ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.10 by root, Sat Mar 24 01:15:22 2007 UTC vs.
Revision 1.106 by root, Tue Jan 19 01:36:34 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 0x00000001	25	#define F_ASCII 0x00000001UL
		26	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	27	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	28	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	29	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	30	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	31	#define F_SPACE_AFTER 0x00000040UL
15	#define F_ALLOW_NONREF 0x00000080	32	#define F_ALLOW_NONREF 0x00000100UL
16	#define F_SHRINK 0x00000100	33	#define F_SHRINK 0x00000200UL
		34	#define F_ALLOW_BLESSED 0x00000400UL
		35	#define F_CONV_BLESSED 0x00000800UL
		36	#define F_RELAXED 0x00001000UL
		37	#define F_ALLOW_UNKNOWN 0x00002000UL
		38	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
17		39
18	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	40	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
19	#define F_DEFAULT 0
20		41
21	#define INIT_SIZE 32 // initial scalar size to be allocated	42	#define INIT_SIZE 32 // initial scalar size to be allocated
		43	#define INDENT_STEP 3 // spaces per indentation level
		44
		45	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
22		46
23	#define SB do {	47	#define SB do {
24	#define SE } while (0)	48	#define SE } while (0)
25		49
26	static HV *json_stash;	50	#if __GNUC__ >= 3
		51	# define expect(expr,value) __builtin_expect ((expr), (value))
		52	# define INLINE static inline
		53	#else
		54	# define expect(expr,value) (expr)
		55	# define INLINE static
		56	#endif
		57
		58	#define expect_false(expr) expect ((expr) != 0, 0)
		59	#define expect_true(expr) expect ((expr) != 0, 1)
		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
		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	}
		110
		111	/////////////////////////////////////////////////////////////////////////////
		112	// utility functions
		113
		114	INLINE SV *
		115	get_bool (const char *name)
		116	{
		117	SV *sv = get_sv (name, 1);
		118
		119	SvREADONLY_on (sv);
		120	SvREADONLY_on (SvRV (sv));
		121
		122	return sv;
		123	}
		124
		125	INLINE void
		126	shrink (SV *sv)
		127	{
		128	sv_utf8_downgrade (sv, 1);
		129
		130	if (SvLEN (sv) > SvCUR (sv) + 1)
		131	{
		132	#ifdef SvPV_shrink_to_cur
		133	SvPV_shrink_to_cur (sv);
		134	#elif defined (SvPV_renew)
		135	SvPV_renew (sv, SvCUR (sv) + 1);
		136	#endif
		137	}
		138	}
		139
		140	// decode an utf-8 character and return it, or (UV)-1 in
		141	// case of an error.
		142	// we special-case "safe" characters from U+80 .. U+7FF,
		143	// but use the very good perl function to parse anything else.
		144	// note that we never call this function for a ascii codepoints
		145	INLINE UV
		146	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		147	{
		148	if (expect_true (len >= 2
		149	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
		150	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
		151	{
		152	*clen = 2;
		153	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		154	}
		155	else
		156	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		157	}
		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, int maxdepth)
		198	{
		199	UV uaccum = 0;
		200	int eaccum = 0;
		201
		202	// if we recurse too deep, skip all remaining digits
		203	// to avoid a stack overflow attack
		204	if (expect_false (--maxdepth <= 0))
		205	while (((U8)*s - '0') < 10)
		206	++s;
		207
		208	for (;;)
		209	{
		210	U8 dig = (U8)*s - '0';
		211
		212	if (expect_false (dig >= 10))
		213	{
		214	if (dig == (U8)((U8)'.' - (U8)'0'))
		215	{
		216	++s;
		217	json_atof_scan1 (s, accum, expo, 1, maxdepth);
		218	}
		219	else if ((dig | ' ') == 'e' - '0')
		220	{
		221	int exp2 = 0;
		222	int neg = 0;
		223
		224	++s;
		225
		226	if (*s == '-')
		227	{
		228	++s;
		229	neg = 1;
		230	}
		231	else if (*s == '+')
		232	++s;
		233
		234	while ((dig = (U8)*s - '0') < 10)
		235	exp2 = exp2 * 10 + *s++ - '0';
		236
		237	*expo += neg ? -exp2 : exp2;
		238	}
		239
		240	break;
		241	}
		242
		243	++s;
		244
		245	uaccum = uaccum * 10 + dig;
		246	++eaccum;
		247
		248	// if we have too many digits, then recurse for more
		249	// we actually do this for rather few digits
		250	if (uaccum >= (UV_MAX - 9) / 10)
		251	{
		252	if (postdp) *expo -= eaccum;
		253	json_atof_scan1 (s, accum, expo, postdp, maxdepth);
		254	if (postdp) *expo += eaccum;
		255
		256	break;
		257	}
		258	}
		259
		260	// this relies greatly on the quality of the pow ()
		261	// implementation of the platform, but a good
		262	// implementation is hard to beat.
		263	if (postdp) *expo -= eaccum;
		264	*accum += uaccum * Perl_pow (10., *expo);
		265	*expo += eaccum;
		266	}
		267
		268	static NV
		269	json_atof (const char *s)
		270	{
		271	NV accum = 0.;
		272	int expo = 0;
		273	int neg = 0;
		274
		275	if (*s == '-')
		276	{
		277	++s;
		278	neg = 1;
		279	}
		280
		281	// a recursion depth of ten gives us >>500 bits
		282	json_atof_scan1 (s, &accum, &expo, 0, 10);
		283
		284	return neg ? -accum : accum;
		285	}
		286	/////////////////////////////////////////////////////////////////////////////
		287	// encoder
27		288
28	// structure used for encoding JSON	289	// structure used for encoding JSON
29	typedef struct	290	typedef struct
30	{	291	{
31	char *cur;	292	char *cur; // SvPVX (sv) + current output position
32	STRLEN len; // SvLEN (sv)
33	char *end; // SvEND (sv)	293	char *end; // SvEND (sv)
34	SV *sv;	294	SV *sv; // result scalar
35	UV flags;	295	JSON json;
36	int max_recurse;	296	U32 indent; // indentation level
37	int indent;	297	UV limit; // escape character values >= this value when encoding
38	} enc_t;	298	} enc_t;
39		299
40	// structure used for decoding JSON	300	INLINE void
41	typedef struct
42	{
43	char *cur;
44	char *end;
45	const char *err;
46	UV flags;
47	} dec_t;
48
49	static UV *
50	SvJSON (SV *sv)
51	{
52	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
53	croak ("object is not of type JSON::XS");
54
55	return &SvUVX (SvRV (sv));
56	}
57
58	static void
59	shrink (SV *sv)
60	{
61	sv_utf8_downgrade (sv, 1);
62	#ifdef SvPV_shrink_to_cur
63	SvPV_shrink_to_cur (sv);
64	#endif
65	}
66
67	/////////////////////////////////////////////////////////////////////////////
68
69	static void
70	need (enc_t *enc, STRLEN len)	301	need (enc_t *enc, STRLEN len)
71	{	302	{
72	if (enc->cur + len >= enc->end)	303	if (expect_false (enc->cur + len >= enc->end))
73	{	304	{
74	STRLEN cur = enc->cur - SvPVX (enc->sv);	305	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
75	SvGROW (enc->sv, cur + len + 1);	306	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
76	enc->cur = SvPVX (enc->sv) + cur;	307	enc->cur = SvPVX (enc->sv) + cur;
77	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	308	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
78	}	309	}
79	}	310	}
80		311
81	static void	312	INLINE void
82	encode_ch (enc_t *enc, char ch)	313	encode_ch (enc_t *enc, char ch)
83	{	314	{
84	need (enc, 1);	315	need (enc, 1);
85	*enc->cur++ = ch;	316	*enc->cur++ = ch;
86	}	317	}
…		…
94		325
95	while (str < end)	326	while (str < end)
96	{	327	{
97	unsigned char ch = *(unsigned char *)str;	328	unsigned char ch = *(unsigned char *)str;
98		329
99	if (ch >= 0x20 && ch < 0x80) // most common case	330	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
100	{	331	{
101	if (ch == '"') // but with slow exceptions	332	if (expect_false (ch == '"')) // but with slow exceptions
102	{	333	{
103	need (enc, len += 1);	334	need (enc, len += 1);
104	*enc->cur++ = '\\';	335	*enc->cur++ = '\\';
105	*enc->cur++ = '"';	336	*enc->cur++ = '"';
106	}	337	}
107	else if (ch == '\\')	338	else if (expect_false (ch == '\\'))
108	{	339	{
109	need (enc, len += 1);	340	need (enc, len += 1);
110	*enc->cur++ = '\\';	341	*enc->cur++ = '\\';
111	*enc->cur++ = '\\';	342	*enc->cur++ = '\\';
112	}	343	}
…		…
130	STRLEN clen;	361	STRLEN clen;
131	UV uch;	362	UV uch;
132		363
133	if (is_utf8)	364	if (is_utf8)
134	{	365	{
135	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	366	uch = decode_utf8 (str, end - str, &clen);
136	if (clen == (STRLEN)-1)	367	if (clen == (STRLEN)-1)
137	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	368	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
138	}	369	}
139	else	370	else
140	{	371	{
141	uch = ch;	372	uch = ch;
142	clen = 1;	373	clen = 1;
143	}	374	}
144		375
145	if (uch > 0x10FFFFUL)	376	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
146	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
147
148	if (uch < 0x80 || enc->flags & F_ASCII)
149	{	377	{
150	if (uch > 0xFFFFUL)	378	if (uch >= 0x10000UL)
151	{	379	{
		380	if (uch >= 0x110000UL)
		381	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		382
152	need (enc, len += 11);	383	need (enc, len += 11);
153	sprintf (enc->cur, "\\u%04x\\u%04x",	384	sprintf (enc->cur, "\\u%04x\\u%04x",
154	(int)((uch - 0x10000) / 0x400 + 0xD800),	385	(int)((uch - 0x10000) / 0x400 + 0xD800),
155	(int)((uch - 0x10000) % 0x400 + 0xDC00));	386	(int)((uch - 0x10000) % 0x400 + 0xDC00));
156	enc->cur += 12;	387	enc->cur += 12;
157	}	388	}
158	else	389	else
159	{	390	{
160	static char hexdigit [16] = "0123456789abcdef";
161	need (enc, len += 5);	391	need (enc, len += 5);
162	*enc->cur++ = '\\';	392	*enc->cur++ = '\\';
163	*enc->cur++ = 'u';	393	*enc->cur++ = 'u';
164	*enc->cur++ = hexdigit [ uch >> 12 ];	394	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
165	*enc->cur++ = hexdigit [(uch >> 8) & 15];	395	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
166	*enc->cur++ = hexdigit [(uch >> 4) & 15];	396	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
167	*enc->cur++ = hexdigit [(uch >> 0) & 15];	397	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
168	}	398	}
169		399
		400	str += clen;
		401	}
		402	else if (enc->json.flags & F_LATIN1)
		403	{
		404	*enc->cur++ = uch;
170	str += clen;	405	str += clen;
171	}	406	}
172	else if (is_utf8)	407	else if (is_utf8)
173	{	408	{
174	need (enc, len += clen);	409	need (enc, len += clen);
…		…
178	}	413	}
179	while (--clen);	414	while (--clen);
180	}	415	}
181	else	416	else
182	{	417	{
183	need (enc, len += 10); // never more than 11 bytes needed	418	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
184	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	419	enc->cur = encode_utf8 (enc->cur, uch);
185	++str;	420	++str;
186	}	421	}
187	}	422	}
188	}	423	}
189	}	424	}
190		425
191	--len;	426	--len;
192	}	427	}
193	}	428	}
194		429
195	#define INDENT SB \	430	INLINE void
		431	encode_indent (enc_t *enc)
		432	{
196	if (enc->flags & F_INDENT) \	433	if (enc->json.flags & F_INDENT)
197	{ \	434	{
198	int i_; \	435	int spaces = enc->indent * INDENT_STEP;
199	need (enc, enc->indent); \	436
200	for (i_ = enc->indent * 3; i_--; )\	437	need (enc, spaces);
		438	memset (enc->cur, ' ', spaces);
		439	enc->cur += spaces;
		440	}
		441	}
		442
		443	INLINE void
		444	encode_space (enc_t *enc)
		445	{
		446	need (enc, 1);
		447	encode_ch (enc, ' ');
		448	}
		449
		450	INLINE void
		451	encode_nl (enc_t *enc)
		452	{
		453	if (enc->json.flags & F_INDENT)
		454	{
		455	need (enc, 1);
201	encode_ch (enc, ' '); \	456	encode_ch (enc, '\n');
202	} \	457	}
203	SE	458	}
204		459
205	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	460	INLINE void
206	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	461	encode_comma (enc_t *enc)
207	#define COMMA SB \	462	{
208	encode_ch (enc, ','); \	463	encode_ch (enc, ',');
		464
209	if (enc->flags & F_INDENT) \	465	if (enc->json.flags & F_INDENT)
210	NL; \	466	encode_nl (enc);
211	else if (enc->flags & F_SPACE_AFTER) \	467	else if (enc->json.flags & F_SPACE_AFTER)
212	SPACE; \	468	encode_space (enc);
213	SE	469	}
214		470
215	static void encode_sv (enc_t *enc, SV *sv);	471	static void encode_sv (enc_t *enc, SV *sv);
216		472
217	static void	473	static void
218	encode_av (enc_t *enc, AV *av)	474	encode_av (enc_t *enc, AV *av)
219	{	475	{
220	int i, len = av_len (av);	476	int i, len = av_len (av);
221		477
		478	if (enc->indent >= enc->json.max_depth)
		479	croak (ERR_NESTING_EXCEEDED);
		480
222	encode_ch (enc, '['); NL;	481	encode_ch (enc, '[');
223	++enc->indent;	482
		483	if (len >= 0)
		484	{
		485	encode_nl (enc); ++enc->indent;
224		486
225	for (i = 0; i <= len; ++i)	487	for (i = 0; i <= len; ++i)
226	{	488	{
227	INDENT;	489	SV **svp = av_fetch (av, i, 0);
228	encode_sv (enc, *av_fetch (av, i, 0));
229		490
		491	encode_indent (enc);
		492
		493	if (svp)
		494	encode_sv (enc, *svp);
		495	else
		496	encode_str (enc, "null", 4, 0);
		497
230	if (i < len)	498	if (i < len)
231	COMMA;	499	encode_comma (enc);
232	}	500	}
233		501
234	NL;	502	encode_nl (enc); --enc->indent; encode_indent (enc);
235		503	}
236	--enc->indent;	504
237	INDENT; encode_ch (enc, ']');	505	encode_ch (enc, ']');
238	}	506	}
239		507
240	static void	508	static void
241	encode_he (enc_t *enc, HE *he)	509	encode_hk (enc_t *enc, HE *he)
242	{	510	{
243	encode_ch (enc, '"');	511	encode_ch (enc, '"');
244		512
245	if (HeKLEN (he) == HEf_SVKEY)	513	if (HeKLEN (he) == HEf_SVKEY)
246	{	514	{
…		…
256	else	524	else
257	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	525	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
258		526
259	encode_ch (enc, '"');	527	encode_ch (enc, '"');
260		528
261	if (enc->flags & F_SPACE_BEFORE) SPACE;	529	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
262	encode_ch (enc, ':');	530	encode_ch (enc, ':');
263	if (enc->flags & F_SPACE_AFTER ) SPACE;	531	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
264	encode_sv (enc, HeVAL (he));
265	}	532	}
266		533
267	// compare hash entries, used when all keys are bytestrings	534	// compare hash entries, used when all keys are bytestrings
268	static int	535	static int
269	he_cmp_fast (const void *a_, const void *b_)	536	he_cmp_fast (const void *a_, const void *b_)
…		…
274	HE *b = *(HE **)b_;	541	HE *b = *(HE **)b_;
275		542
276	STRLEN la = HeKLEN (a);	543	STRLEN la = HeKLEN (a);
277	STRLEN lb = HeKLEN (b);	544	STRLEN lb = HeKLEN (b);
278		545
279	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	546	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
280	cmp = la < lb ? -1 : la == lb ? 0 : 1;	547	cmp = lb - la;
281		548
282	return cmp;	549	return cmp;
283	}	550	}
284		551
285	// compare hash entries, used when some keys are sv's or utf-x	552	// compare hash entries, used when some keys are sv's or utf-x
286	static int	553	static int
287	he_cmp_slow (const void *a, const void *b)	554	he_cmp_slow (const void *a, const void *b)
288	{	555	{
289	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	556	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
290	}	557	}
291		558
292	static void	559	static void
293	encode_hv (enc_t *enc, HV *hv)	560	encode_hv (enc_t *enc, HV *hv)
294	{	561	{
295	int count, i;	562	HE *he;
296		563
297	encode_ch (enc, '{'); NL; ++enc->indent;	564	if (enc->indent >= enc->json.max_depth)
		565	croak (ERR_NESTING_EXCEEDED);
298		566
299	if ((count = hv_iterinit (hv)))	567	encode_ch (enc, '{');
300	{	568
301	// for canonical output we have to sort by keys first	569	// for canonical output we have to sort by keys first
302	// actually, this is mostly due to the stupid so-called	570	// actually, this is mostly due to the stupid so-called
303	// security workaround added somewhere in 5.8.x.	571	// security workaround added somewhere in 5.8.x
304	// that randomises hash orderings	572	// that randomises hash orderings
305	if (enc->flags & F_CANONICAL)	573	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
		574	{
		575	int count = hv_iterinit (hv);
		576
		577	if (SvMAGICAL (hv))
306	{	578	{
307	HE *he, *hes [count];	579	// need to count by iterating. could improve by dynamically building the vector below
		580	// but I don't care for the speed of this special case.
		581	// note also that we will run into undefined behaviour when the two iterations
		582	// do not result in the same count, something I might care for in some later release.
		583
		584	count = 0;
		585	while (hv_iternext (hv))
		586	++count;
		587
		588	hv_iterinit (hv);
		589	}
		590
		591	if (count)
		592	{
308	int fast = 1;	593	int i, fast = 1;
		594	#if defined(__BORLANDC__) || defined(_MSC_VER)
		595	HE **hes = _alloca (count * sizeof (HE));
		596	#else
		597	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		598	#endif
309		599
310	i = 0;	600	i = 0;
311	while ((he = hv_iternext (hv)))	601	while ((he = hv_iternext (hv)))
312	{	602	{
313	hes [i++] = he;	603	hes [i++] = he;
…		…
335		625
336	FREETMPS;	626	FREETMPS;
337	LEAVE;	627	LEAVE;
338	}	628	}
339		629
340	for (i = 0; i < count; ++i)	630	encode_nl (enc); ++enc->indent;
		631
		632	while (count--)
341	{	633	{
342	INDENT;	634	encode_indent (enc);
		635	he = hes [count];
343	encode_he (enc, hes [i]);	636	encode_hk (enc, he);
		637	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
344		638
345	if (i < count - 1)	639	if (count)
346	COMMA;	640	encode_comma (enc);
347	}	641	}
348		642
		643	encode_nl (enc); --enc->indent; encode_indent (enc);
		644	}
		645	}
		646	else
		647	{
		648	if (hv_iterinit (hv) || SvMAGICAL (hv))
		649	if ((he = hv_iternext (hv)))
349	NL;	650	{
		651	encode_nl (enc); ++enc->indent;
		652
		653	for (;;)
		654	{
		655	encode_indent (enc);
		656	encode_hk (enc, he);
		657	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		658
		659	if (!(he = hv_iternext (hv)))
		660	break;
		661
		662	encode_comma (enc);
		663	}
		664
		665	encode_nl (enc); --enc->indent; encode_indent (enc);
		666	}
		667	}
		668
		669	encode_ch (enc, '}');
		670	}
		671
		672	// encode objects, arrays and special \0=false and \1=true values.
		673	static void
		674	encode_rv (enc_t *enc, SV *sv)
		675	{
		676	svtype svt;
		677
		678	SvGETMAGIC (sv);
		679	svt = SvTYPE (sv);
		680
		681	if (expect_false (SvOBJECT (sv)))
		682	{
		683	HV *stash = !JSON_SLOW || json_boolean_stash
		684	? json_boolean_stash
		685	: gv_stashpv ("JSON::XS::Boolean", 1);
		686
		687	if (SvSTASH (sv) == stash)
		688	{
		689	if (SvIV (sv))
		690	encode_str (enc, "true", 4, 0);
		691	else
		692	encode_str (enc, "false", 5, 0);
350	}	693	}
351	else	694	else
352	{	695	{
353	SV *sv;	696	#if 0
354	HE *he = hv_iternext (hv);	697	if (0 && sv_derived_from (rv, "JSON::Literal"))
355
356	for (;;)
357	{	698	{
358	INDENT;	699	// not yet
359	encode_he (enc, he);
360
361	if (!(he = hv_iternext (hv)))
362	break;
363
364	COMMA;
365	}	700	}
366		701	#endif
		702	if (enc->json.flags & F_CONV_BLESSED)
367	NL;	703	{
		704	// we re-bless the reference to get overload and other niceties right
		705	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
		706
		707	if (to_json)
		708	{
		709	dSP;
		710
		711	ENTER; SAVETMPS; PUSHMARK (SP);
		712	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		713
		714	// calling with G_SCALAR ensures that we always get a 1 return value
		715	PUTBACK;
		716	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		717	SPAGAIN;
		718
		719	// catch this surprisingly common error
		720	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		721	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		722
		723	sv = POPs;
		724	PUTBACK;
		725
		726	encode_sv (enc, sv);
		727
		728	FREETMPS; LEAVE;
		729	}
		730	else if (enc->json.flags & F_ALLOW_BLESSED)
		731	encode_str (enc, "null", 4, 0);
		732	else
		733	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		734	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		735	}
		736	else if (enc->json.flags & F_ALLOW_BLESSED)
		737	encode_str (enc, "null", 4, 0);
		738	else
		739	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		740	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
368	}	741	}
369	}	742	}
		743	else if (svt == SVt_PVHV)
		744	encode_hv (enc, (HV *)sv);
		745	else if (svt == SVt_PVAV)
		746	encode_av (enc, (AV *)sv);
		747	else if (svt < SVt_PVAV)
		748	{
		749	STRLEN len = 0;
		750	char *pv = svt ? SvPV (sv, len) : 0;
370		751
371	--enc->indent; INDENT; encode_ch (enc, '}');	752	if (len == 1 && *pv == '1')
		753	encode_str (enc, "true", 4, 0);
		754	else if (len == 1 && *pv == '0')
		755	encode_str (enc, "false", 5, 0);
		756	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		757	encode_str (enc, "null", 4, 0);
		758	else
		759	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		760	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		761	}
		762	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		763	encode_str (enc, "null", 4, 0);
		764	else
		765	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		766	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
372	}	767	}
373		768
374	static void	769	static void
375	encode_sv (enc_t *enc, SV *sv)	770	encode_sv (enc_t *enc, SV *sv)
376	{	771	{
…		…
384	encode_str (enc, str, len, SvUTF8 (sv));	779	encode_str (enc, str, len, SvUTF8 (sv));
385	encode_ch (enc, '"');	780	encode_ch (enc, '"');
386	}	781	}
387	else if (SvNOKp (sv))	782	else if (SvNOKp (sv))
388	{	783	{
		784	// trust that perl will do the right thing w.r.t. JSON syntax.
389	need (enc, NV_DIG + 32);	785	need (enc, NV_DIG + 32);
390	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	786	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
391	enc->cur += strlen (enc->cur);	787	enc->cur += strlen (enc->cur);
392	}	788	}
393	else if (SvIOKp (sv))	789	else if (SvIOKp (sv))
394	{	790	{
		791	// we assume we can always read an IV as a UV and vice versa
		792	// we assume two's complement
		793	// we assume no aliasing issues in the union
		794	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		795	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		796	{
		797	// optimise the "small number case"
		798	// code will likely be branchless and use only a single multiplication
		799	// works for numbers up to 59074
		800	I32 i = SvIVX (sv);
		801	U32 u;
		802	char digit, nz = 0;
		803
395	need (enc, 64);	804	need (enc, 6);
		805
		806	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		807	u = i < 0 ? -i : i;
		808
		809	// convert to 4.28 fixed-point representation
		810	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		811
		812	// now output digit by digit, each time masking out the integer part
		813	// and multiplying by 5 while moving the decimal point one to the right,
		814	// resulting in a net multiplication by 10.
		815	// we always write the digit to memory but conditionally increment
		816	// the pointer, to enable the use of conditional move instructions.
		817	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		818	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		819	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		820	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		821	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		822	}
		823	else
		824	{
		825	// large integer, use the (rather slow) snprintf way.
		826	need (enc, IVUV_MAXCHARS);
396	enc->cur +=	827	enc->cur +=
397	SvIsUV(sv)	828	SvIsUV(sv)
398	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	829	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
399	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	830	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		831	}
400	}	832	}
401	else if (SvROK (sv))	833	else if (SvROK (sv))
402	{	834	encode_rv (enc, SvRV (sv));
403	SV *rv = SvRV (sv);	835	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
404
405	if (!--enc->max_recurse)
406	croak ("data structure too deep (hit recursion limit)");
407
408	switch (SvTYPE (rv))
409	{
410	case SVt_PVAV: encode_av (enc, (AV *)rv); break;
411	case SVt_PVHV: encode_hv (enc, (HV *)rv); break;
412
413	default:
414	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
415	SvPV_nolen (sv));
416	}
417	}
418	else if (!SvOK (sv))
419	encode_str (enc, "null", 4, 0);	836	encode_str (enc, "null", 4, 0);
420	else	837	else
421	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	838	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
422	SvPV_nolen (sv), SvFLAGS (sv));	839	SvPV_nolen (sv), SvFLAGS (sv));
423	}	840	}
424		841
425	static SV *	842	static SV *
426	encode_json (SV *scalar, UV flags)	843	encode_json (SV *scalar, JSON *json)
427	{	844	{
		845	enc_t enc;
		846
428	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	847	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
429	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	848	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
430		849
431	enc_t enc;	850	enc.json = *json;
432	enc.flags = flags;
433	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	851	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
434	enc.cur = SvPVX (enc.sv);	852	enc.cur = SvPVX (enc.sv);
435	enc.end = SvEND (enc.sv);	853	enc.end = SvEND (enc.sv);
436	enc.max_recurse = 0;
437	enc.indent = 0;	854	enc.indent = 0;
		855	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		856	: enc.json.flags & F_LATIN1 ? 0x000100UL
		857	: 0x110000UL;
438		858
439	SvPOK_only (enc.sv);	859	SvPOK_only (enc.sv);
440	encode_sv (&enc, scalar);	860	encode_sv (&enc, scalar);
		861	encode_nl (&enc);
441		862
		863	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		864	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		865
442	if (!(flags & (F_ASCII | F_UTF8)))	866	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
443	SvUTF8_on (enc.sv);	867	SvUTF8_on (enc.sv);
444		868
445	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
446
447	if (enc.flags & F_SHRINK)	869	if (enc.json.flags & F_SHRINK)
448	shrink (enc.sv);	870	shrink (enc.sv);
449		871
450	return enc.sv;	872	return enc.sv;
451	}	873	}
452		874
453	/////////////////////////////////////////////////////////////////////////////	875	/////////////////////////////////////////////////////////////////////////////
		876	// decoder
454		877
455	#define WS \	878	// structure used for decoding JSON
		879	typedef struct
		880	{
		881	char *cur; // current parser pointer
		882	char *end; // end of input string
		883	const char *err; // parse error, if != 0
		884	JSON json;
		885	U32 depth; // recursion depth
		886	U32 maxdepth; // recursion depth limit
		887	} dec_t;
		888
		889	INLINE void
		890	decode_comment (dec_t *dec)
		891	{
		892	// only '#'-style comments allowed a.t.m.
		893
		894	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		895	++dec->cur;
		896	}
		897
		898	INLINE void
		899	decode_ws (dec_t *dec)
		900	{
456	for (;;) \	901	for (;;)
457	{ \	902	{
458	char ch = *dec->cur; \	903	char ch = *dec->cur;
		904
459	if (ch > 0x20 \	905	if (ch > 0x20)
		906	{
		907	if (expect_false (ch == '#'))
		908	{
		909	if (dec->json.flags & F_RELAXED)
		910	decode_comment (dec);
		911	else
		912	break;
		913	}
		914	else
		915	break;
		916	}
460	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	917	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
461	break; \	918	break; // parse error, but let higher level handle it, gives better error messages
		919
462	++dec->cur; \	920	++dec->cur;
463	}	921	}
		922	}
464		923
465	#define ERR(reason) SB dec->err = reason; goto fail; SE	924	#define ERR(reason) SB dec->err = reason; goto fail; SE
		925
466	#define EXPECT_CH(ch) SB \	926	#define EXPECT_CH(ch) SB \
467	if (*dec->cur != ch) \	927	if (*dec->cur != ch) \
468	ERR (# ch " expected"); \	928	ERR (# ch " expected"); \
469	++dec->cur; \	929	++dec->cur; \
470	SE	930	SE
471		931
		932	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
		933	#define DEC_DEC_DEPTH --dec->depth
		934
472	static SV *decode_sv (dec_t *dec);	935	static SV *decode_sv (dec_t *dec);
473		936
474	static signed char decode_hexdigit[256];	937	static signed char decode_hexdigit[256];
475		938
476	static UV	939	static UV
477	decode_4hex (dec_t *dec)	940	decode_4hex (dec_t *dec)
478	{	941	{
479	signed char d1, d2, d3, d4;	942	signed char d1, d2, d3, d4;
		943	unsigned char *cur = (unsigned char *)dec->cur;
480		944
481	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	945	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
482	if (d1 < 0) ERR ("four hexadecimal digits expected");	946	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
483	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	947	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
484	if (d2 < 0) ERR ("four hexadecimal digits expected");	948	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
485	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
486	if (d3 < 0) ERR ("four hexadecimal digits expected");
487	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
488	if (d4 < 0) ERR ("four hexadecimal digits expected");
489		949
490	dec->cur += 4;	950	dec->cur += 4;
491		951
492	return ((UV)d1) << 12	952	return ((UV)d1) << 12
493	| ((UV)d2) << 8	953	| ((UV)d2) << 8
…		…
496		956
497	fail:	957	fail:
498	return (UV)-1;	958	return (UV)-1;
499	}	959	}
500		960
501	#define APPEND_GROW(n) SB \
502	if (cur + (n) >= end) \
503	{ \
504	STRLEN ofs = cur - SvPVX (sv); \
505	SvGROW (sv, ofs + (n) + 1); \
506	cur = SvPVX (sv) + ofs; \
507	end = SvEND (sv); \
508	} \
509	SE
510
511	#define APPEND_CH(ch) SB \
512	APPEND_GROW (1); \
513	*cur++ = (ch); \
514	SE
515
516	static SV *	961	static SV *
517	decode_str (dec_t *dec)	962	decode_str (dec_t *dec)
518	{	963	{
519	SV *sv = NEWSV (0,2);	964	SV *sv = 0;
520	int utf8 = 0;	965	int utf8 = 0;
521	char *cur = SvPVX (sv);	966	char *dec_cur = dec->cur;
522	char *end = SvEND (sv);
523		967
524	for (;;)	968	do
525	{	969	{
526	unsigned char ch = *(unsigned char *)dec->cur;	970	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		971	char *cur = buf;
527		972
528	if (ch == '"')	973	do
529	break;
530	else if (ch == '\\')
531	{	974	{
532	switch (*++dec->cur)	975	unsigned char ch = *(unsigned char *)dec_cur++;
		976
		977	if (expect_false (ch == '"'))
533	{	978	{
534	case '\\':	979	--dec_cur;
535	case '/':	980	break;
536	case '"': APPEND_CH (*dec->cur++); break;	981	}
537		982	else if (expect_false (ch == '\\'))
538	case 'b': APPEND_CH ('\010'); ++dec->cur; break;	983	{
539	case 't': APPEND_CH ('\011'); ++dec->cur; break;	984	switch (*dec_cur)
540	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
541	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
542	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
543
544	case 'u':
545	{	985	{
546	UV lo, hi;	986	case '\\':
547	++dec->cur;	987	case '/':
		988	case '"': *cur++ = *dec_cur++; break;
548		989
549	hi = decode_4hex (dec);	990	case 'b': ++dec_cur; *cur++ = '\010'; break;
550	if (hi == (UV)-1)	991	case 't': ++dec_cur; *cur++ = '\011'; break;
551	goto fail;	992	case 'n': ++dec_cur; *cur++ = '\012'; break;
		993	case 'f': ++dec_cur; *cur++ = '\014'; break;
		994	case 'r': ++dec_cur; *cur++ = '\015'; break;
552		995
553	// possibly a surrogate pair	996	case 'u':
554	if (hi >= 0xd800 && hi < 0xdc00)
555	{	997	{
556	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	998	UV lo, hi;
557	ERR ("missing low surrogate character in surrogate pair");	999	++dec_cur;
558		1000
559	dec->cur += 2;	1001	dec->cur = dec_cur;
560
561	lo = decode_4hex (dec);	1002	hi = decode_4hex (dec);
		1003	dec_cur = dec->cur;
562	if (lo == (UV)-1)	1004	if (hi == (UV)-1)
563	goto fail;	1005	goto fail;
564		1006
		1007	// possibly a surrogate pair
		1008	if (hi >= 0xd800)
		1009	if (hi < 0xdc00)
		1010	{
		1011	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		1012	ERR ("missing low surrogate character in surrogate pair");
		1013
		1014	dec_cur += 2;
		1015
		1016	dec->cur = dec_cur;
		1017	lo = decode_4hex (dec);
		1018	dec_cur = dec->cur;
		1019	if (lo == (UV)-1)
		1020	goto fail;
		1021
565	if (lo < 0xdc00 || lo >= 0xe000)	1022	if (lo < 0xdc00 || lo >= 0xe000)
566	ERR ("surrogate pair expected");	1023	ERR ("surrogate pair expected");
567		1024
568	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	1025	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		1026	}
		1027	else if (hi < 0xe000)
		1028	ERR ("missing high surrogate character in surrogate pair");
		1029
		1030	if (hi >= 0x80)
		1031	{
		1032	utf8 = 1;
		1033
		1034	cur = encode_utf8 (cur, hi);
		1035	}
		1036	else
		1037	*cur++ = hi;
569	}	1038	}
570	else if (hi >= 0xdc00 && hi < 0xe000)
571	ERR ("missing high surrogate character in surrogate pair");
572
573	if (hi >= 0x80)
574	{	1039	break;
575	utf8 = 1;
576		1040
577	APPEND_GROW (4); // at most 4 bytes for 21 bits
578	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
579	}
580	else	1041	default:
581	APPEND_CH (hi);	1042	--dec_cur;
		1043	ERR ("illegal backslash escape sequence in string");
582	}	1044	}
583	break;	1045	}
		1046	else if (expect_true (ch >= 0x20 && ch < 0x80))
		1047	*cur++ = ch;
		1048	else if (ch >= 0x80)
		1049	{
		1050	STRLEN clen;
584		1051
585	default:
586	--dec->cur;	1052	--dec_cur;
587	ERR ("illegal backslash escape sequence in string");	1053
		1054	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		1055	if (clen == (STRLEN)-1)
		1056	ERR ("malformed UTF-8 character in JSON string");
		1057
		1058	do
		1059	*cur++ = *dec_cur++;
		1060	while (--clen);
		1061
		1062	utf8 = 1;
		1063	}
		1064	else
		1065	{
		1066	--dec_cur;
		1067
		1068	if (!ch)
		1069	ERR ("unexpected end of string while parsing JSON string");
		1070	else
		1071	ERR ("invalid character encountered while parsing JSON string");
588	}	1072	}
589	}	1073	}
590	else if (ch >= 0x20 && ch <= 0x7f)	1074	while (cur < buf + SHORT_STRING_LEN);
591	APPEND_CH (*dec->cur++);	1075
592	else if (ch >= 0x80)
593	{	1076	{
594	STRLEN clen;	1077	STRLEN len = cur - buf;
595	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
596	if (clen == (STRLEN)-1)
597	ERR ("malformed UTF-8 character in JSON string");
598		1078
599	APPEND_GROW (clen);	1079	if (sv)
600	do
601	{	1080	{
602	*cur++ = *dec->cur++;	1081	STRLEN cur = SvCUR (sv);
		1082
		1083	if (SvLEN (sv) <= cur + len)
		1084	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1085
		1086	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		1087	SvCUR_set (sv, SvCUR (sv) + len);
603	}	1088	}
604	while (--clen);
605
606	utf8 = 1;
607	}
608	else if (dec->cur == dec->end)
609	ERR ("unexpected end of string while parsing json string");
610	else	1089	else
611	ERR ("invalid character encountered");	1090	sv = newSVpvn (buf, len);
612	}	1091	}
		1092	}
		1093	while (*dec_cur != '"');
613		1094
614	++dec->cur;	1095	++dec_cur;
615		1096
616	SvCUR_set (sv, cur - SvPVX (sv));	1097	if (sv)
617		1098	{
618	SvPOK_only (sv);	1099	SvPOK_only (sv);
619	*SvEND (sv) = 0;	1100	*SvEND (sv) = 0;
620		1101
621	if (utf8)	1102	if (utf8)
622	SvUTF8_on (sv);	1103	SvUTF8_on (sv);
		1104	}
		1105	else
		1106	sv = newSVpvn ("", 0);
623		1107
624	if (dec->flags & F_SHRINK)	1108	dec->cur = dec_cur;
625	shrink (sv);
626
627	return sv;	1109	return sv;
628		1110
629	fail:	1111	fail:
630	SvREFCNT_dec (sv);	1112	dec->cur = dec_cur;
631	return 0;	1113	return 0;
632	}	1114	}
633		1115
634	static SV *	1116	static SV *
635	decode_num (dec_t *dec)	1117	decode_num (dec_t *dec)
…		…
693	is_nv = 1;	1175	is_nv = 1;
694	}	1176	}
695		1177
696	if (!is_nv)	1178	if (!is_nv)
697	{	1179	{
698	UV uv;	1180	int len = dec->cur - start;
699	int numtype = grok_number (start, dec->cur - start, &uv);	1181
700	if (numtype & IS_NUMBER_IN_UV)	1182	// special case the rather common 1..5-digit-int case
701	if (numtype & IS_NUMBER_NEG)	1183	if (*start == '-')
		1184	switch (len)
702	{	1185	{
703	if (uv < (UV)IV_MIN)	1186	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
704	return newSViv (-(IV)uv);	1187	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
		1188	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1189	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1190	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
705	}	1191	}
		1192	else
		1193	switch (len)
		1194	{
		1195	case 1: return newSViv ( start [0] - '0' * 1);
		1196	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1197	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1198	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1199	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1200	}
		1201
		1202	{
		1203	UV uv;
		1204	int numtype = grok_number (start, len, &uv);
		1205	if (numtype & IS_NUMBER_IN_UV)
		1206	if (numtype & IS_NUMBER_NEG)
		1207	{
		1208	if (uv < (UV)IV_MIN)
		1209	return newSViv (-(IV)uv);
		1210	}
706	else	1211	else
707	return newSVuv (uv);	1212	return newSVuv (uv);
708	}	1213	}
709		1214
		1215	len -= *start == '-' ? 1 : 0;
		1216
		1217	// does not fit into IV or UV, try NV
		1218	if (len <= NV_DIG)
		1219	// fits into NV without loss of precision
		1220	return newSVnv (json_atof (start));
		1221
		1222	// everything else fails, convert it to a string
		1223	return newSVpvn (start, dec->cur - start);
		1224	}
		1225
		1226	// loss of precision here
710	return newSVnv (Atof (start));	1227	return newSVnv (json_atof (start));
711		1228
712	fail:	1229	fail:
713	return 0;	1230	return 0;
714	}	1231	}
715		1232
716	static SV *	1233	static SV *
717	decode_av (dec_t *dec)	1234	decode_av (dec_t *dec)
718	{	1235	{
719	AV *av = newAV ();	1236	AV *av = newAV ();
720		1237
721	WS;	1238	DEC_INC_DEPTH;
		1239	decode_ws (dec);
		1240
722	if (*dec->cur == ']')	1241	if (*dec->cur == ']')
723	++dec->cur;	1242	++dec->cur;
724	else	1243	else
725	for (;;)	1244	for (;;)
726	{	1245	{
…		…
730	if (!value)	1249	if (!value)
731	goto fail;	1250	goto fail;
732		1251
733	av_push (av, value);	1252	av_push (av, value);
734		1253
735	WS;	1254	decode_ws (dec);
736		1255
737	if (*dec->cur == ']')	1256	if (*dec->cur == ']')
738	{	1257	{
739	++dec->cur;	1258	++dec->cur;
740	break;	1259	break;
…		…
742		1261
743	if (*dec->cur != ',')	1262	if (*dec->cur != ',')
744	ERR (", or ] expected while parsing array");	1263	ERR (", or ] expected while parsing array");
745		1264
746	++dec->cur;	1265	++dec->cur;
		1266
		1267	decode_ws (dec);
		1268
		1269	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1270	{
		1271	++dec->cur;
		1272	break;
		1273	}
747	}	1274	}
748		1275
		1276	DEC_DEC_DEPTH;
749	return newRV_noinc ((SV *)av);	1277	return newRV_noinc ((SV *)av);
750		1278
751	fail:	1279	fail:
752	SvREFCNT_dec (av);	1280	SvREFCNT_dec (av);
		1281	DEC_DEC_DEPTH;
753	return 0;	1282	return 0;
754	}	1283	}
755		1284
756	static SV *	1285	static SV *
757	decode_hv (dec_t *dec)	1286	decode_hv (dec_t *dec)
758	{	1287	{
		1288	SV *sv;
759	HV *hv = newHV ();	1289	HV *hv = newHV ();
760		1290
761	WS;	1291	DEC_INC_DEPTH;
		1292	decode_ws (dec);
		1293
762	if (*dec->cur == '}')	1294	if (*dec->cur == '}')
763	++dec->cur;	1295	++dec->cur;
764	else	1296	else
765	for (;;)	1297	for (;;)
766	{	1298	{
767	SV *key, *value;
768
769	WS; EXPECT_CH ('"');	1299	EXPECT_CH ('"');
770		1300
771	key = decode_str (dec);	1301	// heuristic: assume that
772	if (!key)	1302	// a) decode_str + hv_store_ent are abysmally slow.
773	goto fail;	1303	// b) most hash keys are short, simple ascii text.
		1304	// => try to "fast-match" such strings to avoid
		1305	// the overhead of decode_str + hv_store_ent.
		1306	{
		1307	SV *value;
		1308	char *p = dec->cur;
		1309	char *e = p + 24; // only try up to 24 bytes
774		1310
775	WS; EXPECT_CH (':');	1311	for (;;)
776
777	value = decode_sv (dec);
778	if (!value)
779	{	1312	{
		1313	// the >= 0x80 is false on most architectures
		1314	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1315	{
		1316	// slow path, back up and use decode_str
		1317	SV *key = decode_str (dec);
		1318	if (!key)
		1319	goto fail;
		1320
		1321	decode_ws (dec); EXPECT_CH (':');
		1322
		1323	decode_ws (dec);
		1324	value = decode_sv (dec);
		1325	if (!value)
		1326	{
		1327	SvREFCNT_dec (key);
		1328	goto fail;
		1329	}
		1330
		1331	hv_store_ent (hv, key, value, 0);
780	SvREFCNT_dec (key);	1332	SvREFCNT_dec (key);
		1333
		1334	break;
		1335	}
		1336	else if (*p == '"')
		1337	{
		1338	// fast path, got a simple key
		1339	char *key = dec->cur;
		1340	int len = p - key;
		1341	dec->cur = p + 1;
		1342
		1343	decode_ws (dec); EXPECT_CH (':');
		1344
		1345	decode_ws (dec);
		1346	value = decode_sv (dec);
		1347	if (!value)
781	goto fail;	1348	goto fail;
		1349
		1350	hv_store (hv, key, len, value, 0);
		1351
		1352	break;
		1353	}
		1354
		1355	++p;
782	}	1356	}
783
784	//TODO: optimise
785	hv_store_ent (hv, key, value, 0);
786
787	WS;	1357	}
		1358
		1359	decode_ws (dec);
788		1360
789	if (*dec->cur == '}')	1361	if (*dec->cur == '}')
790	{	1362	{
791	++dec->cur;	1363	++dec->cur;
792	break;	1364	break;
…		…
794		1366
795	if (*dec->cur != ',')	1367	if (*dec->cur != ',')
796	ERR (", or } expected while parsing object/hash");	1368	ERR (", or } expected while parsing object/hash");
797		1369
798	++dec->cur;	1370	++dec->cur;
		1371
		1372	decode_ws (dec);
		1373
		1374	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1375	{
		1376	++dec->cur;
		1377	break;
		1378	}
799	}	1379	}
800		1380
		1381	DEC_DEC_DEPTH;
801	return newRV_noinc ((SV *)hv);	1382	sv = newRV_noinc ((SV *)hv);
		1383
		1384	// check filter callbacks
		1385	if (dec->json.flags & F_HOOK)
		1386	{
		1387	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1388	{
		1389	HE *cb, *he;
		1390
		1391	hv_iterinit (hv);
		1392	he = hv_iternext (hv);
		1393	hv_iterinit (hv);
		1394
		1395	// the next line creates a mortal sv each time its called.
		1396	// might want to optimise this for common cases.
		1397	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1398
		1399	if (cb)
		1400	{
		1401	dSP;
		1402	int count;
		1403
		1404	ENTER; SAVETMPS; PUSHMARK (SP);
		1405	XPUSHs (HeVAL (he));
		1406
		1407	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1408
		1409	if (count == 1)
		1410	{
		1411	sv = newSVsv (POPs);
		1412	FREETMPS; LEAVE;
		1413	return sv;
		1414	}
		1415
		1416	FREETMPS; LEAVE;
		1417	}
		1418	}
		1419
		1420	if (dec->json.cb_object)
		1421	{
		1422	dSP;
		1423	int count;
		1424
		1425	ENTER; SAVETMPS; PUSHMARK (SP);
		1426	XPUSHs (sv_2mortal (sv));
		1427
		1428	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1429
		1430	if (count == 1)
		1431	{
		1432	sv = newSVsv (POPs);
		1433	FREETMPS; LEAVE;
		1434	return sv;
		1435	}
		1436
		1437	SvREFCNT_inc (sv);
		1438	FREETMPS; LEAVE;
		1439	}
		1440	}
		1441
		1442	return sv;
802		1443
803	fail:	1444	fail:
804	SvREFCNT_dec (hv);	1445	SvREFCNT_dec (hv);
		1446	DEC_DEC_DEPTH;
805	return 0;	1447	return 0;
806	}	1448	}
807		1449
808	static SV *	1450	static SV *
809	decode_sv (dec_t *dec)	1451	decode_sv (dec_t *dec)
810	{	1452	{
811	WS;	1453	// the beauty of JSON: you need exactly one character lookahead
		1454	// to parse everything.
812	switch (*dec->cur)	1455	switch (*dec->cur)
813	{	1456	{
814	case '"': ++dec->cur; return decode_str (dec);	1457	case '"': ++dec->cur; return decode_str (dec);
815	case '[': ++dec->cur; return decode_av (dec);	1458	case '[': ++dec->cur; return decode_av (dec);
816	case '{': ++dec->cur; return decode_hv (dec);	1459	case '{': ++dec->cur; return decode_hv (dec);
817		1460
818	case '-':	1461	case '-':
819	case '0': case '1': case '2': case '3': case '4':	1462	case '0': case '1': case '2': case '3': case '4':
820	case '5': case '6': case '7': case '8': case '9':	1463	case '5': case '6': case '7': case '8': case '9':
821	return decode_num (dec);	1464	return decode_num (dec);
822		1465
823	case 't':	1466	case 't':
824	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1467	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
825	{	1468	{
826	dec->cur += 4;	1469	dec->cur += 4;
		1470	#if JSON_SLOW
		1471	json_true = get_bool ("JSON::XS::true");
		1472	#endif
827	return newSViv (1);	1473	return newSVsv (json_true);
828	}	1474	}
829	else	1475	else
830	ERR ("'true' expected");	1476	ERR ("'true' expected");
831		1477
832	break;	1478	break;
833		1479
834	case 'f':	1480	case 'f':
835	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1481	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
836	{	1482	{
837	dec->cur += 5;	1483	dec->cur += 5;
		1484	#if JSON_SLOW
		1485	json_false = get_bool ("JSON::XS::false");
		1486	#endif
838	return newSViv (0);	1487	return newSVsv (json_false);
839	}	1488	}
840	else	1489	else
841	ERR ("'false' expected");	1490	ERR ("'false' expected");
842		1491
843	break;	1492	break;
…		…
852	ERR ("'null' expected");	1501	ERR ("'null' expected");
853		1502
854	break;	1503	break;
855		1504
856	default:	1505	default:
857	ERR ("malformed json string, neither array, object, number, string or atom");	1506	ERR ("malformed JSON string, neither array, object, number, string or atom");
858	break;	1507	break;
859	}	1508	}
860		1509
861	fail:	1510	fail:
862	return 0;	1511	return 0;
863	}	1512	}
864		1513
865	static SV *	1514	static SV *
866	decode_json (SV *string, UV flags)	1515	decode_json (SV *string, JSON *json, char **offset_return)
867	{	1516	{
		1517	dec_t dec;
868	SV *sv;	1518	SV *sv;
869		1519
		1520	/* work around bugs in 5.10 where manipulating magic values
		1521	* will perl ignore the magic in subsequent accesses
		1522	*/
		1523	/*SvGETMAGIC (string);*/
		1524	if (SvMAGICAL (string))
		1525	string = sv_2mortal (newSVsv (string));
		1526
		1527	SvUPGRADE (string, SVt_PV);
		1528
		1529	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1530	* assertion with -DDEBUGGING, although SvCUR is documented to
		1531	* return the xpv_cur field which certainly exists after upgrading.
		1532	* according to nicholas clark, calling SvPOK fixes this.
		1533	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1534	* and hope for the best.
		1535	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1536	* assertion business is seriously broken, try yet another workaround
		1537	* for the broken -DDEBUGGING.
		1538	*/
		1539	{
		1540	#ifdef DEBUGGING
		1541	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1542	#else
		1543	STRLEN offset = SvCUR (string);
		1544	#endif
		1545
		1546	if (offset > json->max_size && json->max_size)
		1547	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1548	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1549	}
		1550
870	if (flags & F_UTF8)	1551	if (json->flags & F_UTF8)
871	sv_utf8_downgrade (string, 0);	1552	sv_utf8_downgrade (string, 0);
872	else	1553	else
873	sv_utf8_upgrade (string);	1554	sv_utf8_upgrade (string);
874		1555
875	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1556	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
876		1557
877	dec_t dec;	1558	dec.json = *json;
878	dec.flags = flags;
879	dec.cur = SvPVX (string);	1559	dec.cur = SvPVX (string);
880	dec.end = SvEND (string);	1560	dec.end = SvEND (string);
881	dec.err = 0;	1561	dec.err = 0;
		1562	dec.depth = 0;
882		1563
		1564	if (dec.json.cb_object || dec.json.cb_sk_object)
		1565	dec.json.flags |= F_HOOK;
		1566
		1567	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1568
		1569	decode_ws (&dec);
883	sv = decode_sv (&dec);	1570	sv = decode_sv (&dec);
884		1571
		1572	if (offset_return)
		1573	*offset_return = dec.cur;
		1574
		1575	if (!(offset_return || !sv))
		1576	{
		1577	// check for trailing garbage
		1578	decode_ws (&dec);
		1579
		1580	if (*dec.cur)
		1581	{
		1582	dec.err = "garbage after JSON object";
		1583	SvREFCNT_dec (sv);
		1584	sv = 0;
		1585	}
		1586	}
		1587
885	if (!sv)	1588	if (!sv)
886	{	1589	{
887	IV offset = dec.flags & F_UTF8
888	? dec.cur - SvPVX (string)
889	: utf8_distance (dec.cur, SvPVX (string));
890	SV *uni = sv_newmortal ();	1590	SV *uni = sv_newmortal ();
891		1591
892	// horrible hack to silence warning inside pv_uni_display	1592	// horrible hack to silence warning inside pv_uni_display
893	COP cop = *PL_curcop;	1593	COP cop = *PL_curcop;
894	cop.cop_warnings = pWARN_NONE;	1594	cop.cop_warnings = pWARN_NONE;
…		…
896	SAVEVPTR (PL_curcop);	1596	SAVEVPTR (PL_curcop);
897	PL_curcop = &cop;	1597	PL_curcop = &cop;
898	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1598	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
899	LEAVE;	1599	LEAVE;
900		1600
901	croak ("%s, at character offset %d (%s)",	1601	croak ("%s, at character offset %d (before \"%s\")",
902	dec.err,	1602	dec.err,
903	(int)offset,	1603	ptr_to_index (string, dec.cur),
904	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1604	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
905	}	1605	}
906		1606
907	sv = sv_2mortal (sv);	1607	sv = sv_2mortal (sv);
908		1608
909	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1609	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
910	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1610	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
911		1611
912	return sv;	1612	return sv;
913	}	1613	}
914		1614
		1615	/////////////////////////////////////////////////////////////////////////////
		1616	// incremental parser
		1617
		1618	static void
		1619	incr_parse (JSON *self)
		1620	{
		1621	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1622
		1623	// the state machine here is a bit convoluted and could be simplified a lot
		1624	// but this would make it slower, so...
		1625
		1626	for (;;)
		1627	{
		1628	//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
		1629	switch (self->incr_mode)
		1630	{
		1631	// only used for initial whitespace skipping
		1632	case INCR_M_WS:
		1633	for (;;)
		1634	{
		1635	if (*p > 0x20)
		1636	{
		1637	if (*p == '#')
		1638	{
		1639	self->incr_mode = INCR_M_C0;
		1640	goto incr_m_c;
		1641	}
		1642	else
		1643	{
		1644	self->incr_mode = INCR_M_JSON;
		1645	goto incr_m_json;
		1646	}
		1647	}
		1648	else if (!*p)
		1649	goto interrupt;
		1650
		1651	++p;
		1652	}
		1653
		1654	// skip a single char inside a string (for \\-processing)
		1655	case INCR_M_BS:
		1656	if (!*p)
		1657	goto interrupt;
		1658
		1659	++p;
		1660	self->incr_mode = INCR_M_STR;
		1661	goto incr_m_str;
		1662
		1663	// inside #-style comments
		1664	case INCR_M_C0:
		1665	case INCR_M_C1:
		1666	incr_m_c:
		1667	for (;;)
		1668	{
		1669	if (*p == '\n')
		1670	{
		1671	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1672	break;
		1673	}
		1674	else if (!*p)
		1675	goto interrupt;
		1676
		1677	++p;
		1678	}
		1679
		1680	break;
		1681
		1682	// inside a string
		1683	case INCR_M_STR:
		1684	incr_m_str:
		1685	for (;;)
		1686	{
		1687	if (*p == '"')
		1688	{
		1689	++p;
		1690	self->incr_mode = INCR_M_JSON;
		1691
		1692	if (!self->incr_nest)
		1693	goto interrupt;
		1694
		1695	goto incr_m_json;
		1696	}
		1697	else if (*p == '\\')
		1698	{
		1699	++p; // "virtually" consumes character after \
		1700
		1701	if (!*p) // if at end of string we have to switch modes
		1702	{
		1703	self->incr_mode = INCR_M_BS;
		1704	goto interrupt;
		1705	}
		1706	}
		1707	else if (!*p)
		1708	goto interrupt;
		1709
		1710	++p;
		1711	}
		1712
		1713	// after initial ws, outside string
		1714	case INCR_M_JSON:
		1715	incr_m_json:
		1716	for (;;)
		1717	{
		1718	switch (*p++)
		1719	{
		1720	case 0:
		1721	--p;
		1722	goto interrupt;
		1723
		1724	case 0x09:
		1725	case 0x0a:
		1726	case 0x0d:
		1727	case 0x20:
		1728	if (!self->incr_nest)
		1729	{
		1730	--p; // do not eat the whitespace, let the next round do it
		1731	goto interrupt;
		1732	}
		1733	break;
		1734
		1735	case '"':
		1736	self->incr_mode = INCR_M_STR;
		1737	goto incr_m_str;
		1738
		1739	case '[':
		1740	case '{':
		1741	if (++self->incr_nest > self->max_depth)
		1742	croak (ERR_NESTING_EXCEEDED);
		1743	break;
		1744
		1745	case ']':
		1746	case '}':
		1747	if (--self->incr_nest <= 0)
		1748	goto interrupt;
		1749	break;
		1750
		1751	case '#':
		1752	self->incr_mode = INCR_M_C1;
		1753	goto incr_m_c;
		1754	}
		1755	}
		1756	}
		1757
		1758	modechange:
		1759	;
		1760	}
		1761
		1762	interrupt:
		1763	self->incr_pos = p - SvPVX (self->incr_text);
		1764	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1765	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
		1766	}
		1767
		1768	/////////////////////////////////////////////////////////////////////////////
		1769	// XS interface functions
		1770
915	MODULE = JSON::XS PACKAGE = JSON::XS	1771	MODULE = JSON::XS PACKAGE = JSON::XS
916		1772
917	BOOT:	1773	BOOT:
918	{	1774	{
919	int i;	1775	int i;
920		1776
921	memset (decode_hexdigit, 0xff, 256);
922	for (i = 10; i--; )	1777	for (i = 0; i < 256; ++i)
923	decode_hexdigit ['0' + i] = i;	1778	decode_hexdigit [i] =
		1779	i >= '0' && i <= '9' ? i - '0'
		1780	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1781	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1782	: -1;
924		1783
925	for (i = 7; i--; )
926	{
927	decode_hexdigit ['a' + i] = 10 + i;
928	decode_hexdigit ['A' + i] = 10 + i;
929	}
930
931	json_stash = gv_stashpv ("JSON::XS", 1);	1784	json_stash = gv_stashpv ("JSON::XS" , 1);
		1785	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1786
		1787	json_true = get_bool ("JSON::XS::true");
		1788	json_false = get_bool ("JSON::XS::false");
		1789
		1790	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
932	}	1791	}
933		1792
934	PROTOTYPES: DISABLE	1793	PROTOTYPES: DISABLE
935		1794
936	SV *new (char *dummy)	1795	void CLONE (...)
937	CODE:	1796	CODE:
938	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1797	json_stash = 0;
		1798	json_boolean_stash = 0;
		1799
		1800	void new (char *klass)
		1801	PPCODE:
		1802	{
		1803	SV *pv = NEWSV (0, sizeof (JSON));
		1804	SvPOK_only (pv);
		1805	json_init ((JSON *)SvPVX (pv));
		1806	XPUSHs (sv_2mortal (sv_bless (
		1807	newRV_noinc (pv),
		1808	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1809	)));
		1810	}
		1811
		1812	void ascii (JSON *self, int enable = 1)
		1813	ALIAS:
		1814	ascii = F_ASCII
		1815	latin1 = F_LATIN1
		1816	utf8 = F_UTF8
		1817	indent = F_INDENT
		1818	canonical = F_CANONICAL
		1819	space_before = F_SPACE_BEFORE
		1820	space_after = F_SPACE_AFTER
		1821	pretty = F_PRETTY
		1822	allow_nonref = F_ALLOW_NONREF
		1823	shrink = F_SHRINK
		1824	allow_blessed = F_ALLOW_BLESSED
		1825	convert_blessed = F_CONV_BLESSED
		1826	relaxed = F_RELAXED
		1827	allow_unknown = F_ALLOW_UNKNOWN
		1828	PPCODE:
		1829	{
		1830	if (enable)
		1831	self->flags |= ix;
		1832	else
		1833	self->flags &= ~ix;
		1834
		1835	XPUSHs (ST (0));
		1836	}
		1837
		1838	void get_ascii (JSON *self)
		1839	ALIAS:
		1840	get_ascii = F_ASCII
		1841	get_latin1 = F_LATIN1
		1842	get_utf8 = F_UTF8
		1843	get_indent = F_INDENT
		1844	get_canonical = F_CANONICAL
		1845	get_space_before = F_SPACE_BEFORE
		1846	get_space_after = F_SPACE_AFTER
		1847	get_allow_nonref = F_ALLOW_NONREF
		1848	get_shrink = F_SHRINK
		1849	get_allow_blessed = F_ALLOW_BLESSED
		1850	get_convert_blessed = F_CONV_BLESSED
		1851	get_relaxed = F_RELAXED
		1852	get_allow_unknown = F_ALLOW_UNKNOWN
		1853	PPCODE:
		1854	XPUSHs (boolSV (self->flags & ix));
		1855
		1856	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		1857	PPCODE:
		1858	self->max_depth = max_depth;
		1859	XPUSHs (ST (0));
		1860
		1861	U32 get_max_depth (JSON *self)
		1862	CODE:
		1863	RETVAL = self->max_depth;
939	OUTPUT:	1864	OUTPUT:
940	RETVAL	1865	RETVAL
941		1866
942	SV *ascii (SV *self, int enable = 1)	1867	void max_size (JSON *self, U32 max_size = 0)
943	ALIAS:	1868	PPCODE:
944	ascii = F_ASCII	1869	self->max_size = max_size;
945	utf8 = F_UTF8	1870	XPUSHs (ST (0));
946	indent = F_INDENT	1871
947	canonical = F_CANONICAL	1872	int get_max_size (JSON *self)
948	space_before = F_SPACE_BEFORE
949	space_after = F_SPACE_AFTER
950	pretty = F_PRETTY
951	allow_nonref = F_ALLOW_NONREF
952	shrink = F_SHRINK
953	CODE:	1873	CODE:
954	{	1874	RETVAL = self->max_size;
955	UV *uv = SvJSON (self);
956	if (enable)
957	*uv |= ix;
958	else
959	*uv &= ~ix;
960
961	RETVAL = newSVsv (self);
962	}
963	OUTPUT:	1875	OUTPUT:
964	RETVAL	1876	RETVAL
965		1877
966	void encode (SV *self, SV *scalar)	1878	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
967	PPCODE:	1879	PPCODE:
968	XPUSHs (encode_json (scalar, *SvJSON (self)));	1880	{
		1881	SvREFCNT_dec (self->cb_object);
		1882	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
969		1883
970	void decode (SV *self, SV *jsonstr)	1884	XPUSHs (ST (0));
		1885	}
		1886
		1887	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
971	PPCODE:	1888	PPCODE:
		1889	{
		1890	if (!self->cb_sk_object)
		1891	self->cb_sk_object = newHV ();
		1892
		1893	if (SvOK (cb))
		1894	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1895	else
		1896	{
		1897	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1898
		1899	if (!HvKEYS (self->cb_sk_object))
		1900	{
		1901	SvREFCNT_dec (self->cb_sk_object);
		1902	self->cb_sk_object = 0;
		1903	}
		1904	}
		1905
		1906	XPUSHs (ST (0));
		1907	}
		1908
		1909	void encode (JSON *self, SV *scalar)
		1910	PPCODE:
		1911	XPUSHs (encode_json (scalar, self));
		1912
		1913	void decode (JSON *self, SV *jsonstr)
		1914	PPCODE:
972	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1915	XPUSHs (decode_json (jsonstr, self, 0));
		1916
		1917	void decode_prefix (JSON *self, SV *jsonstr)
		1918	PPCODE:
		1919	{
		1920	char *offset;
		1921	EXTEND (SP, 2);
		1922	PUSHs (decode_json (jsonstr, self, &offset));
		1923	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		1924	}
		1925
		1926	void incr_parse (JSON *self, SV *jsonstr = 0)
		1927	PPCODE:
		1928	{
		1929	if (!self->incr_text)
		1930	self->incr_text = newSVpvn ("", 0);
		1931
		1932	// append data, if any
		1933	if (jsonstr)
		1934	{
		1935	if (SvUTF8 (jsonstr))
		1936	{
		1937	if (!SvUTF8 (self->incr_text))
		1938	{
		1939	/* utf-8-ness differs, need to upgrade */
		1940	sv_utf8_upgrade (self->incr_text);
		1941
		1942	if (self->incr_pos)
		1943	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1944	- (U8 *)SvPVX (self->incr_text);
		1945	}
		1946	}
		1947	else if (SvUTF8 (self->incr_text))
		1948	sv_utf8_upgrade (jsonstr);
		1949
		1950	{
		1951	STRLEN len;
		1952	const char *str = SvPV (jsonstr, len);
		1953	STRLEN cur = SvCUR (self->incr_text);
		1954
		1955	if (SvLEN (self->incr_text) <= cur + len)
		1956	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1957
		1958	Move (str, SvEND (self->incr_text), len, char);
		1959	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1960	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1961	}
		1962	}
		1963
		1964	if (GIMME_V != G_VOID)
		1965	do
		1966	{
		1967	char *offset;
		1968
		1969	if (!INCR_DONE (self))
		1970	{
		1971	incr_parse (self);
		1972
		1973	if (self->incr_pos > self->max_size && self->max_size)
		1974	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1975	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1976
		1977	if (!INCR_DONE (self))
		1978	break;
		1979	}
		1980
		1981	XPUSHs (decode_json (self->incr_text, self, &offset));
		1982
		1983	self->incr_pos -= offset - SvPVX (self->incr_text);
		1984	self->incr_nest = 0;
		1985	self->incr_mode = 0;
		1986
		1987	sv_chop (self->incr_text, offset);
		1988	}
		1989	while (GIMME_V == G_ARRAY);
		1990	}
		1991
		1992	SV *incr_text (JSON *self)
		1993	ATTRS: lvalue
		1994	CODE:
		1995	{
		1996	if (self->incr_pos)
		1997	croak ("incr_text can not be called when the incremental parser already started parsing");
		1998
		1999	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		2000	}
		2001	OUTPUT:
		2002	RETVAL
		2003
		2004	void incr_skip (JSON *self)
		2005	CODE:
		2006	{
		2007	if (self->incr_pos)
		2008	{
		2009	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2010	self->incr_pos = 0;
		2011	self->incr_nest = 0;
		2012	self->incr_mode = 0;
		2013	}
		2014	}
		2015
		2016	void incr_reset (JSON *self)
		2017	CODE:
		2018	{
		2019	SvREFCNT_dec (self->incr_text);
		2020	self->incr_text = 0;
		2021	self->incr_pos = 0;
		2022	self->incr_nest = 0;
		2023	self->incr_mode = 0;
		2024	}
		2025
		2026	void DESTROY (JSON *self)
		2027	CODE:
		2028	SvREFCNT_dec (self->cb_sk_object);
		2029	SvREFCNT_dec (self->cb_object);
		2030	SvREFCNT_dec (self->incr_text);
973		2031
974	PROTOTYPES: ENABLE	2032	PROTOTYPES: ENABLE
975		2033
976	void to_json (SV *scalar)	2034	void encode_json (SV *scalar)
		2035	ALIAS:
		2036	to_json_ = 0
		2037	encode_json = F_UTF8
977	PPCODE:	2038	PPCODE:
		2039	{
		2040	JSON json;
		2041	json_init (&json);
		2042	json.flags |= ix;
978	XPUSHs (encode_json (scalar, F_UTF8));	2043	XPUSHs (encode_json (scalar, &json));
		2044	}
979		2045
980	void from_json (SV *jsonstr)	2046	void decode_json (SV *jsonstr)
		2047	ALIAS:
		2048	from_json_ = 0
		2049	decode_json = F_UTF8
981	PPCODE:	2050	PPCODE:
		2051	{
		2052	JSON json;
		2053	json_init (&json);
		2054	json.flags |= ix;
982	XPUSHs (decode_json (jsonstr, F_UTF8));	2055	XPUSHs (decode_json (jsonstr, &json, 0));
		2056	}
983		2057

Diff Legend

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