:	1509	default:
940	ERR ("malformed json string, neither array, object, number, string or atom");	1510	ERR ("malformed JSON string, neither array, object, number, string or atom");
941	break;	1511	break;
942	}	1512	}
943		1513
944	fail:	1514	fail:
945	return 0;	1515	return 0;
946	}	1516	}
947		1517
948	static SV *	1518	static SV *
949	decode_json (SV *string, U32 flags)	1519	decode_json (SV *string, JSON *json, char **offset_return)
950	{	1520	{
		1521	dec_t dec;
951	SV *sv;	1522	SV *sv;
952		1523
953	if (flags & F_UTF8)	1524	/* work around bugs in 5.10 where manipulating magic values
		1525	* will perl ignore the magic in subsequent accesses
		1526	*/
		1527	/*SvGETMAGIC (string);*/
		1528	if (SvMAGICAL (string))
		1529	string = sv_2mortal (newSVsv (string));
		1530
		1531	SvUPGRADE (string, SVt_PV);
		1532
		1533	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1534	* assertion with -DDEBUGGING, although SvCUR is documented to
		1535	* return the xpv_cur field which certainly exists after upgrading.
		1536	* according to nicholas clark, calling SvPOK fixes this.
		1537	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1538	* and hope for the best.
		1539	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1540	* assertion business is seriously broken, try yet another workaround
		1541	* for the broken -DDEBUGGING.
		1542	*/
		1543	{
		1544	#ifdef DEBUGGING
		1545	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1546	#else
		1547	STRLEN offset = SvCUR (string);
		1548	#endif
		1549
		1550	if (offset > json->max_size && json->max_size)
		1551	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1552	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1553	}
		1554
		1555	if (DECODE_WANTS_OCTETS (json))
954	sv_utf8_downgrade (string, 0);	1556	sv_utf8_downgrade (string, 0);
955	else	1557	else
956	sv_utf8_upgrade (string);	1558	sv_utf8_upgrade (string);
957		1559
958	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1560	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
959		1561
960	dec_t dec;	1562	dec.json = *json;
961	dec.flags = flags;
962	dec.cur = SvPVX (string);	1563	dec.cur = SvPVX (string);
963	dec.end = SvEND (string);	1564	dec.end = SvEND (string);
964	dec.err = 0;	1565	dec.err = 0;
965	dec.depth = 0;	1566	dec.depth = 0;
966	dec.maxdepth = DEC_DEPTH (dec.flags);
967		1567
968	*SvEND (sv) = 0; // this shou[ld basically be a nop, too	1568	if (dec.json.cb_object || dec.json.cb_sk_object)
		1569	dec.json.flags |= F_HOOK;
		1570
		1571	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1572
		1573	decode_ws (&dec);
969	sv = decode_sv (&dec);	1574	sv = decode_sv (&dec);
970		1575
		1576	if (offset_return)
		1577	*offset_return = dec.cur;
		1578
		1579	if (!(offset_return || !sv))
		1580	{
		1581	// check for trailing garbage
		1582	decode_ws (&dec);
		1583
		1584	if (*dec.cur)
		1585	{
		1586	dec.err = "garbage after JSON object";
		1587	SvREFCNT_dec (sv);
		1588	sv = 0;
		1589	}
		1590	}
		1591
971	if (!sv)	1592	if (!sv)
972	{	1593	{
973	IV offset = dec.flags & F_UTF8
974	? dec.cur - SvPVX (string)
975	: utf8_distance (dec.cur, SvPVX (string));
976	SV *uni = sv_newmortal ();	1594	SV *uni = sv_newmortal ();
977		1595
978	// horrible hack to silence warning inside pv_uni_display	1596	// horrible hack to silence warning inside pv_uni_display
979	COP cop = *PL_curcop;	1597	COP cop = *PL_curcop;
980	cop.cop_warnings = pWARN_NONE;	1598	cop.cop_warnings = pWARN_NONE;
…		…
982	SAVEVPTR (PL_curcop);	1600	SAVEVPTR (PL_curcop);
983	PL_curcop = &cop;	1601	PL_curcop = &cop;
984	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1602	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
985	LEAVE;	1603	LEAVE;
986		1604
987	croak ("%s, at character offset %d (%s)",	1605	croak ("%s, at character offset %d (before \"%s\")",
988	dec.err,	1606	dec.err,
989	(int)offset,	1607	ptr_to_index (string, dec.cur),
990	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1608	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
991	}	1609	}
992		1610
993	sv = sv_2mortal (sv);	1611	sv = sv_2mortal (sv);
994		1612
995	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1613	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
996	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1614	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
997		1615
998	return sv;	1616	return sv;
		1617	}
		1618
		1619	/////////////////////////////////////////////////////////////////////////////
		1620	// incremental parser
		1621
		1622	static void
		1623	incr_parse (JSON *self)
		1624	{
		1625	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1626
		1627	// the state machine here is a bit convoluted and could be simplified a lot
		1628	// but this would make it slower, so...
		1629
		1630	for (;;)
		1631	{
		1632	//printf ("loop pod %d *p<%c><%s>, mode %d nest %d\n", p - SvPVX (self->incr_text), *p, p, self->incr_mode, self->incr_nest);//D
		1633	switch (self->incr_mode)
		1634	{
		1635	// only used for initial whitespace skipping
		1636	case INCR_M_WS:
		1637	for (;;)
		1638	{
		1639	if (*p > 0x20)
		1640	{
		1641	if (*p == '#')
		1642	{
		1643	self->incr_mode = INCR_M_C0;
		1644	goto incr_m_c;
		1645	}
		1646	else
		1647	{
		1648	self->incr_mode = INCR_M_JSON;
		1649	goto incr_m_json;
		1650	}
		1651	}
		1652	else if (!*p)
		1653	goto interrupt;
		1654
		1655	++p;
		1656	}
		1657
		1658	// skip a single char inside a string (for \\-processing)
		1659	case INCR_M_BS:
		1660	if (!*p)
		1661	goto interrupt;
		1662
		1663	++p;
		1664	self->incr_mode = INCR_M_STR;
		1665	goto incr_m_str;
		1666
		1667	// inside #-style comments
		1668	case INCR_M_C0:
		1669	case INCR_M_C1:
		1670	incr_m_c:
		1671	for (;;)
		1672	{
		1673	if (*p == '\n')
		1674	{
		1675	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1676	break;
		1677	}
		1678	else if (!*p)
		1679	goto interrupt;
		1680
		1681	++p;
		1682	}
		1683
		1684	break;
		1685
		1686	// inside a string
		1687	case INCR_M_STR:
		1688	incr_m_str:
		1689	for (;;)
		1690	{
		1691	if (*p == '"')
		1692	{
		1693	++p;
		1694	self->incr_mode = INCR_M_JSON;
		1695
		1696	if (!self->incr_nest)
		1697	goto interrupt;
		1698
		1699	goto incr_m_json;
		1700	}
		1701	else if (*p == '\\')
		1702	{
		1703	++p; // "virtually" consumes character after \
		1704
		1705	if (!*p) // if at end of string we have to switch modes
		1706	{
		1707	self->incr_mode = INCR_M_BS;
		1708	goto interrupt;
		1709	}
		1710	}
		1711	else if (!*p)
		1712	goto interrupt;
		1713
		1714	++p;
		1715	}
		1716
		1717	// after initial ws, outside string
		1718	case INCR_M_JSON:
		1719	incr_m_json:
		1720	for (;;)
		1721	{
		1722	switch (*p++)
		1723	{
		1724	case 0:
		1725	--p;
		1726	goto interrupt;
		1727
		1728	case 0x09:
		1729	case 0x0a:
		1730	case 0x0d:
		1731	case 0x20:
		1732	if (!self->incr_nest)
		1733	{
		1734	--p; // do not eat the whitespace, let the next round do it
		1735	goto interrupt;
		1736	}
		1737	break;
		1738
		1739	case '"':
		1740	self->incr_mode = INCR_M_STR;
		1741	goto incr_m_str;
		1742
		1743	case '[':
		1744	case '{':
		1745	if (++self->incr_nest > self->max_depth)
		1746	croak (ERR_NESTING_EXCEEDED);
		1747	break;
		1748
		1749	case ']':
		1750	case '}':
		1751	if (--self->incr_nest <= 0)
		1752	goto interrupt;
		1753	break;
		1754
		1755	case '#':
		1756	self->incr_mode = INCR_M_C1;
		1757	goto incr_m_c;
		1758	}
		1759	}
		1760	}
		1761
		1762	modechange:
		1763	;
		1764	}
		1765
		1766	interrupt:
		1767	self->incr_pos = p - SvPVX (self->incr_text);
		1768	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1769	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
999	}	1770	}
1000		1771
1001	/////////////////////////////////////////////////////////////////////////////	1772	/////////////////////////////////////////////////////////////////////////////
1002	// XS interface functions	1773	// XS interface functions
1003		1774
1004	MODULE = JSON::XS PACKAGE = JSON::XS	1775	MODULE = JSON::XS PACKAGE = JSON::XS
1005		1776
1006	BOOT:	1777	BOOT:
1007	{	1778	{
1008	int i;	1779	int i;
1009
1010	memset (decode_hexdigit, 0xff, 256);
1011		1780
1012	for (i = 0; i < 256; ++i)	1781	for (i = 0; i < 256; ++i)
1013	decode_hexdigit [i] =	1782	decode_hexdigit [i] =
1014	i >= '0' && i <= '9' ? i - '0'	1783	i >= '0' && i <= '9' ? i - '0'
1015	: i >= 'a' && i <= 'f' ? i - 'a' + 10	1784	: i >= 'a' && i <= 'f' ? i - 'a' + 10
1016	: i >= 'A' && i <= 'F' ? i - 'A' + 10	1785	: i >= 'A' && i <= 'F' ? i - 'A' + 10
1017	: -1;	1786	: -1;
1018		1787
1019	json_stash = gv_stashpv ("JSON::XS", 1);	1788	json_stash = gv_stashpv ("JSON::XS" , 1);
		1789	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1790
		1791	json_true = get_bool ("JSON::XS::true");
		1792	json_false = get_bool ("JSON::XS::false");
		1793
		1794	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1020	}	1795	}
1021		1796
1022	PROTOTYPES: DISABLE	1797	PROTOTYPES: DISABLE
1023		1798
1024	SV *new (char *dummy)	1799	void CLONE (...)
1025	CODE:	1800	CODE:
1026	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1801	json_stash = 0;
		1802	json_boolean_stash = 0;
		1803
		1804	void new (char *klass)
		1805	PPCODE:
		1806	{
		1807	SV *pv = NEWSV (0, sizeof (JSON));
		1808	SvPOK_only (pv);
		1809	json_init ((JSON *)SvPVX (pv));
		1810	XPUSHs (sv_2mortal (sv_bless (
		1811	newRV_noinc (pv),
		1812	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1813	)));
		1814	}
		1815
		1816	void ascii (JSON *self, int enable = 1)
		1817	ALIAS:
		1818	ascii = F_ASCII
		1819	latin1 = F_LATIN1
		1820	utf8 = F_UTF8
		1821	indent = F_INDENT
		1822	canonical = F_CANONICAL
		1823	space_before = F_SPACE_BEFORE
		1824	space_after = F_SPACE_AFTER
		1825	pretty = F_PRETTY
		1826	allow_nonref = F_ALLOW_NONREF
		1827	shrink = F_SHRINK
		1828	allow_blessed = F_ALLOW_BLESSED
		1829	convert_blessed = F_CONV_BLESSED
		1830	relaxed = F_RELAXED
		1831	allow_unknown = F_ALLOW_UNKNOWN
		1832	PPCODE:
		1833	{
		1834	if (enable)
		1835	self->flags |= ix;
		1836	else
		1837	self->flags &= ~ix;
		1838
		1839	XPUSHs (ST (0));
		1840	}
		1841
		1842	void get_ascii (JSON *self)
		1843	ALIAS:
		1844	get_ascii = F_ASCII
		1845	get_latin1 = F_LATIN1
		1846	get_utf8 = F_UTF8
		1847	get_indent = F_INDENT
		1848	get_canonical = F_CANONICAL
		1849	get_space_before = F_SPACE_BEFORE
		1850	get_space_after = F_SPACE_AFTER
		1851	get_allow_nonref = F_ALLOW_NONREF
		1852	get_shrink = F_SHRINK
		1853	get_allow_blessed = F_ALLOW_BLESSED
		1854	get_convert_blessed = F_CONV_BLESSED
		1855	get_relaxed = F_RELAXED
		1856	get_allow_unknown = F_ALLOW_UNKNOWN
		1857	PPCODE:
		1858	XPUSHs (boolSV (self->flags & ix));
		1859
		1860	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		1861	PPCODE:
		1862	self->max_depth = max_depth;
		1863	XPUSHs (ST (0));
		1864
		1865	U32 get_max_depth (JSON *self)
		1866	CODE:
		1867	RETVAL = self->max_depth;
1027	OUTPUT:	1868	OUTPUT:
1028	RETVAL	1869	RETVAL
1029		1870
1030	SV *ascii (SV *self, int enable = 1)	1871	void max_size (JSON *self, U32 max_size = 0)
1031	ALIAS:	1872	PPCODE:
1032	ascii = F_ASCII	1873	self->max_size = max_size;
1033	utf8 = F_UTF8	1874	XPUSHs (ST (0));
1034	indent = F_INDENT	1875
1035	canonical = F_CANONICAL	1876	int get_max_size (JSON *self)
1036	space_before = F_SPACE_BEFORE
1037	space_after = F_SPACE_AFTER
1038	pretty = F_PRETTY
1039	allow_nonref = F_ALLOW_NONREF
1040	shrink = F_SHRINK
1041	CODE:	1877	CODE:
1042	{	1878	RETVAL = self->max_size;
1043	UV *uv = SvJSON (self);
1044	if (enable)
1045	*uv |= ix;
1046	else
1047	*uv &= ~ix;
1048
1049	RETVAL = newSVsv (self);
1050	}
1051	OUTPUT:	1879	OUTPUT:
1052	RETVAL	1880	RETVAL
1053		1881
1054	SV *max_depth (SV *self, int max_depth = 0x80000000UL)	1882	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
		1883	PPCODE:
		1884	{
		1885	SvREFCNT_dec (self->cb_object);
		1886	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1887
		1888	XPUSHs (ST (0));
		1889	}
		1890
		1891	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
		1892	PPCODE:
		1893	{
		1894	if (!self->cb_sk_object)
		1895	self->cb_sk_object = newHV ();
		1896
		1897	if (SvOK (cb))
		1898	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1899	else
		1900	{
		1901	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1902
		1903	if (!HvKEYS (self->cb_sk_object))
		1904	{
		1905	SvREFCNT_dec (self->cb_sk_object);
		1906	self->cb_sk_object = 0;
		1907	}
		1908	}
		1909
		1910	XPUSHs (ST (0));
		1911	}
		1912
		1913	void encode (JSON *self, SV *scalar)
		1914	PPCODE:
		1915	XPUSHs (encode_json (scalar, self));
		1916
		1917	void decode (JSON *self, SV *jsonstr)
		1918	PPCODE:
		1919	XPUSHs (decode_json (jsonstr, self, 0));
		1920
		1921	void decode_prefix (JSON *self, SV *jsonstr)
		1922	PPCODE:
		1923	{
		1924	char *offset;
		1925	EXTEND (SP, 2);
		1926	PUSHs (decode_json (jsonstr, self, &offset));
		1927	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		1928	}
		1929
		1930	void incr_parse (JSON *self, SV *jsonstr = 0)
		1931	PPCODE:
		1932	{
		1933	if (!self->incr_text)
		1934	self->incr_text = newSVpvn ("", 0);
		1935
		1936	/* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
		1937	if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
		1938	if (DECODE_WANTS_OCTETS (self))
		1939	{
		1940	if (self->incr_pos)
		1941	self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
		1942	(U8 *)SvPVX (self->incr_text) + self->incr_pos);
		1943
		1944	sv_utf8_downgrade (self->incr_text, 0);
		1945	}
		1946	else
		1947	{
		1948	sv_utf8_upgrade (self->incr_text);
		1949
		1950	if (self->incr_pos)
		1951	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1952	- (U8 *)SvPVX (self->incr_text);
		1953	}
		1954
		1955	// append data, if any
		1956	if (jsonstr)
		1957	{
		1958	/* make sure both strings have same encoding */
		1959	if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
		1960	if (SvUTF8 (jsonstr))
		1961	sv_utf8_downgrade (jsonstr, 0);
		1962	else
		1963	sv_utf8_upgrade (jsonstr);
		1964
		1965	/* and then just blindly append */
		1966	{
		1967	STRLEN len;
		1968	const char *str = SvPV (jsonstr, len);
		1969	STRLEN cur = SvCUR (self->incr_text);
		1970
		1971	if (SvLEN (self->incr_text) <= cur + len)
		1972	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1973
		1974	Move (str, SvEND (self->incr_text), len, char);
		1975	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1976	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1977	}
		1978	}
		1979
		1980	if (GIMME_V != G_VOID)
		1981	do
		1982	{
		1983	char *offset;
		1984
		1985	if (!INCR_DONE (self))
		1986	{
		1987	incr_parse (self);
		1988
		1989	if (self->incr_pos > self->max_size && self->max_size)
		1990	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1991	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1992
		1993	if (!INCR_DONE (self))
		1994	break;
		1995	}
		1996
		1997	XPUSHs (decode_json (self->incr_text, self, &offset));
		1998
		1999	self->incr_pos -= offset - SvPVX (self->incr_text);
		2000	self->incr_nest = 0;
		2001	self->incr_mode = 0;
		2002
		2003	sv_chop (self->incr_text, offset);
		2004	}
		2005	while (GIMME_V == G_ARRAY);
		2006	}
		2007
		2008	SV *incr_text (JSON *self)
		2009	ATTRS: lvalue
1055	CODE:	2010	CODE:
1056	{	2011	{
1057	UV *uv = SvJSON (self);	2012	if (self->incr_pos)
1058	UV log2 = 0;	2013	croak ("incr_text can not be called when the incremental parser already started parsing");
1059		2014
1060	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	2015	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
1061
1062	while ((1UL << log2) < max_depth)
1063	++log2;
1064
1065	*uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1066
1067	RETVAL = newSVsv (self);
1068	}	2016	}
1069	OUTPUT:	2017	OUTPUT:
1070	RETVAL	2018	RETVAL
1071		2019
		2020	void incr_skip (JSON *self)
		2021	CODE:
		2022	{
		2023	if (self->incr_pos)
		2024	{
		2025	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2026	self->incr_pos = 0;
		2027	self->incr_nest = 0;
		2028	self->incr_mode = 0;
		2029	}
		2030	}
		2031
		2032	void incr_reset (JSON *self)
		2033	CODE:
		2034	{
		2035	SvREFCNT_dec (self->incr_text);
		2036	self->incr_text = 0;
		2037	self->incr_pos = 0;
		2038	self->incr_nest = 0;
		2039	self->incr_mode = 0;
		2040	}
		2041
		2042	void DESTROY (JSON *self)
		2043	CODE:
		2044	SvREFCNT_dec (self->cb_sk_object);
		2045	SvREFCNT_dec (self->cb_object);
		2046	SvREFCNT_dec (self->incr_text);
		2047
		2048	PROTOTYPES: ENABLE
		2049
1072	void encode (SV *self, SV *scalar)	2050	void encode_json (SV *scalar)
		2051	ALIAS:
		2052	to_json_ = 0
		2053	encode_json = F_UTF8
1073	PPCODE:	2054	PPCODE:
		2055	{
		2056	JSON json;
		2057	json_init (&json);
		2058	json.flags |= ix;
1074	XPUSHs (encode_json (scalar, *SvJSON (self)));	2059	XPUSHs (encode_json (scalar, &json));
		2060	}
1075		2061
1076	void decode (SV *self, SV *jsonstr)	2062	void decode_json (SV *jsonstr)
		2063	ALIAS:
		2064	from_json_ = 0
		2065	decode_json = F_UTF8
1077	PPCODE:	2066	PPCODE:
		2067	{
		2068	JSON json;
		2069	json_init (&json);
		2070	json.flags |= ix;
1078	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	2071	XPUSHs (decode_json (jsonstr, &json, 0));
		2072	}
1079		2073
1080	PROTOTYPES: ENABLE
1081
1082	void to_json (SV *scalar)
1083	ALIAS:
1084	objToJson = 0
1085	PPCODE:
1086	XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8));
1087
1088	void from_json (SV *jsonstr)
1089	ALIAS:
1090	jsonToObj = 0
1091	PPCODE:
1092	XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8));
1093

Diff Legend

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