ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.35 by root, Wed Jun 6 14:52:49 2007 UTC vs.
Revision 1.104 by root, Tue Jan 19 00:31:13 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"	8	#include <stdio.h>
		9	#include <limits.h>
		10	#include <float.h>
9		11
10	#if defined(__BORLANDC__) || defined(_MSC_VER)	12	#if defined(__BORLANDC__) || defined(_MSC_VER)
11	# define snprintf _snprintf // C compilers have this in stdio.h	13	# define snprintf _snprintf // C compilers have this in stdio.h
12	#endif	14	#endif
13		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
14	#define F_ASCII 0x00000001UL	25	#define F_ASCII 0x00000001UL
15	#define F_LATIN1 0x00000002UL	26	#define F_LATIN1 0x00000002UL
16	#define F_UTF8 0x00000004UL	27	#define F_UTF8 0x00000004UL
17	#define F_INDENT 0x00000008UL	28	#define F_INDENT 0x00000008UL
18	#define F_CANONICAL 0x00000010UL	29	#define F_CANONICAL 0x00000010UL
19	#define F_SPACE_BEFORE 0x00000020UL	30	#define F_SPACE_BEFORE 0x00000020UL
20	#define F_SPACE_AFTER 0x00000040UL	31	#define F_SPACE_AFTER 0x00000040UL
21	#define F_ALLOW_NONREF 0x00000100UL	32	#define F_ALLOW_NONREF 0x00000100UL
22	#define F_SHRINK 0x00000200UL	33	#define F_SHRINK 0x00000200UL
		34	#define F_ALLOW_BLESSED 0x00000400UL
		35	#define F_CONV_BLESSED 0x00000800UL
23	#define F_MAXDEPTH 0xf8000000UL	36	#define F_RELAXED 0x00001000UL
24	#define S_MAXDEPTH 27	37	#define F_ALLOW_UNKNOWN 0x00002000UL
25		38	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
26	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
27
28	// F_SELFCONVERT? <=> to_json/toJson
29	// F_BLESSED? <=> { $__class__$ => }
30		39
31	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	40	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
32	#define F_DEFAULT (9UL << S_MAXDEPTH)
33		41
34	#define INIT_SIZE 32 // initial scalar size to be allocated	42	#define INIT_SIZE 32 // initial scalar size to be allocated
35	#define INDENT_STEP 3 // spaces per indentation level	43	#define INDENT_STEP 3 // spaces per indentation level
36		44
37	#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
38		46
39	#define SB do {	47	#define SB do {
40	#define SE } while (0)	48	#define SE } while (0)
41		49
42	#if __GNUC__ >= 3	50	#if __GNUC__ >= 3
43	# define expect(expr,value) __builtin_expect ((expr),(value))	51	# define expect(expr,value) __builtin_expect ((expr), (value))
44	# define inline inline	52	# define INLINE static inline
45	#else	53	#else
46	# define expect(expr,value) (expr)	54	# define expect(expr,value) (expr)
47	# define inline static	55	# define INLINE static
48	#endif	56	#endif
49		57
50	#define expect_false(expr) expect ((expr) != 0, 0)	58	#define expect_false(expr) expect ((expr) != 0, 0)
51	#define expect_true(expr) expect ((expr) != 0, 1)	59	#define expect_true(expr) expect ((expr) != 0, 1)
52		60
		61	#define IN_RANGE_INC(type,val,beg,end) \
		62	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		63	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		64
		65	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
		66
		67	#ifdef USE_ITHREADS
		68	# define JSON_SLOW 1
		69	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		70	#else
		71	# define JSON_SLOW 0
		72	# define JSON_STASH json_stash
		73	#endif
		74
53	static HV *json_stash; // JSON::XS::	75	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		76	static SV *json_true, *json_false;
		77
		78	enum {
		79	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		80	INCR_M_STR, // inside string
		81	INCR_M_BS, // inside backslash
		82	INCR_M_C0, // inside comment in initial whitespace sequence
		83	INCR_M_C1, // inside comment in other places
		84	INCR_M_JSON // outside anything, count nesting
		85	};
		86
		87	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		88
		89	typedef struct {
		90	U32 flags;
		91	U32 max_depth;
		92	STRLEN max_size;
		93
		94	SV *cb_object;
		95	HV *cb_sk_object;
		96
		97	// for the incremental parser
		98	SV *incr_text; // the source text so far
		99	STRLEN incr_pos; // the current offset into the text
		100	int incr_nest; // {[]}-nesting level
		101	unsigned char incr_mode;
		102	} JSON;
		103
		104	INLINE void
		105	json_init (JSON *json)
		106	{
		107	Zero (json, 1, JSON);
		108	json->max_depth = 512;
		109	}
54		110
55	/////////////////////////////////////////////////////////////////////////////	111	/////////////////////////////////////////////////////////////////////////////
56	// utility functions	112	// utility functions
57		113
58	static UV *	114	INLINE SV *
59	SvJSON (SV *sv)	115	get_bool (const char *name)
60	{	116	{
61	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))	117	SV *sv = get_sv (name, 1);
62	croak ("object is not of type JSON::XS");
63		118
64	return &SvUVX (SvRV (sv));	119	SvREADONLY_on (sv);
65	}	120	SvREADONLY_on (SvRV (sv));
66		121
67	static void	122	return sv;
		123	}
		124
		125	INLINE void
68	shrink (SV *sv)	126	shrink (SV *sv)
69	{	127	{
70	sv_utf8_downgrade (sv, 1);	128	sv_utf8_downgrade (sv, 1);
		129
71	if (SvLEN (sv) > SvCUR (sv) + 1)	130	if (SvLEN (sv) > SvCUR (sv) + 1)
72	{	131	{
73	#ifdef SvPV_shrink_to_cur	132	#ifdef SvPV_shrink_to_cur
74	SvPV_shrink_to_cur (sv);	133	SvPV_shrink_to_cur (sv);
75	#elif defined (SvPV_renew)	134	#elif defined (SvPV_renew)
…		…
81	// decode an utf-8 character and return it, or (UV)-1 in	140	// decode an utf-8 character and return it, or (UV)-1 in
82	// case of an error.	141	// case of an error.
83	// we special-case "safe" characters from U+80 .. U+7FF,	142	// we special-case "safe" characters from U+80 .. U+7FF,
84	// but use the very good perl function to parse anything else.	143	// but use the very good perl function to parse anything else.
85	// note that we never call this function for a ascii codepoints	144	// note that we never call this function for a ascii codepoints
86	inline UV	145	INLINE UV
87	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)	146	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
88	{	147	{
89	if (expect_false (s[0] > 0xdf || s[0] < 0xc2))	148	if (expect_true (len >= 2
90	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	149	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
91	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	150	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
92	{	151	{
93	*clen = 2;	152	*clen = 2;
94	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);	153	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
95	}	154	}
96	else	155	else
97	{	156	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
98	*clen = (STRLEN)-1;	157	}
99	return (UV)-1;	158
		159	// likewise for encoding, also never called for ascii codepoints
		160	// this function takes advantage of this fact, although current gccs
		161	// seem to optimise the check for >= 0x80 away anyways
		162	INLINE unsigned char *
		163	encode_utf8 (unsigned char *s, UV ch)
		164	{
		165	if (expect_false (ch < 0x000080))
		166	*s++ = ch;
		167	else if (expect_true (ch < 0x000800))
		168	*s++ = 0xc0 | ( ch >> 6),
		169	*s++ = 0x80 | ( ch & 0x3f);
		170	else if ( ch < 0x010000)
		171	*s++ = 0xe0 | ( ch >> 12),
		172	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		173	*s++ = 0x80 | ( ch & 0x3f);
		174	else if ( ch < 0x110000)
		175	*s++ = 0xf0 | ( ch >> 18),
		176	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		177	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		178	*s++ = 0x80 | ( ch & 0x3f);
		179
		180	return s;
		181	}
		182
		183	// convert offset pointer to character index, sv must be string
		184	static STRLEN
		185	ptr_to_index (SV *sv, char *offset)
		186	{
		187	return SvUTF8 (sv)
		188	? utf8_distance (offset, SvPVX (sv))
		189	: offset - SvPVX (sv);
		190	}
		191
		192	/////////////////////////////////////////////////////////////////////////////
		193	// fp hell
		194
		195	// scan a group of digits, and a trailing exponent
		196	static void
		197	json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp)
		198	{
		199	UV uaccum = 0;
		200	int eaccum = 0;
		201
		202	for (;;)
100	}	203	{
101	}	204	U8 dig = (U8)*s - '0';
102		205
		206	if (expect_false (dig >= 10))
		207	{
		208	if (dig == (U8)((U8)'.' - (U8)'0'))
		209	{
		210	++s;
		211	json_atof_scan1 (s, accum, expo, 1);
		212	}
		213	else if ((dig | ' ') == 'e' - '0')
		214	{
		215	int exp2 = 0;
		216	int neg = 0;
		217
		218	++s;
		219
		220	if (*s == '-')
		221	{
		222	++s;
		223	neg = 1;
		224	}
		225	else if (*s == '+')
		226	++s;
		227
		228	while ((dig = (U8)*s - '0') < 10)
		229	exp2 = exp2 * 10 + *s++ - '0';
		230
		231	*expo += neg ? -exp2 : exp2;
		232	}
		233
		234	break;
		235	}
		236
		237	++s;
		238
		239	uaccum = uaccum * 10 + dig;
		240	++eaccum;
		241
		242	// if we have too many digits, then recurse for more
		243	// we actually do this for rather few digits
		244	if (uaccum >= (UV_MAX - 9) / 10)
		245	{
		246	if (postdp) *expo -= eaccum;
		247	json_atof_scan1 (s, accum, expo, postdp);
		248	if (postdp) *expo += eaccum;
		249
		250	break;
		251	}
		252	}
		253
		254	if (postdp) *expo -= eaccum;
		255	*accum += uaccum * pow (10., *expo);
		256	*expo += eaccum;
		257	}
		258
		259	static NV
		260	json_atof (const char *s)
		261	{
		262	NV accum = 0.;
		263	int expo = 0;
		264	int neg = 0;
		265
		266	if (*s == '-')
		267	{
		268	++s;
		269	neg = 1;
		270	}
		271
		272	json_atof_scan1 (s, &accum, &expo, 0);
		273
		274	return neg ? -accum : accum;
		275	}
103	/////////////////////////////////////////////////////////////////////////////	276	/////////////////////////////////////////////////////////////////////////////
104	// encoder	277	// encoder
105		278
106	// structure used for encoding JSON	279	// structure used for encoding JSON
107	typedef struct	280	typedef struct
108	{	281	{
109	char *cur; // SvPVX (sv) + current output position	282	char *cur; // SvPVX (sv) + current output position
110	char *end; // SvEND (sv)	283	char *end; // SvEND (sv)
111	SV *sv; // result scalar	284	SV *sv; // result scalar
112	U32 flags; // F_*	285	JSON json;
113	U32 indent; // indentation level	286	U32 indent; // indentation level
114	U32 maxdepth; // max. indentation/recursion level	287	UV limit; // escape character values >= this value when encoding
115	} enc_t;	288	} enc_t;
116		289
117	inline void	290	INLINE void
118	need (enc_t *enc, STRLEN len)	291	need (enc_t *enc, STRLEN len)
119	{	292	{
120	if (expect_false (enc->cur + len >= enc->end))	293	if (expect_false (enc->cur + len >= enc->end))
121	{	294	{
122	STRLEN cur = enc->cur - SvPVX (enc->sv);	295	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
123	SvGROW (enc->sv, cur + len + 1);	296	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
124	enc->cur = SvPVX (enc->sv) + cur;	297	enc->cur = SvPVX (enc->sv) + cur;
125	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	298	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
126	}	299	}
127	}	300	}
128		301
129	inline void	302	INLINE void
130	encode_ch (enc_t *enc, char ch)	303	encode_ch (enc_t *enc, char ch)
131	{	304	{
132	need (enc, 1);	305	need (enc, 1);
133	*enc->cur++ = ch;	306	*enc->cur++ = ch;
134	}	307	}
…		…
178	STRLEN clen;	351	STRLEN clen;
179	UV uch;	352	UV uch;
180		353
181	if (is_utf8)	354	if (is_utf8)
182	{	355	{
183	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
184	uch = decode_utf8 (str, end - str, &clen);	356	uch = decode_utf8 (str, end - str, &clen);
185	if (clen == (STRLEN)-1)	357	if (clen == (STRLEN)-1)
186	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	358	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
187	}	359	}
188	else	360	else
189	{	361	{
190	uch = ch;	362	uch = ch;
191	clen = 1;	363	clen = 1;
192	}	364	}
193		365
194	if (uch > 0x10FFFFUL)	366	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
195	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
196
197	if (uch < 0x80 || enc->flags & F_ASCII || (enc->flags & F_LATIN1 && uch > 0xFF))
198	{	367	{
199	if (uch > 0xFFFFUL)	368	if (uch >= 0x10000UL)
200	{	369	{
		370	if (uch >= 0x110000UL)
		371	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		372
201	need (enc, len += 11);	373	need (enc, len += 11);
202	sprintf (enc->cur, "\\u%04x\\u%04x",	374	sprintf (enc->cur, "\\u%04x\\u%04x",
203	(int)((uch - 0x10000) / 0x400 + 0xD800),	375	(int)((uch - 0x10000) / 0x400 + 0xD800),
204	(int)((uch - 0x10000) % 0x400 + 0xDC00));	376	(int)((uch - 0x10000) % 0x400 + 0xDC00));
205	enc->cur += 12;	377	enc->cur += 12;
206	}	378	}
207	else	379	else
208	{	380	{
209	static char hexdigit [16] = "0123456789abcdef";
210	need (enc, len += 5);	381	need (enc, len += 5);
211	*enc->cur++ = '\\';	382	*enc->cur++ = '\\';
212	*enc->cur++ = 'u';	383	*enc->cur++ = 'u';
213	*enc->cur++ = hexdigit [ uch >> 12 ];	384	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
214	*enc->cur++ = hexdigit [(uch >> 8) & 15];	385	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
215	*enc->cur++ = hexdigit [(uch >> 4) & 15];	386	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
216	*enc->cur++ = hexdigit [(uch >> 0) & 15];	387	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
217	}	388	}
218		389
219	str += clen;	390	str += clen;
220	}	391	}
221	else if (enc->flags & F_LATIN1)	392	else if (enc->json.flags & F_LATIN1)
222	{	393	{
223	*enc->cur++ = uch;	394	*enc->cur++ = uch;
224	str += clen;	395	str += clen;
225	}	396	}
226	else if (is_utf8)	397	else if (is_utf8)
…		…
233	while (--clen);	404	while (--clen);
234	}	405	}
235	else	406	else
236	{	407	{
237	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	408	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
238	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	409	enc->cur = encode_utf8 (enc->cur, uch);
239	++str;	410	++str;
240	}	411	}
241	}	412	}
242	}	413	}
243	}	414	}
244		415
245	--len;	416	--len;
246	}	417	}
247	}	418	}
248		419
249	inline void	420	INLINE void
250	encode_indent (enc_t *enc)	421	encode_indent (enc_t *enc)
251	{	422	{
252	if (enc->flags & F_INDENT)	423	if (enc->json.flags & F_INDENT)
253	{	424	{
254	int spaces = enc->indent * INDENT_STEP;	425	int spaces = enc->indent * INDENT_STEP;
255		426
256	need (enc, spaces);	427	need (enc, spaces);
257	memset (enc->cur, ' ', spaces);	428	memset (enc->cur, ' ', spaces);
258	enc->cur += spaces;	429	enc->cur += spaces;
259	}	430	}
260	}	431	}
261		432
262	inline void	433	INLINE void
263	encode_space (enc_t *enc)	434	encode_space (enc_t *enc)
264	{	435	{
265	need (enc, 1);	436	need (enc, 1);
266	encode_ch (enc, ' ');	437	encode_ch (enc, ' ');
267	}	438	}
268		439
269	inline void	440	INLINE void
270	encode_nl (enc_t *enc)	441	encode_nl (enc_t *enc)
271	{	442	{
272	if (enc->flags & F_INDENT)	443	if (enc->json.flags & F_INDENT)
273	{	444	{
274	need (enc, 1);	445	need (enc, 1);
275	encode_ch (enc, '\n');	446	encode_ch (enc, '\n');
276	}	447	}
277	}	448	}
278		449
279	inline void	450	INLINE void
280	encode_comma (enc_t *enc)	451	encode_comma (enc_t *enc)
281	{	452	{
282	encode_ch (enc, ',');	453	encode_ch (enc, ',');
283		454
284	if (enc->flags & F_INDENT)	455	if (enc->json.flags & F_INDENT)
285	encode_nl (enc);	456	encode_nl (enc);
286	else if (enc->flags & F_SPACE_AFTER)	457	else if (enc->json.flags & F_SPACE_AFTER)
287	encode_space (enc);	458	encode_space (enc);
288	}	459	}
289		460
290	static void encode_sv (enc_t *enc, SV *sv);	461	static void encode_sv (enc_t *enc, SV *sv);
291		462
292	static void	463	static void
293	encode_av (enc_t *enc, AV *av)	464	encode_av (enc_t *enc, AV *av)
294	{	465	{
295	int i, len = av_len (av);	466	int i, len = av_len (av);
296		467
297	if (enc->indent >= enc->maxdepth)	468	if (enc->indent >= enc->json.max_depth)
298	croak ("data structure too deep (hit recursion limit)");	469	croak (ERR_NESTING_EXCEEDED);
299		470
300	encode_ch (enc, '['); encode_nl (enc);	471	encode_ch (enc, '[');
301	++enc->indent;	472
		473	if (len >= 0)
		474	{
		475	encode_nl (enc); ++enc->indent;
302		476
303	for (i = 0; i <= len; ++i)	477	for (i = 0; i <= len; ++i)
304	{	478	{
		479	SV **svp = av_fetch (av, i, 0);
		480
305	encode_indent (enc);	481	encode_indent (enc);
306	encode_sv (enc, *av_fetch (av, i, 0));
307		482
		483	if (svp)
		484	encode_sv (enc, *svp);
		485	else
		486	encode_str (enc, "null", 4, 0);
		487
308	if (i < len)	488	if (i < len)
309	encode_comma (enc);	489	encode_comma (enc);
310	}	490	}
311		491
		492	encode_nl (enc); --enc->indent; encode_indent (enc);
		493	}
		494
312	encode_nl (enc);	495	encode_ch (enc, ']');
313
314	--enc->indent;
315	encode_indent (enc); encode_ch (enc, ']');
316	}	496	}
317		497
318	static void	498	static void
319	encode_he (enc_t *enc, HE *he)	499	encode_hk (enc_t *enc, HE *he)
320	{	500	{
321	encode_ch (enc, '"');	501	encode_ch (enc, '"');
322		502
323	if (HeKLEN (he) == HEf_SVKEY)	503	if (HeKLEN (he) == HEf_SVKEY)
324	{	504	{
…		…
334	else	514	else
335	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	515	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
336		516
337	encode_ch (enc, '"');	517	encode_ch (enc, '"');
338		518
339	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	519	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
340	encode_ch (enc, ':');	520	encode_ch (enc, ':');
341	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	521	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
342	encode_sv (enc, HeVAL (he));
343	}	522	}
344		523
345	// compare hash entries, used when all keys are bytestrings	524	// compare hash entries, used when all keys are bytestrings
346	static int	525	static int
347	he_cmp_fast (const void *a_, const void *b_)	526	he_cmp_fast (const void *a_, const void *b_)
…		…
352	HE *b = *(HE **)b_;	531	HE *b = *(HE **)b_;
353		532
354	STRLEN la = HeKLEN (a);	533	STRLEN la = HeKLEN (a);
355	STRLEN lb = HeKLEN (b);	534	STRLEN lb = HeKLEN (b);
356		535
357	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	536	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
358	cmp = la - lb;	537	cmp = lb - la;
359		538
360	return cmp;	539	return cmp;
361	}	540	}
362		541
363	// compare hash entries, used when some keys are sv's or utf-x	542	// compare hash entries, used when some keys are sv's or utf-x
364	static int	543	static int
365	he_cmp_slow (const void *a, const void *b)	544	he_cmp_slow (const void *a, const void *b)
366	{	545	{
367	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	546	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
368	}	547	}
369		548
370	static void	549	static void
371	encode_hv (enc_t *enc, HV *hv)	550	encode_hv (enc_t *enc, HV *hv)
372	{	551	{
373	int count, i;	552	HE *he;
374		553
375	if (enc->indent >= enc->maxdepth)	554	if (enc->indent >= enc->json.max_depth)
376	croak ("data structure too deep (hit recursion limit)");	555	croak (ERR_NESTING_EXCEEDED);
377		556
378	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	557	encode_ch (enc, '{');
379		558
380	if ((count = hv_iterinit (hv)))
381	{
382	// for canonical output we have to sort by keys first	559	// for canonical output we have to sort by keys first
383	// actually, this is mostly due to the stupid so-called	560	// actually, this is mostly due to the stupid so-called
384	// security workaround added somewhere in 5.8.x.	561	// security workaround added somewhere in 5.8.x
385	// that randomises hash orderings	562	// that randomises hash orderings
386	if (enc->flags & F_CANONICAL)	563	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
		564	{
		565	int count = hv_iterinit (hv);
		566
		567	if (SvMAGICAL (hv))
387	{	568	{
		569	// need to count by iterating. could improve by dynamically building the vector below
		570	// but I don't care for the speed of this special case.
		571	// note also that we will run into undefined behaviour when the two iterations
		572	// do not result in the same count, something I might care for in some later release.
		573
		574	count = 0;
		575	while (hv_iternext (hv))
		576	++count;
		577
		578	hv_iterinit (hv);
		579	}
		580
		581	if (count)
		582	{
388	int fast = 1;	583	int i, fast = 1;
389	HE *he;
390	#if defined(__BORLANDC__) || defined(_MSC_VER)	584	#if defined(__BORLANDC__) || defined(_MSC_VER)
391	HE **hes = _alloca (count * sizeof (HE));	585	HE **hes = _alloca (count * sizeof (HE));
392	#else	586	#else
393	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode	587	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
394	#endif	588	#endif
…		…
421		615
422	FREETMPS;	616	FREETMPS;
423	LEAVE;	617	LEAVE;
424	}	618	}
425		619
426	for (i = 0; i < count; ++i)	620	encode_nl (enc); ++enc->indent;
		621
		622	while (count--)
427	{	623	{
428	encode_indent (enc);	624	encode_indent (enc);
		625	he = hes [count];
429	encode_he (enc, hes [i]);	626	encode_hk (enc, he);
		627	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
430		628
431	if (i < count - 1)	629	if (count)
432	encode_comma (enc);	630	encode_comma (enc);
433	}	631	}
434		632
		633	encode_nl (enc); --enc->indent; encode_indent (enc);
		634	}
		635	}
		636	else
		637	{
		638	if (hv_iterinit (hv) || SvMAGICAL (hv))
		639	if ((he = hv_iternext (hv)))
		640	{
		641	encode_nl (enc); ++enc->indent;
		642
		643	for (;;)
		644	{
435	encode_nl (enc);	645	encode_indent (enc);
		646	encode_hk (enc, he);
		647	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		648
		649	if (!(he = hv_iternext (hv)))
		650	break;
		651
		652	encode_comma (enc);
		653	}
		654
		655	encode_nl (enc); --enc->indent; encode_indent (enc);
		656	}
		657	}
		658
		659	encode_ch (enc, '}');
		660	}
		661
		662	// encode objects, arrays and special \0=false and \1=true values.
		663	static void
		664	encode_rv (enc_t *enc, SV *sv)
		665	{
		666	svtype svt;
		667
		668	SvGETMAGIC (sv);
		669	svt = SvTYPE (sv);
		670
		671	if (expect_false (SvOBJECT (sv)))
		672	{
		673	HV *stash = !JSON_SLOW || json_boolean_stash
		674	? json_boolean_stash
		675	: gv_stashpv ("JSON::XS::Boolean", 1);
		676
		677	if (SvSTASH (sv) == stash)
		678	{
		679	if (SvIV (sv))
		680	encode_str (enc, "true", 4, 0);
		681	else
		682	encode_str (enc, "false", 5, 0);
436	}	683	}
437	else	684	else
438	{	685	{
439	HE *he = hv_iternext (hv);	686	#if 0
440		687	if (0 && sv_derived_from (rv, "JSON::Literal"))
441	for (;;)
442	{	688	{
443	encode_indent (enc);	689	// not yet
444	encode_he (enc, he);
445
446	if (!(he = hv_iternext (hv)))
447	break;
448
449	encode_comma (enc);
450	}	690	}
		691	#endif
		692	if (enc->json.flags & F_CONV_BLESSED)
		693	{
		694	// we re-bless the reference to get overload and other niceties right
		695	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
451		696
		697	if (to_json)
		698	{
		699	dSP;
		700
		701	ENTER; SAVETMPS; PUSHMARK (SP);
		702	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		703
		704	// calling with G_SCALAR ensures that we always get a 1 return value
		705	PUTBACK;
		706	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		707	SPAGAIN;
		708
		709	// catch this surprisingly common error
		710	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		711	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		712
		713	sv = POPs;
		714	PUTBACK;
		715
452	encode_nl (enc);	716	encode_sv (enc, sv);
		717
		718	FREETMPS; LEAVE;
		719	}
		720	else if (enc->json.flags & F_ALLOW_BLESSED)
		721	encode_str (enc, "null", 4, 0);
		722	else
		723	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		724	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		725	}
		726	else if (enc->json.flags & F_ALLOW_BLESSED)
		727	encode_str (enc, "null", 4, 0);
		728	else
		729	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		730	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
453	}	731	}
454	}	732	}
455
456	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
457	}
458
459	// encode objects, arrays and special \0=false and \1=true values.
460	static void
461	encode_rv (enc_t *enc, SV *sv)
462	{
463	svtype svt;
464
465	SvGETMAGIC (sv);
466	svt = SvTYPE (sv);
467
468	if (svt == SVt_PVHV)	733	else if (svt == SVt_PVHV)
469	encode_hv (enc, (HV *)sv);	734	encode_hv (enc, (HV *)sv);
470	else if (svt == SVt_PVAV)	735	else if (svt == SVt_PVAV)
471	encode_av (enc, (AV *)sv);	736	encode_av (enc, (AV *)sv);
472	else if (svt < SVt_PVAV)	737	else if (svt < SVt_PVAV)
473	{	738	{
474	if (SvNIOK (sv) && SvIV (sv) == 0)	739	STRLEN len = 0;
		740	char *pv = svt ? SvPV (sv, len) : 0;
		741
		742	if (len == 1 && *pv == '1')
		743	encode_str (enc, "true", 4, 0);
		744	else if (len == 1 && *pv == '0')
475	encode_str (enc, "false", 5, 0);	745	encode_str (enc, "false", 5, 0);
476	else if (SvNIOK (sv) && SvIV (sv) == 1)	746	else if (enc->json.flags & F_ALLOW_UNKNOWN)
477	encode_str (enc, "true", 4, 0);	747	encode_str (enc, "null", 4, 0);
478	else	748	else
479	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	749	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
480	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	750	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
481	}	751	}
		752	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		753	encode_str (enc, "null", 4, 0);
482	else	754	else
483	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	755	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
484	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	756	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
485	}	757	}
486		758
…		…
497	encode_str (enc, str, len, SvUTF8 (sv));	769	encode_str (enc, str, len, SvUTF8 (sv));
498	encode_ch (enc, '"');	770	encode_ch (enc, '"');
499	}	771	}
500	else if (SvNOKp (sv))	772	else if (SvNOKp (sv))
501	{	773	{
		774	// trust that perl will do the right thing w.r.t. JSON syntax.
502	need (enc, NV_DIG + 32);	775	need (enc, NV_DIG + 32);
503	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	776	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
504	enc->cur += strlen (enc->cur);	777	enc->cur += strlen (enc->cur);
505	}	778	}
506	else if (SvIOKp (sv))	779	else if (SvIOKp (sv))
507	{	780	{
508	// we assume we can always read an IV as a UV	781	// we assume we can always read an IV as a UV and vice versa
509	if (SvUV (sv) & ~(UV)0x7fff)	782	// we assume two's complement
		783	// we assume no aliasing issues in the union
		784	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		785	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
510	{	786	{
		787	// optimise the "small number case"
		788	// code will likely be branchless and use only a single multiplication
		789	// works for numbers up to 59074
		790	I32 i = SvIVX (sv);
		791	U32 u;
		792	char digit, nz = 0;
		793
511	need (enc, 32);	794	need (enc, 6);
512	enc->cur +=	795
513	SvIsUV(sv)	796	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
514	? snprintf (enc->cur, 32, "%"UVuf, (UV)SvUVX (sv))	797	u = i < 0 ? -i : i;
515	: snprintf (enc->cur, 32, "%"IVdf, (IV)SvIVX (sv));	798
		799	// convert to 4.28 fixed-point representation
		800	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		801
		802	// now output digit by digit, each time masking out the integer part
		803	// and multiplying by 5 while moving the decimal point one to the right,
		804	// resulting in a net multiplication by 10.
		805	// we always write the digit to memory but conditionally increment
		806	// the pointer, to enable the use of conditional move instructions.
		807	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		808	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		809	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		810	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		811	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
516	}	812	}
517	else	813	else
518	{	814	{
519	// optimise the "small number case"	815	// large integer, use the (rather slow) snprintf way.
520	// code will likely be branchless and use only a single multiplication
521	I32 i = SvIV (sv);
522	U32 u;
523
524	need (enc, 6);	816	need (enc, IVUV_MAXCHARS);
525		817	enc->cur +=
526	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;	818	SvIsUV(sv)
527	u = i < 0 ? -i : i;	819	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
528		820	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
529	// convert to 4.28 fixed-point representation
530	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
531
532	char digit, nz = 0;
533
534	digit = u >> 28; *enc->cur = digit + '0'; nz |= digit; enc->cur += nz ? 1 : 0; u = (u & 0xfffffff) * 5;
535	digit = u >> 27; *enc->cur = digit + '0'; nz |= digit; enc->cur += nz ? 1 : 0; u = (u & 0x7ffffff) * 5;
536	digit = u >> 26; *enc->cur = digit + '0'; nz |= digit; enc->cur += nz ? 1 : 0; u = (u & 0x3ffffff) * 5;
537	digit = u >> 25; *enc->cur = digit + '0'; nz |= digit; enc->cur += nz ? 1 : 0; u = (u & 0x1ffffff) * 5;
538	digit = u >> 24; *enc->cur = digit + '0'; nz |= digit; enc->cur += 1;
539	}	821	}
540	}	822	}
541	else if (SvROK (sv))	823	else if (SvROK (sv))
542	encode_rv (enc, SvRV (sv));	824	encode_rv (enc, SvRV (sv));
543	else if (!SvOK (sv))	825	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
544	encode_str (enc, "null", 4, 0);	826	encode_str (enc, "null", 4, 0);
545	else	827	else
546	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	828	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
547	SvPV_nolen (sv), SvFLAGS (sv));	829	SvPV_nolen (sv), SvFLAGS (sv));
548	}	830	}
549		831
550	static SV *	832	static SV *
551	encode_json (SV *scalar, U32 flags)	833	encode_json (SV *scalar, JSON *json)
552	{	834	{
553	enc_t enc;	835	enc_t enc;
554		836
555	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	837	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
556	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	838	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
557		839
558	enc.flags = flags;	840	enc.json = *json;
559	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	841	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
560	enc.cur = SvPVX (enc.sv);	842	enc.cur = SvPVX (enc.sv);
561	enc.end = SvEND (enc.sv);	843	enc.end = SvEND (enc.sv);
562	enc.indent = 0;	844	enc.indent = 0;
563	enc.maxdepth = DEC_DEPTH (flags);	845	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		846	: enc.json.flags & F_LATIN1 ? 0x000100UL
		847	: 0x110000UL;
564		848
565	SvPOK_only (enc.sv);	849	SvPOK_only (enc.sv);
566	encode_sv (&enc, scalar);	850	encode_sv (&enc, scalar);
		851	encode_nl (&enc);
567		852
568	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	853	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
569	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	854	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
570		855
571	if (!(flags & (F_ASCII | F_LATIN1 | F_UTF8)))	856	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
572	SvUTF8_on (enc.sv);	857	SvUTF8_on (enc.sv);
573		858
574	if (enc.flags & F_SHRINK)	859	if (enc.json.flags & F_SHRINK)
575	shrink (enc.sv);	860	shrink (enc.sv);
576		861
577	return enc.sv;	862	return enc.sv;
578	}	863	}
579		864
…		…
584	typedef struct	869	typedef struct
585	{	870	{
586	char *cur; // current parser pointer	871	char *cur; // current parser pointer
587	char *end; // end of input string	872	char *end; // end of input string
588	const char *err; // parse error, if != 0	873	const char *err; // parse error, if != 0
589	U32 flags; // F_*	874	JSON json;
590	U32 depth; // recursion depth	875	U32 depth; // recursion depth
591	U32 maxdepth; // recursion depth limit	876	U32 maxdepth; // recursion depth limit
592	} dec_t;	877	} dec_t;
593		878
594	inline void	879	INLINE void
		880	decode_comment (dec_t *dec)
		881	{
		882	// only '#'-style comments allowed a.t.m.
		883
		884	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		885	++dec->cur;
		886	}
		887
		888	INLINE void
595	decode_ws (dec_t *dec)	889	decode_ws (dec_t *dec)
596	{	890	{
597	for (;;)	891	for (;;)
598	{	892	{
599	char ch = *dec->cur;	893	char ch = *dec->cur;
600		894
601	if (ch > 0x20	895	if (ch > 0x20)
		896	{
		897	if (expect_false (ch == '#'))
		898	{
		899	if (dec->json.flags & F_RELAXED)
		900	decode_comment (dec);
		901	else
		902	break;
		903	}
		904	else
		905	break;
		906	}
602	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	907	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
603	break;	908	break; // parse error, but let higher level handle it, gives better error messages
604		909
605	++dec->cur;	910	++dec->cur;
606	}	911	}
607	}	912	}
608		913
…		…
612	if (*dec->cur != ch) \	917	if (*dec->cur != ch) \
613	ERR (# ch " expected"); \	918	ERR (# ch " expected"); \
614	++dec->cur; \	919	++dec->cur; \
615	SE	920	SE
616		921
617	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	922	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
618	#define DEC_DEC_DEPTH --dec->depth	923	#define DEC_DEC_DEPTH --dec->depth
619		924
620	static SV *decode_sv (dec_t *dec);	925	static SV *decode_sv (dec_t *dec);
621		926
622	static signed char decode_hexdigit[256];	927	static signed char decode_hexdigit[256];
…		…
625	decode_4hex (dec_t *dec)	930	decode_4hex (dec_t *dec)
626	{	931	{
627	signed char d1, d2, d3, d4;	932	signed char d1, d2, d3, d4;
628	unsigned char *cur = (unsigned char *)dec->cur;	933	unsigned char *cur = (unsigned char *)dec->cur;
629		934
630	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("four hexadecimal digits expected");	935	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
631	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("four hexadecimal digits expected");	936	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
632	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("four hexadecimal digits expected");	937	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
633	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("four hexadecimal digits expected");	938	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
634		939
635	dec->cur += 4;	940	dec->cur += 4;
636		941
637	return ((UV)d1) << 12	942	return ((UV)d1) << 12
638	| ((UV)d2) << 8	943	| ((UV)d2) << 8
…		…
646	static SV *	951	static SV *
647	decode_str (dec_t *dec)	952	decode_str (dec_t *dec)
648	{	953	{
649	SV *sv = 0;	954	SV *sv = 0;
650	int utf8 = 0;	955	int utf8 = 0;
		956	char *dec_cur = dec->cur;
651		957
652	do	958	do
653	{	959	{
654	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];	960	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
655	char *cur = buf;	961	char *cur = buf;
656		962
657	do	963	do
658	{	964	{
659	unsigned char ch = *(unsigned char *)dec->cur++;	965	unsigned char ch = *(unsigned char *)dec_cur++;
660		966
661	if (expect_false (ch == '"'))	967	if (expect_false (ch == '"'))
662	{	968	{
663	--dec->cur;	969	--dec_cur;
664	break;	970	break;
665	}	971	}
666	else if (expect_false (ch == '\\'))	972	else if (expect_false (ch == '\\'))
667	{	973	{
668	switch (*dec->cur)	974	switch (*dec_cur)
669	{	975	{
670	case '\\':	976	case '\\':
671	case '/':	977	case '/':
672	case '"': *cur++ = *dec->cur++; break;	978	case '"': *cur++ = *dec_cur++; break;
673		979
674	case 'b': ++dec->cur; *cur++ = '\010'; break;	980	case 'b': ++dec_cur; *cur++ = '\010'; break;
675	case 't': ++dec->cur; *cur++ = '\011'; break;	981	case 't': ++dec_cur; *cur++ = '\011'; break;
676	case 'n': ++dec->cur; *cur++ = '\012'; break;	982	case 'n': ++dec_cur; *cur++ = '\012'; break;
677	case 'f': ++dec->cur; *cur++ = '\014'; break;	983	case 'f': ++dec_cur; *cur++ = '\014'; break;
678	case 'r': ++dec->cur; *cur++ = '\015'; break;	984	case 'r': ++dec_cur; *cur++ = '\015'; break;
679		985
680	case 'u':	986	case 'u':
681	{	987	{
682	UV lo, hi;	988	UV lo, hi;
683	++dec->cur;	989	++dec_cur;
684		990
		991	dec->cur = dec_cur;
685	hi = decode_4hex (dec);	992	hi = decode_4hex (dec);
		993	dec_cur = dec->cur;
686	if (hi == (UV)-1)	994	if (hi == (UV)-1)
687	goto fail;	995	goto fail;
688		996
689	// possibly a surrogate pair	997	// possibly a surrogate pair
690	if (hi >= 0xd800)	998	if (hi >= 0xd800)
691	if (hi < 0xdc00)	999	if (hi < 0xdc00)
692	{	1000	{
693	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	1001	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
694	ERR ("missing low surrogate character in surrogate pair");	1002	ERR ("missing low surrogate character in surrogate pair");
695		1003
696	dec->cur += 2;	1004	dec_cur += 2;
697		1005
		1006	dec->cur = dec_cur;
698	lo = decode_4hex (dec);	1007	lo = decode_4hex (dec);
		1008	dec_cur = dec->cur;
699	if (lo == (UV)-1)	1009	if (lo == (UV)-1)
700	goto fail;	1010	goto fail;
701		1011
702	if (lo < 0xdc00 || lo >= 0xe000)	1012	if (lo < 0xdc00 || lo >= 0xe000)
703	ERR ("surrogate pair expected");	1013	ERR ("surrogate pair expected");
…		…
709		1019
710	if (hi >= 0x80)	1020	if (hi >= 0x80)
711	{	1021	{
712	utf8 = 1;	1022	utf8 = 1;
713		1023
714	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	1024	cur = encode_utf8 (cur, hi);
715	}	1025	}
716	else	1026	else
717	*cur++ = hi;	1027	*cur++ = hi;
718	}	1028	}
719	break;	1029	break;
720		1030
721	default:	1031	default:
722	--dec->cur;	1032	--dec_cur;
723	ERR ("illegal backslash escape sequence in string");	1033	ERR ("illegal backslash escape sequence in string");
724	}	1034	}
725	}	1035	}
726	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	1036	else if (expect_true (ch >= 0x20 && ch < 0x80))
727	*cur++ = ch;	1037	*cur++ = ch;
728	else if (ch >= 0x80)	1038	else if (ch >= 0x80)
729	{	1039	{
730	STRLEN clen;	1040	STRLEN clen;
731	UV uch;
732		1041
733	--dec->cur;	1042	--dec_cur;
734		1043
735	uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	1044	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
736	if (clen == (STRLEN)-1)	1045	if (clen == (STRLEN)-1)
737	ERR ("malformed UTF-8 character in JSON string");	1046	ERR ("malformed UTF-8 character in JSON string");
738		1047
739	do	1048	do
740	*cur++ = *dec->cur++;	1049	*cur++ = *dec_cur++;
741	while (--clen);	1050	while (--clen);
742		1051
743	utf8 = 1;	1052	utf8 = 1;
744	}	1053	}
745	else	1054	else
746	{	1055	{
747	--dec->cur;	1056	--dec_cur;
748		1057
749	if (!ch)	1058	if (!ch)
750	ERR ("unexpected end of string while parsing JSON string");	1059	ERR ("unexpected end of string while parsing JSON string");
751	else	1060	else
752	ERR ("invalid character encountered while parsing JSON string");	1061	ERR ("invalid character encountered while parsing JSON string");
…		…
757	{	1066	{
758	STRLEN len = cur - buf;	1067	STRLEN len = cur - buf;
759		1068
760	if (sv)	1069	if (sv)
761	{	1070	{
762	SvGROW (sv, SvCUR (sv) + len + 1);	1071	STRLEN cur = SvCUR (sv);
		1072
		1073	if (SvLEN (sv) <= cur + len)
		1074	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1075
763	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	1076	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
764	SvCUR_set (sv, SvCUR (sv) + len);	1077	SvCUR_set (sv, SvCUR (sv) + len);
765	}	1078	}
766	else	1079	else
767	sv = newSVpvn (buf, len);	1080	sv = newSVpvn (buf, len);
768	}	1081	}
769	}	1082	}
770	while (*dec->cur != '"');	1083	while (*dec_cur != '"');
771		1084
772	++dec->cur;	1085	++dec_cur;
773		1086
774	if (sv)	1087	if (sv)
775	{	1088	{
776	SvPOK_only (sv);	1089	SvPOK_only (sv);
777	*SvEND (sv) = 0;	1090	*SvEND (sv) = 0;
…		…
780	SvUTF8_on (sv);	1093	SvUTF8_on (sv);
781	}	1094	}
782	else	1095	else
783	sv = newSVpvn ("", 0);	1096	sv = newSVpvn ("", 0);
784		1097
		1098	dec->cur = dec_cur;
785	return sv;	1099	return sv;
786		1100
787	fail:	1101	fail:
		1102	dec->cur = dec_cur;
788	return 0;	1103	return 0;
789	}	1104	}
790		1105
791	static SV *	1106	static SV *
792	decode_num (dec_t *dec)	1107	decode_num (dec_t *dec)
…		…
850	is_nv = 1;	1165	is_nv = 1;
851	}	1166	}
852		1167
853	if (!is_nv)	1168	if (!is_nv)
854	{	1169	{
		1170	int len = dec->cur - start;
		1171
855	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1172	// special case the rather common 1..5-digit-int case
856	if (*start == '-')	1173	if (*start == '-')
857	switch (dec->cur - start)	1174	switch (len)
858	{	1175	{
859	case 2: return newSViv (-( start [1] - '0' ));	1176	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
860	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1177	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
861	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));	1178	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
862	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));	1179	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1180	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
863	}	1181	}
864	else	1182	else
865	switch (dec->cur - start)	1183	switch (len)
866	{	1184	{
867	case 1: return newSViv ( start [0] - '0' );	1185	case 1: return newSViv ( start [0] - '0' * 1);
868	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1186	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
869	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);	1187	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
870	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);	1188	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1189	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
871	}	1190	}
872		1191
873	{	1192	{
874	UV uv;	1193	UV uv;
875	int numtype = grok_number (start, dec->cur - start, &uv);	1194	int numtype = grok_number (start, len, &uv);
876	if (numtype & IS_NUMBER_IN_UV)	1195	if (numtype & IS_NUMBER_IN_UV)
877	if (numtype & IS_NUMBER_NEG)	1196	if (numtype & IS_NUMBER_NEG)
878	{	1197	{
879	if (uv < (UV)IV_MIN)	1198	if (uv < (UV)IV_MIN)
880	return newSViv (-(IV)uv);	1199	return newSViv (-(IV)uv);
881	}	1200	}
882	else	1201	else
883	return newSVuv (uv);	1202	return newSVuv (uv);
884	}	1203	}
885	}
886		1204
		1205	len -= *start == '-' ? 1 : 0;
		1206
		1207	// does not fit into IV or UV, try NV
		1208	if (len <= NV_DIG)
		1209	// fits into NV without loss of precision
		1210	return newSVnv (json_atof (start));
		1211
		1212	// everything else fails, convert it to a string
		1213	return newSVpvn (start, dec->cur - start);
		1214	}
		1215
		1216	// loss of precision here
887	return newSVnv (Atof (start));	1217	return newSVnv (json_atof (start));
888		1218
889	fail:	1219	fail:
890	return 0;	1220	return 0;
891	}	1221	}
892		1222
…		…
921		1251
922	if (*dec->cur != ',')	1252	if (*dec->cur != ',')
923	ERR (", or ] expected while parsing array");	1253	ERR (", or ] expected while parsing array");
924		1254
925	++dec->cur;	1255	++dec->cur;
		1256
		1257	decode_ws (dec);
		1258
		1259	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1260	{
		1261	++dec->cur;
		1262	break;
		1263	}
926	}	1264	}
927		1265
928	DEC_DEC_DEPTH;	1266	DEC_DEC_DEPTH;
929	return newRV_noinc ((SV *)av);	1267	return newRV_noinc ((SV *)av);
930		1268
…		…
935	}	1273	}
936		1274
937	static SV *	1275	static SV *
938	decode_hv (dec_t *dec)	1276	decode_hv (dec_t *dec)
939	{	1277	{
		1278	SV *sv;
940	HV *hv = newHV ();	1279	HV *hv = newHV ();
941		1280
942	DEC_INC_DEPTH;	1281	DEC_INC_DEPTH;
943	decode_ws (dec);	1282	decode_ws (dec);
944		1283
945	if (*dec->cur == '}')	1284	if (*dec->cur == '}')
946	++dec->cur;	1285	++dec->cur;
947	else	1286	else
948	for (;;)	1287	for (;;)
949	{	1288	{
		1289	EXPECT_CH ('"');
		1290
		1291	// heuristic: assume that
		1292	// a) decode_str + hv_store_ent are abysmally slow.
		1293	// b) most hash keys are short, simple ascii text.
		1294	// => try to "fast-match" such strings to avoid
		1295	// the overhead of decode_str + hv_store_ent.
		1296	{
950	SV *key, *value;	1297	SV *value;
		1298	char *p = dec->cur;
		1299	char *e = p + 24; // only try up to 24 bytes
951		1300
952	decode_ws (dec); EXPECT_CH ('"');	1301	for (;;)
953
954	key = decode_str (dec);
955	if (!key)
956	goto fail;
957
958	decode_ws (dec); EXPECT_CH (':');
959
960	value = decode_sv (dec);
961	if (!value)
962	{	1302	{
		1303	// the >= 0x80 is false on most architectures
		1304	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1305	{
		1306	// slow path, back up and use decode_str
		1307	SV *key = decode_str (dec);
		1308	if (!key)
		1309	goto fail;
		1310
		1311	decode_ws (dec); EXPECT_CH (':');
		1312
		1313	decode_ws (dec);
		1314	value = decode_sv (dec);
		1315	if (!value)
		1316	{
		1317	SvREFCNT_dec (key);
		1318	goto fail;
		1319	}
		1320
		1321	hv_store_ent (hv, key, value, 0);
963	SvREFCNT_dec (key);	1322	SvREFCNT_dec (key);
		1323
		1324	break;
		1325	}
		1326	else if (*p == '"')
		1327	{
		1328	// fast path, got a simple key
		1329	char *key = dec->cur;
		1330	int len = p - key;
		1331	dec->cur = p + 1;
		1332
		1333	decode_ws (dec); EXPECT_CH (':');
		1334
		1335	decode_ws (dec);
		1336	value = decode_sv (dec);
		1337	if (!value)
964	goto fail;	1338	goto fail;
		1339
		1340	hv_store (hv, key, len, value, 0);
		1341
		1342	break;
		1343	}
		1344
		1345	++p;
965	}	1346	}
966		1347	}
967	hv_store_ent (hv, key, value, 0);
968	SvREFCNT_dec (key);
969		1348
970	decode_ws (dec);	1349	decode_ws (dec);
971		1350
972	if (*dec->cur == '}')	1351	if (*dec->cur == '}')
973	{	1352	{
…		…
977		1356
978	if (*dec->cur != ',')	1357	if (*dec->cur != ',')
979	ERR (", or } expected while parsing object/hash");	1358	ERR (", or } expected while parsing object/hash");
980		1359
981	++dec->cur;	1360	++dec->cur;
		1361
		1362	decode_ws (dec);
		1363
		1364	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1365	{
		1366	++dec->cur;
		1367	break;
		1368	}
982	}	1369	}
983		1370
984	DEC_DEC_DEPTH;	1371	DEC_DEC_DEPTH;
985	return newRV_noinc ((SV *)hv);	1372	sv = newRV_noinc ((SV *)hv);
		1373
		1374	// check filter callbacks
		1375	if (dec->json.flags & F_HOOK)
		1376	{
		1377	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1378	{
		1379	HE *cb, *he;
		1380
		1381	hv_iterinit (hv);
		1382	he = hv_iternext (hv);
		1383	hv_iterinit (hv);
		1384
		1385	// the next line creates a mortal sv each time its called.
		1386	// might want to optimise this for common cases.
		1387	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1388
		1389	if (cb)
		1390	{
		1391	dSP;
		1392	int count;
		1393
		1394	ENTER; SAVETMPS; PUSHMARK (SP);
		1395	XPUSHs (HeVAL (he));
		1396
		1397	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1398
		1399	if (count == 1)
		1400	{
		1401	sv = newSVsv (POPs);
		1402	FREETMPS; LEAVE;
		1403	return sv;
		1404	}
		1405
		1406	FREETMPS; LEAVE;
		1407	}
		1408	}
		1409
		1410	if (dec->json.cb_object)
		1411	{
		1412	dSP;
		1413	int count;
		1414
		1415	ENTER; SAVETMPS; PUSHMARK (SP);
		1416	XPUSHs (sv_2mortal (sv));
		1417
		1418	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1419
		1420	if (count == 1)
		1421	{
		1422	sv = newSVsv (POPs);
		1423	FREETMPS; LEAVE;
		1424	return sv;
		1425	}
		1426
		1427	SvREFCNT_inc (sv);
		1428	FREETMPS; LEAVE;
		1429	}
		1430	}
		1431
		1432	return sv;
986		1433
987	fail:	1434	fail:
988	SvREFCNT_dec (hv);	1435	SvREFCNT_dec (hv);
989	DEC_DEC_DEPTH;	1436	DEC_DEC_DEPTH;
990	return 0;	1437	return 0;
991	}	1438	}
992		1439
993	static SV *	1440	static SV *
994	decode_sv (dec_t *dec)	1441	decode_sv (dec_t *dec)
995	{	1442	{
996	decode_ws (dec);	1443	// the beauty of JSON: you need exactly one character lookahead
		1444	// to parse everything.
997	switch (*dec->cur)	1445	switch (*dec->cur)
998	{	1446	{
999	case '"': ++dec->cur; return decode_str (dec);	1447	case '"': ++dec->cur; return decode_str (dec);
1000	case '[': ++dec->cur; return decode_av (dec);	1448	case '[': ++dec->cur; return decode_av (dec);
1001	case '{': ++dec->cur; return decode_hv (dec);	1449	case '{': ++dec->cur; return decode_hv (dec);
1002		1450
1003	case '-':	1451	case '-':
1004	case '0': case '1': case '2': case '3': case '4':	1452	case '0': case '1': case '2': case '3': case '4':
1005	case '5': case '6': case '7': case '8': case '9':	1453	case '5': case '6': case '7': case '8': case '9':
1006	return decode_num (dec);	1454	return decode_num (dec);
1007		1455
1008	case 't':	1456	case 't':
1009	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1457	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1010	{	1458	{
1011	dec->cur += 4;	1459	dec->cur += 4;
		1460	#if JSON_SLOW
		1461	json_true = get_bool ("JSON::XS::true");
		1462	#endif
1012	return newSViv (1);	1463	return newSVsv (json_true);
1013	}	1464	}
1014	else	1465	else
1015	ERR ("'true' expected");	1466	ERR ("'true' expected");
1016		1467
1017	break;	1468	break;
1018		1469
1019	case 'f':	1470	case 'f':
1020	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1471	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1021	{	1472	{
1022	dec->cur += 5;	1473	dec->cur += 5;
		1474	#if JSON_SLOW
		1475	json_false = get_bool ("JSON::XS::false");
		1476	#endif
1023	return newSViv (0);	1477	return newSVsv (json_false);
1024	}	1478	}
1025	else	1479	else
1026	ERR ("'false' expected");	1480	ERR ("'false' expected");
1027		1481
1028	break;	1482	break;
…		…
1046	fail:	1500	fail:
1047	return 0;	1501	return 0;
1048	}	1502	}
1049		1503
1050	static SV *	1504	static SV *
1051	decode_json (SV *string, U32 flags, UV *offset_return)	1505	decode_json (SV *string, JSON *json, char **offset_return)
1052	{	1506	{
1053	dec_t dec;	1507	dec_t dec;
1054	UV offset;
1055	SV *sv;	1508	SV *sv;
1056		1509
		1510	/* work around bugs in 5.10 where manipulating magic values
		1511	* will perl ignore the magic in subsequent accesses
		1512	*/
1057	SvGETMAGIC (string);	1513	/*SvGETMAGIC (string);*/
		1514	if (SvMAGICAL (string))
		1515	string = sv_2mortal (newSVsv (string));
		1516
1058	SvUPGRADE (string, SVt_PV);	1517	SvUPGRADE (string, SVt_PV);
1059		1518
		1519	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1520	* assertion with -DDEBUGGING, although SvCUR is documented to
		1521	* return the xpv_cur field which certainly exists after upgrading.
		1522	* according to nicholas clark, calling SvPOK fixes this.
		1523	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1524	* and hope for the best.
		1525	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1526	* assertion business is seriously broken, try yet another workaround
		1527	* for the broken -DDEBUGGING.
		1528	*/
		1529	{
		1530	#ifdef DEBUGGING
		1531	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1532	#else
		1533	STRLEN offset = SvCUR (string);
		1534	#endif
		1535
		1536	if (offset > json->max_size && json->max_size)
		1537	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1538	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1539	}
		1540
1060	if (flags & F_UTF8)	1541	if (json->flags & F_UTF8)
1061	sv_utf8_downgrade (string, 0);	1542	sv_utf8_downgrade (string, 0);
1062	else	1543	else
1063	sv_utf8_upgrade (string);	1544	sv_utf8_upgrade (string);
1064		1545
1065	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1546	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1066		1547
1067	dec.flags = flags;	1548	dec.json = *json;
1068	dec.cur = SvPVX (string);	1549	dec.cur = SvPVX (string);
1069	dec.end = SvEND (string);	1550	dec.end = SvEND (string);
1070	dec.err = 0;	1551	dec.err = 0;
1071	dec.depth = 0;	1552	dec.depth = 0;
1072	dec.maxdepth = DEC_DEPTH (dec.flags);	1553
		1554	if (dec.json.cb_object || dec.json.cb_sk_object)
		1555	dec.json.flags |= F_HOOK;
1073		1556
1074	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1557	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1558
		1559	decode_ws (&dec);
1075	sv = decode_sv (&dec);	1560	sv = decode_sv (&dec);
		1561
		1562	if (offset_return)
		1563	*offset_return = dec.cur;
1076		1564
1077	if (!(offset_return || !sv))	1565	if (!(offset_return || !sv))
1078	{	1566	{
1079	// check for trailing garbage	1567	// check for trailing garbage
1080	decode_ws (&dec);	1568	decode_ws (&dec);
…		…
1083	{	1571	{
1084	dec.err = "garbage after JSON object";	1572	dec.err = "garbage after JSON object";
1085	SvREFCNT_dec (sv);	1573	SvREFCNT_dec (sv);
1086	sv = 0;	1574	sv = 0;
1087	}	1575	}
1088	}
1089
1090	if (offset_return || !sv)
1091	{
1092	offset = dec.flags & F_UTF8
1093	? dec.cur - SvPVX (string)
1094	: utf8_distance (dec.cur, SvPVX (string));
1095
1096	if (offset_return)
1097	*offset_return = offset;
1098	}	1576	}
1099		1577
1100	if (!sv)	1578	if (!sv)
1101	{	1579	{
1102	SV *uni = sv_newmortal ();	1580	SV *uni = sv_newmortal ();
…		…
1108	SAVEVPTR (PL_curcop);	1586	SAVEVPTR (PL_curcop);
1109	PL_curcop = &cop;	1587	PL_curcop = &cop;
1110	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1588	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1111	LEAVE;	1589	LEAVE;
1112		1590
1113	croak ("%s, at character offset %d [\"%s\"]",	1591	croak ("%s, at character offset %d (before \"%s\")",
1114	dec.err,	1592	dec.err,
1115	(int)offset,	1593	ptr_to_index (string, dec.cur),
1116	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1594	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1117	}	1595	}
1118		1596
1119	sv = sv_2mortal (sv);	1597	sv = sv_2mortal (sv);
1120		1598
1121	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1599	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1122	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1600	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1123		1601
1124	return sv;	1602	return sv;
		1603	}
		1604
		1605	/////////////////////////////////////////////////////////////////////////////
		1606	// incremental parser
		1607
		1608	static void
		1609	incr_parse (JSON *self)
		1610	{
		1611	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1612
		1613	// the state machine here is a bit convoluted and could be simplified a lot
		1614	// but this would make it slower, so...
		1615
		1616	for (;;)
		1617	{
		1618	//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
		1619	switch (self->incr_mode)
		1620	{
		1621	// only used for initial whitespace skipping
		1622	case INCR_M_WS:
		1623	for (;;)
		1624	{
		1625	if (*p > 0x20)
		1626	{
		1627	if (*p == '#')
		1628	{
		1629	self->incr_mode = INCR_M_C0;
		1630	goto incr_m_c;
		1631	}
		1632	else
		1633	{
		1634	self->incr_mode = INCR_M_JSON;
		1635	goto incr_m_json;
		1636	}
		1637	}
		1638	else if (!*p)
		1639	goto interrupt;
		1640
		1641	++p;
		1642	}
		1643
		1644	// skip a single char inside a string (for \\-processing)
		1645	case INCR_M_BS:
		1646	if (!*p)
		1647	goto interrupt;
		1648
		1649	++p;
		1650	self->incr_mode = INCR_M_STR;
		1651	goto incr_m_str;
		1652
		1653	// inside #-style comments
		1654	case INCR_M_C0:
		1655	case INCR_M_C1:
		1656	incr_m_c:
		1657	for (;;)
		1658	{
		1659	if (*p == '\n')
		1660	{
		1661	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1662	break;
		1663	}
		1664	else if (!*p)
		1665	goto interrupt;
		1666
		1667	++p;
		1668	}
		1669
		1670	break;
		1671
		1672	// inside a string
		1673	case INCR_M_STR:
		1674	incr_m_str:
		1675	for (;;)
		1676	{
		1677	if (*p == '"')
		1678	{
		1679	++p;
		1680	self->incr_mode = INCR_M_JSON;
		1681
		1682	if (!self->incr_nest)
		1683	goto interrupt;
		1684
		1685	goto incr_m_json;
		1686	}
		1687	else if (*p == '\\')
		1688	{
		1689	++p; // "virtually" consumes character after \
		1690
		1691	if (!*p) // if at end of string we have to switch modes
		1692	{
		1693	self->incr_mode = INCR_M_BS;
		1694	goto interrupt;
		1695	}
		1696	}
		1697	else if (!*p)
		1698	goto interrupt;
		1699
		1700	++p;
		1701	}
		1702
		1703	// after initial ws, outside string
		1704	case INCR_M_JSON:
		1705	incr_m_json:
		1706	for (;;)
		1707	{
		1708	switch (*p++)
		1709	{
		1710	case 0:
		1711	--p;
		1712	goto interrupt;
		1713
		1714	case 0x09:
		1715	case 0x0a:
		1716	case 0x0d:
		1717	case 0x20:
		1718	if (!self->incr_nest)
		1719	{
		1720	--p; // do not eat the whitespace, let the next round do it
		1721	goto interrupt;
		1722	}
		1723	break;
		1724
		1725	case '"':
		1726	self->incr_mode = INCR_M_STR;
		1727	goto incr_m_str;
		1728
		1729	case '[':
		1730	case '{':
		1731	if (++self->incr_nest > self->max_depth)
		1732	croak (ERR_NESTING_EXCEEDED);
		1733	break;
		1734
		1735	case ']':
		1736	case '}':
		1737	if (--self->incr_nest <= 0)
		1738	goto interrupt;
		1739	break;
		1740
		1741	case '#':
		1742	self->incr_mode = INCR_M_C1;
		1743	goto incr_m_c;
		1744	}
		1745	}
		1746	}
		1747
		1748	modechange:
		1749	;
		1750	}
		1751
		1752	interrupt:
		1753	self->incr_pos = p - SvPVX (self->incr_text);
		1754	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1755	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1125	}	1756	}
1126		1757
1127	/////////////////////////////////////////////////////////////////////////////	1758	/////////////////////////////////////////////////////////////////////////////
1128	// XS interface functions	1759	// XS interface functions
1129		1760
1130	MODULE = JSON::XS PACKAGE = JSON::XS	1761	MODULE = JSON::XS PACKAGE = JSON::XS
1131		1762
1132	BOOT:	1763	BOOT:
1133	{	1764	{
1134	int i;	1765	int i;
1135
1136	memset (decode_hexdigit, 0xff, 256);
1137		1766
1138	for (i = 0; i < 256; ++i)	1767	for (i = 0; i < 256; ++i)
1139	decode_hexdigit [i] =	1768	decode_hexdigit [i] =
1140	i >= '0' && i <= '9' ? i - '0'	1769	i >= '0' && i <= '9' ? i - '0'
1141	: i >= 'a' && i <= 'f' ? i - 'a' + 10	1770	: i >= 'a' && i <= 'f' ? i - 'a' + 10
1142	: i >= 'A' && i <= 'F' ? i - 'A' + 10	1771	: i >= 'A' && i <= 'F' ? i - 'A' + 10
1143	: -1;	1772	: -1;
1144		1773
1145	json_stash = gv_stashpv ("JSON::XS", 1);	1774	json_stash = gv_stashpv ("JSON::XS" , 1);
		1775	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1776
		1777	json_true = get_bool ("JSON::XS::true");
		1778	json_false = get_bool ("JSON::XS::false");
		1779
		1780	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1146	}	1781	}
1147		1782
1148	PROTOTYPES: DISABLE	1783	PROTOTYPES: DISABLE
1149		1784
1150	SV *new (char *dummy)	1785	void CLONE (...)
1151	CODE:	1786	CODE:
1152	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1787	json_stash = 0;
		1788	json_boolean_stash = 0;
		1789
		1790	void new (char *klass)
		1791	PPCODE:
		1792	{
		1793	SV *pv = NEWSV (0, sizeof (JSON));
		1794	SvPOK_only (pv);
		1795	json_init ((JSON *)SvPVX (pv));
		1796	XPUSHs (sv_2mortal (sv_bless (
		1797	newRV_noinc (pv),
		1798	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1799	)));
		1800	}
		1801
		1802	void ascii (JSON *self, int enable = 1)
		1803	ALIAS:
		1804	ascii = F_ASCII
		1805	latin1 = F_LATIN1
		1806	utf8 = F_UTF8
		1807	indent = F_INDENT
		1808	canonical = F_CANONICAL
		1809	space_before = F_SPACE_BEFORE
		1810	space_after = F_SPACE_AFTER
		1811	pretty = F_PRETTY
		1812	allow_nonref = F_ALLOW_NONREF
		1813	shrink = F_SHRINK
		1814	allow_blessed = F_ALLOW_BLESSED
		1815	convert_blessed = F_CONV_BLESSED
		1816	relaxed = F_RELAXED
		1817	allow_unknown = F_ALLOW_UNKNOWN
		1818	PPCODE:
		1819	{
		1820	if (enable)
		1821	self->flags |= ix;
		1822	else
		1823	self->flags &= ~ix;
		1824
		1825	XPUSHs (ST (0));
		1826	}
		1827
		1828	void get_ascii (JSON *self)
		1829	ALIAS:
		1830	get_ascii = F_ASCII
		1831	get_latin1 = F_LATIN1
		1832	get_utf8 = F_UTF8
		1833	get_indent = F_INDENT
		1834	get_canonical = F_CANONICAL
		1835	get_space_before = F_SPACE_BEFORE
		1836	get_space_after = F_SPACE_AFTER
		1837	get_allow_nonref = F_ALLOW_NONREF
		1838	get_shrink = F_SHRINK
		1839	get_allow_blessed = F_ALLOW_BLESSED
		1840	get_convert_blessed = F_CONV_BLESSED
		1841	get_relaxed = F_RELAXED
		1842	get_allow_unknown = F_ALLOW_UNKNOWN
		1843	PPCODE:
		1844	XPUSHs (boolSV (self->flags & ix));
		1845
		1846	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		1847	PPCODE:
		1848	self->max_depth = max_depth;
		1849	XPUSHs (ST (0));
		1850
		1851	U32 get_max_depth (JSON *self)
		1852	CODE:
		1853	RETVAL = self->max_depth;
1153	OUTPUT:	1854	OUTPUT:
1154	RETVAL	1855	RETVAL
1155		1856
1156	SV *ascii (SV *self, int enable = 1)	1857	void max_size (JSON *self, U32 max_size = 0)
1157	ALIAS:	1858	PPCODE:
1158	ascii = F_ASCII	1859	self->max_size = max_size;
1159	latin1 = F_LATIN1	1860	XPUSHs (ST (0));
1160	utf8 = F_UTF8	1861
1161	indent = F_INDENT	1862	int get_max_size (JSON *self)
1162	canonical = F_CANONICAL
1163	space_before = F_SPACE_BEFORE
1164	space_after = F_SPACE_AFTER
1165	pretty = F_PRETTY
1166	allow_nonref = F_ALLOW_NONREF
1167	shrink = F_SHRINK
1168	CODE:	1863	CODE:
1169	{	1864	RETVAL = self->max_size;
1170	UV *uv = SvJSON (self);
1171	if (enable)
1172	*uv |= ix;
1173	else
1174	*uv &= ~ix;
1175
1176	RETVAL = newSVsv (self);
1177	}
1178	OUTPUT:	1865	OUTPUT:
1179	RETVAL	1866	RETVAL
1180		1867
1181	SV *max_depth (SV *self, UV max_depth = 0x80000000UL)	1868	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
		1869	PPCODE:
		1870	{
		1871	SvREFCNT_dec (self->cb_object);
		1872	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1873
		1874	XPUSHs (ST (0));
		1875	}
		1876
		1877	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
		1878	PPCODE:
		1879	{
		1880	if (!self->cb_sk_object)
		1881	self->cb_sk_object = newHV ();
		1882
		1883	if (SvOK (cb))
		1884	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1885	else
		1886	{
		1887	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1888
		1889	if (!HvKEYS (self->cb_sk_object))
		1890	{
		1891	SvREFCNT_dec (self->cb_sk_object);
		1892	self->cb_sk_object = 0;
		1893	}
		1894	}
		1895
		1896	XPUSHs (ST (0));
		1897	}
		1898
		1899	void encode (JSON *self, SV *scalar)
		1900	PPCODE:
		1901	XPUSHs (encode_json (scalar, self));
		1902
		1903	void decode (JSON *self, SV *jsonstr)
		1904	PPCODE:
		1905	XPUSHs (decode_json (jsonstr, self, 0));
		1906
		1907	void decode_prefix (JSON *self, SV *jsonstr)
		1908	PPCODE:
		1909	{
		1910	char *offset;
		1911	EXTEND (SP, 2);
		1912	PUSHs (decode_json (jsonstr, self, &offset));
		1913	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		1914	}
		1915
		1916	void incr_parse (JSON *self, SV *jsonstr = 0)
		1917	PPCODE:
		1918	{
		1919	if (!self->incr_text)
		1920	self->incr_text = newSVpvn ("", 0);
		1921
		1922	// append data, if any
		1923	if (jsonstr)
		1924	{
		1925	if (SvUTF8 (jsonstr))
		1926	{
		1927	if (!SvUTF8 (self->incr_text))
		1928	{
		1929	/* utf-8-ness differs, need to upgrade */
		1930	sv_utf8_upgrade (self->incr_text);
		1931
		1932	if (self->incr_pos)
		1933	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1934	- (U8 *)SvPVX (self->incr_text);
		1935	}
		1936	}
		1937	else if (SvUTF8 (self->incr_text))
		1938	sv_utf8_upgrade (jsonstr);
		1939
		1940	{
		1941	STRLEN len;
		1942	const char *str = SvPV (jsonstr, len);
		1943	STRLEN cur = SvCUR (self->incr_text);
		1944
		1945	if (SvLEN (self->incr_text) <= cur + len)
		1946	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1947
		1948	Move (str, SvEND (self->incr_text), len, char);
		1949	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1950	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1951	}
		1952	}
		1953
		1954	if (GIMME_V != G_VOID)
		1955	do
		1956	{
		1957	char *offset;
		1958
		1959	if (!INCR_DONE (self))
		1960	{
		1961	incr_parse (self);
		1962
		1963	if (self->incr_pos > self->max_size && self->max_size)
		1964	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1965	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1966
		1967	if (!INCR_DONE (self))
		1968	break;
		1969	}
		1970
		1971	XPUSHs (decode_json (self->incr_text, self, &offset));
		1972
		1973	self->incr_pos -= offset - SvPVX (self->incr_text);
		1974	self->incr_nest = 0;
		1975	self->incr_mode = 0;
		1976
		1977	sv_chop (self->incr_text, offset);
		1978	}
		1979	while (GIMME_V == G_ARRAY);
		1980	}
		1981
		1982	SV *incr_text (JSON *self)
		1983	ATTRS: lvalue
1182	CODE:	1984	CODE:
1183	{	1985	{
1184	UV *uv = SvJSON (self);	1986	if (self->incr_pos)
1185	UV log2 = 0;	1987	croak ("incr_text can not be called when the incremental parser already started parsing");
1186		1988
1187	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	1989	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
1188
1189	while ((1UL << log2) < max_depth)
1190	++log2;
1191
1192	*uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1193
1194	RETVAL = newSVsv (self);
1195	}	1990	}
1196	OUTPUT:	1991	OUTPUT:
1197	RETVAL	1992	RETVAL
1198		1993
		1994	void incr_skip (JSON *self)
		1995	CODE:
		1996	{
		1997	if (self->incr_pos)
		1998	{
		1999	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2000	self->incr_pos = 0;
		2001	self->incr_nest = 0;
		2002	self->incr_mode = 0;
		2003	}
		2004	}
		2005
		2006	void incr_reset (JSON *self)
		2007	CODE:
		2008	{
		2009	SvREFCNT_dec (self->incr_text);
		2010	self->incr_text = 0;
		2011	self->incr_pos = 0;
		2012	self->incr_nest = 0;
		2013	self->incr_mode = 0;
		2014	}
		2015
		2016	void DESTROY (JSON *self)
		2017	CODE:
		2018	SvREFCNT_dec (self->cb_sk_object);
		2019	SvREFCNT_dec (self->cb_object);
		2020	SvREFCNT_dec (self->incr_text);
		2021
		2022	PROTOTYPES: ENABLE
		2023
1199	void encode (SV *self, SV *scalar)	2024	void encode_json (SV *scalar)
		2025	ALIAS:
		2026	to_json_ = 0
		2027	encode_json = F_UTF8
1200	PPCODE:	2028	PPCODE:
		2029	{
		2030	JSON json;
		2031	json_init (&json);
		2032	json.flags |= ix;
1201	XPUSHs (encode_json (scalar, *SvJSON (self)));	2033	XPUSHs (encode_json (scalar, &json));
		2034	}
1202		2035
1203	void decode (SV *self, SV *jsonstr)	2036	void decode_json (SV *jsonstr)
		2037	ALIAS:
		2038	from_json_ = 0
		2039	decode_json = F_UTF8
1204	PPCODE:	2040	PPCODE:
		2041	{
		2042	JSON json;
		2043	json_init (&json);
		2044	json.flags |= ix;
1205	XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));	2045	XPUSHs (decode_json (jsonstr, &json, 0));
1206
1207	void decode_prefix (SV *self, SV *jsonstr)
1208	PPCODE:
1209	{
1210	UV offset;
1211	EXTEND (SP, 2);
1212	PUSHs (decode_json (jsonstr, *SvJSON (self), &offset));
1213	PUSHs (sv_2mortal (newSVuv (offset)));
1214	}	2046	}
1215		2047
1216	PROTOTYPES: ENABLE
1217
1218	void to_json (SV *scalar)
1219	ALIAS:
1220	objToJson = 0
1221	PPCODE:
1222	XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8));
1223
1224	void from_json (SV *jsonstr)
1225	ALIAS:
1226	jsonToObj = 0
1227	PPCODE:
1228	XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8, 0));
1229

Diff Legend

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