ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.4 by root, Thu Mar 22 21:13:58 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>
		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_JSON_RPC 0x00000040
16	#define F_ALLOW_NONREF 0x00000080	32	#define F_ALLOW_NONREF 0x00000100UL
		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)
		198	{
		199	UV uaccum = 0;
		200	int eaccum = 0;
		201
		202	for (;;)
		203	{
		204	U8 dig = (U8)*s - '0';
		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	}
		276	/////////////////////////////////////////////////////////////////////////////
		277	// encoder
27		278
28	// structure used for encoding JSON	279	// structure used for encoding JSON
29	typedef struct	280	typedef struct
30	{	281	{
31	char *cur;	282	char *cur; // SvPVX (sv) + current output position
32	STRLEN len; // SvLEN (sv)
33	char *end; // SvEND (sv)	283	char *end; // SvEND (sv)
34	SV *sv;	284	SV *sv; // result scalar
35	UV flags;	285	JSON json;
36	int max_recurse;	286	U32 indent; // indentation level
37	int indent;	287	UV limit; // escape character values >= this value when encoding
38	} enc_t;	288	} enc_t;
39		289
40	// structure used for decoding JSON	290	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	/////////////////////////////////////////////////////////////////////////////
59
60	static void
61	need (enc_t *enc, STRLEN len)	291	need (enc_t *enc, STRLEN len)
62	{	292	{
63	if (enc->cur + len >= enc->end)	293	if (expect_false (enc->cur + len >= enc->end))
64	{	294	{
65	STRLEN cur = enc->cur - SvPVX (enc->sv);	295	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
66	SvGROW (enc->sv, cur + len + 1);	296	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
67	enc->cur = SvPVX (enc->sv) + cur;	297	enc->cur = SvPVX (enc->sv) + cur;
68	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	298	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
69	}	299	}
70	}	300	}
71		301
72	static void	302	INLINE void
73	encode_ch (enc_t *enc, char ch)	303	encode_ch (enc_t *enc, char ch)
74	{	304	{
75	need (enc, 1);	305	need (enc, 1);
76	*enc->cur++ = ch;	306	*enc->cur++ = ch;
77	}	307	}
…		…
85		315
86	while (str < end)	316	while (str < end)
87	{	317	{
88	unsigned char ch = *(unsigned char *)str;	318	unsigned char ch = *(unsigned char *)str;
89		319
90	if (ch == '"')	320	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
91	{	321	{
		322	if (expect_false (ch == '"')) // but with slow exceptions
		323	{
92	need (enc, len += 1);	324	need (enc, len += 1);
93	*enc->cur++ = '\\';	325	*enc->cur++ = '\\';
94	*enc->cur++ = '"';	326	*enc->cur++ = '"';
95	++str;
96	}	327	}
97	else if (ch == '\\')	328	else if (expect_false (ch == '\\'))
98	{	329	{
99	need (enc, len += 1);	330	need (enc, len += 1);
100	*enc->cur++ = '\\';	331	*enc->cur++ = '\\';
101	*enc->cur++ = '\\';	332	*enc->cur++ = '\\';
102	++str;
103	}	333	}
104	else if (ch >= 0x20 && ch < 0x80) // most common case	334	else
105	{
106	*enc->cur++ = ch;	335	*enc->cur++ = ch;
107	++str;	336
108	}
109	else if (ch == '\015')
110	{
111	need (enc, len += 1);
112	*enc->cur++ = '\\';
113	*enc->cur++ = 'r';
114	++str;
115	}
116	else if (ch == '\012')
117	{
118	need (enc, len += 1);
119	*enc->cur++ = '\\';
120	*enc->cur++ = 'n';
121	++str;	337	++str;
122	}	338	}
123	else	339	else
124	{	340	{
125	STRLEN clen;	341	switch (ch)
126	UV uch;
127
128	if (is_utf8)
129	{	342	{
130	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	343	case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
131	if (clen < 0)	344	case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
132	croak ("malformed UTF-8 character in string, cannot convert to JSON");	345	case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
133	}	346	case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
134	else	347	case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
135	{
136	uch = ch;
137	clen = 1;
138	}
139		348
140	if (uch < 0x80 || enc->flags & F_ASCII)	349	default:
141	{
142	if (uch > 0xFFFFUL)
143	{	350	{
		351	STRLEN clen;
		352	UV uch;
		353
		354	if (is_utf8)
		355	{
		356	uch = decode_utf8 (str, end - str, &clen);
		357	if (clen == (STRLEN)-1)
		358	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
		359	}
		360	else
		361	{
		362	uch = ch;
		363	clen = 1;
		364	}
		365
		366	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
		367	{
		368	if (uch >= 0x10000UL)
		369	{
		370	if (uch >= 0x110000UL)
		371	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		372
144	need (enc, len += 11);	373	need (enc, len += 11);
145	sprintf (enc->cur, "\\u%04x\\u%04x",	374	sprintf (enc->cur, "\\u%04x\\u%04x",
146	(uch - 0x10000) / 0x400 + 0xD800,	375	(int)((uch - 0x10000) / 0x400 + 0xD800),
147	(uch - 0x10000) % 0x400 + 0xDC00);	376	(int)((uch - 0x10000) % 0x400 + 0xDC00));
148	enc->cur += 12;	377	enc->cur += 12;
		378	}
		379	else
		380	{
		381	need (enc, len += 5);
		382	*enc->cur++ = '\\';
		383	*enc->cur++ = 'u';
		384	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
		385	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
		386	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
		387	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
		388	}
		389
		390	str += clen;
		391	}
		392	else if (enc->json.flags & F_LATIN1)
		393	{
		394	*enc->cur++ = uch;
		395	str += clen;
		396	}
		397	else if (is_utf8)
		398	{
		399	need (enc, len += clen);
		400	do
		401	{
		402	*enc->cur++ = *str++;
		403	}
		404	while (--clen);
		405	}
		406	else
		407	{
		408	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
		409	enc->cur = encode_utf8 (enc->cur, uch);
		410	++str;
		411	}
149	}	412	}
150	else
151	{
152	static char hexdigit [16] = "0123456789abcdef";
153	need (enc, len += 5);
154	*enc->cur++ = '\\';
155	*enc->cur++ = 'u';
156	*enc->cur++ = hexdigit [ uch >> 12 ];
157	*enc->cur++ = hexdigit [(uch >> 8) & 15];
158	*enc->cur++ = hexdigit [(uch >> 4) & 15];
159	*enc->cur++ = hexdigit [(uch >> 0) & 15];
160	}
161
162	str += clen;
163	}
164	else if (is_utf8)
165	{
166	need (enc, len += clen);
167	while (clen--)
168	*enc->cur++ = *str++;
169	}
170	else
171	{
172	need (enc, 10); // never more than 11 bytes needed
173	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
174	++str;
175	}	413	}
176	}	414	}
177		415
178	--len;	416	--len;
179	}	417	}
180	}	418	}
181		419
182	#define INDENT SB \	420	INLINE void
		421	encode_indent (enc_t *enc)
		422	{
183	if (enc->flags & F_INDENT) \	423	if (enc->json.flags & F_INDENT)
184	{ \	424	{
185	int i_; \	425	int spaces = enc->indent * INDENT_STEP;
186	need (enc, enc->indent); \	426
187	for (i_ = enc->indent * 3; i_--; )\	427	need (enc, spaces);
		428	memset (enc->cur, ' ', spaces);
		429	enc->cur += spaces;
		430	}
		431	}
		432
		433	INLINE void
		434	encode_space (enc_t *enc)
		435	{
		436	need (enc, 1);
		437	encode_ch (enc, ' ');
		438	}
		439
		440	INLINE void
		441	encode_nl (enc_t *enc)
		442	{
		443	if (enc->json.flags & F_INDENT)
		444	{
		445	need (enc, 1);
188	encode_ch (enc, ' '); \	446	encode_ch (enc, '\n');
189	} \	447	}
190	SE	448	}
191		449
192	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	450	INLINE void
193	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	451	encode_comma (enc_t *enc)
194	#define COMMA SB \	452	{
195	encode_ch (enc, ','); \	453	encode_ch (enc, ',');
		454
196	if (enc->flags & F_INDENT) \	455	if (enc->json.flags & F_INDENT)
197	NL; \	456	encode_nl (enc);
198	else if (enc->flags & F_SPACE_AFTER) \	457	else if (enc->json.flags & F_SPACE_AFTER)
199	SPACE; \	458	encode_space (enc);
200	SE	459	}
201		460
202	static void encode_sv (enc_t *enc, SV *sv);	461	static void encode_sv (enc_t *enc, SV *sv);
203		462
204	static void	463	static void
205	encode_av (enc_t *enc, AV *av)	464	encode_av (enc_t *enc, AV *av)
206	{	465	{
207	int i, len = av_len (av);	466	int i, len = av_len (av);
208		467
		468	if (enc->indent >= enc->json.max_depth)
		469	croak (ERR_NESTING_EXCEEDED);
		470
209	encode_ch (enc, '['); NL;	471	encode_ch (enc, '[');
210	++enc->indent;	472
		473	if (len >= 0)
		474	{
		475	encode_nl (enc); ++enc->indent;
211		476
212	for (i = 0; i <= len; ++i)	477	for (i = 0; i <= len; ++i)
213	{	478	{
214	INDENT;	479	SV **svp = av_fetch (av, i, 0);
215	encode_sv (enc, *av_fetch (av, i, 0));
216		480
		481	encode_indent (enc);
		482
		483	if (svp)
		484	encode_sv (enc, *svp);
		485	else
		486	encode_str (enc, "null", 4, 0);
		487
217	if (i < len)	488	if (i < len)
218	COMMA;	489	encode_comma (enc);
219	}	490	}
220		491
221	NL;	492	encode_nl (enc); --enc->indent; encode_indent (enc);
222		493	}
223	--enc->indent;	494
224	INDENT; encode_ch (enc, ']');	495	encode_ch (enc, ']');
225	}	496	}
226		497
227	static void	498	static void
228	encode_he (enc_t *enc, HE *he)	499	encode_hk (enc_t *enc, HE *he)
229	{	500	{
230	encode_ch (enc, '"');	501	encode_ch (enc, '"');
231		502
232	if (HeKLEN (he) == HEf_SVKEY)	503	if (HeKLEN (he) == HEf_SVKEY)
233	{	504	{
…		…
243	else	514	else
244	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	515	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
245		516
246	encode_ch (enc, '"');	517	encode_ch (enc, '"');
247		518
248	if (enc->flags & F_SPACE_BEFORE) SPACE;	519	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
249	encode_ch (enc, ':');	520	encode_ch (enc, ':');
250	if (enc->flags & F_SPACE_AFTER ) SPACE;	521	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
251	encode_sv (enc, HeVAL (he));
252	}	522	}
253		523
254	// compare hash entries, used when all keys are bytestrings	524	// compare hash entries, used when all keys are bytestrings
255	static int	525	static int
256	he_cmp_fast (const void *a_, const void *b_)	526	he_cmp_fast (const void *a_, const void *b_)
…		…
261	HE *b = *(HE **)b_;	531	HE *b = *(HE **)b_;
262		532
263	STRLEN la = HeKLEN (a);	533	STRLEN la = HeKLEN (a);
264	STRLEN lb = HeKLEN (b);	534	STRLEN lb = HeKLEN (b);
265		535
266	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	536	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
267	cmp = la < lb ? -1 : la == lb ? 0 : 1;	537	cmp = lb - la;
268		538
269	return cmp;	539	return cmp;
270	}	540	}
271		541
272	// 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
273	static int	543	static int
274	he_cmp_slow (const void *a, const void *b)	544	he_cmp_slow (const void *a, const void *b)
275	{	545	{
276	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	546	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
277	}	547	}
278		548
279	static void	549	static void
280	encode_hv (enc_t *enc, HV *hv)	550	encode_hv (enc_t *enc, HV *hv)
281	{	551	{
282	int count, i;	552	HE *he;
283		553
284	encode_ch (enc, '{'); NL; ++enc->indent;	554	if (enc->indent >= enc->json.max_depth)
		555	croak (ERR_NESTING_EXCEEDED);
285		556
286	if ((count = hv_iterinit (hv)))	557	encode_ch (enc, '{');
287	{	558
288	// for canonical output we have to sort by keys first	559	// for canonical output we have to sort by keys first
289	// actually, this is mostly due to the stupid so-called	560	// actually, this is mostly due to the stupid so-called
290	// security workaround added somewhere in 5.8.x.	561	// security workaround added somewhere in 5.8.x
291	// that randomises hash orderings	562	// that randomises hash orderings
292	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))
293	{	568	{
294	HE *he, *hes [count];	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	{
295	int fast = 1;	583	int i, fast = 1;
		584	#if defined(__BORLANDC__) || defined(_MSC_VER)
		585	HE **hes = _alloca (count * sizeof (HE));
		586	#else
		587	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		588	#endif
296		589
297	i = 0;	590	i = 0;
298	while ((he = hv_iternext (hv)))	591	while ((he = hv_iternext (hv)))
299	{	592	{
300	hes [i++] = he;	593	hes [i++] = he;
…		…
306		599
307	if (fast)	600	if (fast)
308	qsort (hes, count, sizeof (HE *), he_cmp_fast);	601	qsort (hes, count, sizeof (HE *), he_cmp_fast);
309	else	602	else
310	{	603	{
311	// hack to disable "use bytes"	604	// hack to forcefully disable "use bytes"
312	COP *oldcop = PL_curcop, cop;	605	COP cop = *PL_curcop;
313	cop.op_private = 0;	606	cop.op_private = 0;
		607
		608	ENTER;
		609	SAVETMPS;
		610
		611	SAVEVPTR (PL_curcop);
314	PL_curcop = &cop;	612	PL_curcop = &cop;
315		613
316	SAVETMPS;
317	qsort (hes, count, sizeof (HE *), he_cmp_slow);	614	qsort (hes, count, sizeof (HE *), he_cmp_slow);
		615
318	FREETMPS;	616	FREETMPS;
319		617	LEAVE;
320	PL_curcop = oldcop;
321	}	618	}
322		619
323	for (i = 0; i < count; ++i)	620	encode_nl (enc); ++enc->indent;
		621
		622	while (count--)
324	{	623	{
325	INDENT;	624	encode_indent (enc);
		625	he = hes [count];
326	encode_he (enc, hes [i]);	626	encode_hk (enc, he);
		627	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
327		628
328	if (i < count - 1)	629	if (count)
329	COMMA;	630	encode_comma (enc);
330	}	631	}
331		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)))
332	NL;	640	{
		641	encode_nl (enc); ++enc->indent;
		642
		643	for (;;)
		644	{
		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);
333	}	683	}
334	else	684	else
335	{	685	{
336	SV *sv;	686	#if 0
337	HE *he = hv_iternext (hv);	687	if (0 && sv_derived_from (rv, "JSON::Literal"))
338
339	for (;;)
340	{	688	{
341	INDENT;	689	// not yet
342	encode_he (enc, he);
343
344	if (!(he = hv_iternext (hv)))
345	break;
346
347	COMMA;
348	}	690	}
349		691	#endif
		692	if (enc->json.flags & F_CONV_BLESSED)
350	NL;	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);
		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
		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))));
351	}	731	}
352	}	732	}
		733	else if (svt == SVt_PVHV)
		734	encode_hv (enc, (HV *)sv);
		735	else if (svt == SVt_PVAV)
		736	encode_av (enc, (AV *)sv);
		737	else if (svt < SVt_PVAV)
		738	{
		739	STRLEN len = 0;
		740	char *pv = svt ? SvPV (sv, len) : 0;
353		741
354	--enc->indent; INDENT; encode_ch (enc, '}');	742	if (len == 1 && *pv == '1')
		743	encode_str (enc, "true", 4, 0);
		744	else if (len == 1 && *pv == '0')
		745	encode_str (enc, "false", 5, 0);
		746	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		747	encode_str (enc, "null", 4, 0);
		748	else
		749	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		750	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		751	}
		752	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		753	encode_str (enc, "null", 4, 0);
		754	else
		755	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		756	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
355	}	757	}
356		758
357	static void	759	static void
358	encode_sv (enc_t *enc, SV *sv)	760	encode_sv (enc_t *enc, SV *sv)
359	{	761	{
…		…
367	encode_str (enc, str, len, SvUTF8 (sv));	769	encode_str (enc, str, len, SvUTF8 (sv));
368	encode_ch (enc, '"');	770	encode_ch (enc, '"');
369	}	771	}
370	else if (SvNOKp (sv))	772	else if (SvNOKp (sv))
371	{	773	{
		774	// trust that perl will do the right thing w.r.t. JSON syntax.
372	need (enc, NV_DIG + 32);	775	need (enc, NV_DIG + 32);
373	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	776	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
374	enc->cur += strlen (enc->cur);	777	enc->cur += strlen (enc->cur);
375	}	778	}
376	else if (SvIOKp (sv))	779	else if (SvIOKp (sv))
377	{	780	{
		781	// we assume we can always read an IV as a UV and vice versa
		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)
		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
378	need (enc, 64);	794	need (enc, 6);
		795
		796	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		797	u = i < 0 ? -i : i;
		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'
		812	}
		813	else
		814	{
		815	// large integer, use the (rather slow) snprintf way.
		816	need (enc, IVUV_MAXCHARS);
379	enc->cur +=	817	enc->cur +=
380	SvIsUV(sv)	818	SvIsUV(sv)
381	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	819	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
382	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	820	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		821	}
383	}	822	}
384	else if (SvROK (sv))	823	else if (SvROK (sv))
385	{	824	encode_rv (enc, SvRV (sv));
386	if (!--enc->max_recurse)	825	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
387	croak ("data structure too deep (hit recursion limit)");
388
389	sv = SvRV (sv);
390
391	switch (SvTYPE (sv))
392	{
393	case SVt_PVAV: encode_av (enc, (AV *)sv); break;
394	case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
395
396	default:
397	croak ("JSON can only represent references to arrays or hashes");
398	}
399	}
400	else if (!SvOK (sv))
401	encode_str (enc, "null", 4, 0);	826	encode_str (enc, "null", 4, 0);
402	else	827	else
403	croak ("encountered perl type that JSON cannot handle");	828	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
		829	SvPV_nolen (sv), SvFLAGS (sv));
404	}	830	}
405		831
406	static SV *	832	static SV *
407	encode_json (SV *scalar, UV flags)	833	encode_json (SV *scalar, JSON *json)
408	{	834	{
409	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
410	croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");
411
412	enc_t enc;	835	enc_t enc;
413	enc.flags = flags;	836
		837	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
		838	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
		839
		840	enc.json = *json;
414	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	841	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
415	enc.cur = SvPVX (enc.sv);	842	enc.cur = SvPVX (enc.sv);
416	enc.end = SvEND (enc.sv);	843	enc.end = SvEND (enc.sv);
417	enc.max_recurse = 0;
418	enc.indent = 0;	844	enc.indent = 0;
		845	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		846	: enc.json.flags & F_LATIN1 ? 0x000100UL
		847	: 0x110000UL;
419		848
420	SvPOK_only (enc.sv);	849	SvPOK_only (enc.sv);
421	encode_sv (&enc, scalar);	850	encode_sv (&enc, scalar);
		851	encode_nl (&enc);
422		852
		853	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		854	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		855
423	if (!(flags & (F_ASCII | F_UTF8)))	856	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
424	SvUTF8_on (enc.sv);	857	SvUTF8_on (enc.sv);
425		858
426	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	859	if (enc.json.flags & F_SHRINK)
		860	shrink (enc.sv);
		861
427	return enc.sv;	862	return enc.sv;
428	}	863	}
429		864
430	/////////////////////////////////////////////////////////////////////////////	865	/////////////////////////////////////////////////////////////////////////////
		866	// decoder
431		867
432	#define WS \	868	// structure used for decoding JSON
		869	typedef struct
		870	{
		871	char *cur; // current parser pointer
		872	char *end; // end of input string
		873	const char *err; // parse error, if != 0
		874	JSON json;
		875	U32 depth; // recursion depth
		876	U32 maxdepth; // recursion depth limit
		877	} dec_t;
		878
		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
		889	decode_ws (dec_t *dec)
		890	{
433	for (;;) \	891	for (;;)
434	{ \	892	{
435	char ch = *dec->cur; \	893	char ch = *dec->cur;
		894
436	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	}
437	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	907	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
438	break; \	908	break; // parse error, but let higher level handle it, gives better error messages
		909
439	++dec->cur; \	910	++dec->cur;
440	}	911	}
		912	}
441		913
442	#define ERR(reason) SB dec->err = reason; goto fail; SE	914	#define ERR(reason) SB dec->err = reason; goto fail; SE
		915
443	#define EXPECT_CH(ch) SB \	916	#define EXPECT_CH(ch) SB \
444	if (*dec->cur != ch) \	917	if (*dec->cur != ch) \
445	ERR (# ch " expected"); \	918	ERR (# ch " expected"); \
446	++dec->cur; \	919	++dec->cur; \
447	SE	920	SE
448		921
		922	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
		923	#define DEC_DEC_DEPTH --dec->depth
		924
449	static SV *decode_sv (dec_t *dec);	925	static SV *decode_sv (dec_t *dec);
450		926
451	static signed char decode_hexdigit[256];	927	static signed char decode_hexdigit[256];
452		928
453	static UV	929	static UV
454	decode_4hex (dec_t *dec)	930	decode_4hex (dec_t *dec)
455	{	931	{
456	signed char d1, d2, d3, d4;	932	signed char d1, d2, d3, d4;
		933	unsigned char *cur = (unsigned char *)dec->cur;
457		934
458	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	935	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
459	if (d1 < 0) ERR ("four hexadecimal digits expected");	936	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
460	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	937	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
461	if (d2 < 0) ERR ("four hexadecimal digits expected");	938	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
462	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
463	if (d3 < 0) ERR ("four hexadecimal digits expected");
464	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
465	if (d4 < 0) ERR ("four hexadecimal digits expected");
466		939
467	dec->cur += 4;	940	dec->cur += 4;
468		941
469	return ((UV)d1) << 12	942	return ((UV)d1) << 12
470	| ((UV)d2) << 8	943	| ((UV)d2) << 8
…		…
473		946
474	fail:	947	fail:
475	return (UV)-1;	948	return (UV)-1;
476	}	949	}
477		950
478	#define APPEND_GROW(n) SB \
479	if (cur + (n) >= end) \
480	{ \
481	STRLEN ofs = cur - SvPVX (sv); \
482	SvGROW (sv, ofs + (n) + 1); \
483	cur = SvPVX (sv) + ofs; \
484	end = SvEND (sv); \
485	} \
486	SE
487
488	#define APPEND_CH(ch) SB \
489	APPEND_GROW (1); \
490	*cur++ = (ch); \
491	SE
492
493	static SV *	951	static SV *
494	decode_str (dec_t *dec)	952	decode_str (dec_t *dec)
495	{	953	{
496	SV *sv = NEWSV (0,2);	954	SV *sv = 0;
497	int utf8 = 0;	955	int utf8 = 0;
498	char *cur = SvPVX (sv);	956	char *dec_cur = dec->cur;
499	char *end = SvEND (sv);
500		957
501	for (;;)	958	do
502	{	959	{
503	unsigned char ch = *(unsigned char *)dec->cur;	960	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		961	char *cur = buf;
504		962
505	if (ch == '"')	963	do
506	break;
507	else if (ch == '\\')
508	{	964	{
509	switch (*++dec->cur)	965	unsigned char ch = *(unsigned char *)dec_cur++;
		966
		967	if (expect_false (ch == '"'))
510	{	968	{
511	case '\\':	969	--dec_cur;
512	case '/':	970	break;
513	case '"': APPEND_CH (*dec->cur++); break;	971	}
514		972	else if (expect_false (ch == '\\'))
515	case 'b': APPEND_CH ('\010'); ++dec->cur; break;	973	{
516	case 't': APPEND_CH ('\011'); ++dec->cur; break;	974	switch (*dec_cur)
517	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
518	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
519	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
520
521	case 'u':
522	{	975	{
523	UV lo, hi;	976	case '\\':
524	++dec->cur;	977	case '/':
		978	case '"': *cur++ = *dec_cur++; break;
525		979
526	hi = decode_4hex (dec);	980	case 'b': ++dec_cur; *cur++ = '\010'; break;
527	if (hi == (UV)-1)	981	case 't': ++dec_cur; *cur++ = '\011'; break;
528	goto fail;	982	case 'n': ++dec_cur; *cur++ = '\012'; break;
		983	case 'f': ++dec_cur; *cur++ = '\014'; break;
		984	case 'r': ++dec_cur; *cur++ = '\015'; break;
529		985
530	// possibly a surrogate pair	986	case 'u':
531	if (hi >= 0xd800 && hi < 0xdc00)
532	{	987	{
533	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	988	UV lo, hi;
534	ERR ("illegal surrogate character");	989	++dec_cur;
535		990
536	dec->cur += 2;	991	dec->cur = dec_cur;
537
538	lo = decode_4hex (dec);	992	hi = decode_4hex (dec);
		993	dec_cur = dec->cur;
539	if (lo == (UV)-1)	994	if (hi == (UV)-1)
540	goto fail;	995	goto fail;
541		996
		997	// possibly a surrogate pair
		998	if (hi >= 0xd800)
		999	if (hi < 0xdc00)
		1000	{
		1001	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		1002	ERR ("missing low surrogate character in surrogate pair");
		1003
		1004	dec_cur += 2;
		1005
		1006	dec->cur = dec_cur;
		1007	lo = decode_4hex (dec);
		1008	dec_cur = dec->cur;
		1009	if (lo == (UV)-1)
		1010	goto fail;
		1011
542	if (lo < 0xdc00 || lo >= 0xe000)	1012	if (lo < 0xdc00 || lo >= 0xe000)
543	ERR ("surrogate pair expected");	1013	ERR ("surrogate pair expected");
544		1014
545	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	1015	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		1016	}
		1017	else if (hi < 0xe000)
		1018	ERR ("missing high surrogate character in surrogate pair");
		1019
		1020	if (hi >= 0x80)
		1021	{
		1022	utf8 = 1;
		1023
		1024	cur = encode_utf8 (cur, hi);
		1025	}
		1026	else
		1027	*cur++ = hi;
546	}	1028	}
547	else if (lo >= 0xdc00 && lo < 0xe000)
548	ERR ("illegal surrogate character");
549
550	if (hi >= 0x80)
551	{	1029	break;
552	utf8 = 1;
553		1030
554	APPEND_GROW (4); // at most 4 bytes for 21 bits
555	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
556	}
557	else	1031	default:
558	APPEND_CH (hi);	1032	--dec_cur;
		1033	ERR ("illegal backslash escape sequence in string");
559	}	1034	}
		1035	}
		1036	else if (expect_true (ch >= 0x20 && ch < 0x80))
		1037	*cur++ = ch;
		1038	else if (ch >= 0x80)
		1039	{
		1040	STRLEN clen;
		1041
		1042	--dec_cur;
		1043
		1044	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		1045	if (clen == (STRLEN)-1)
		1046	ERR ("malformed UTF-8 character in JSON string");
		1047
		1048	do
		1049	*cur++ = *dec_cur++;
		1050	while (--clen);
		1051
		1052	utf8 = 1;
		1053	}
		1054	else
		1055	{
		1056	--dec_cur;
		1057
		1058	if (!ch)
		1059	ERR ("unexpected end of string while parsing JSON string");
560	break;	1060	else
		1061	ERR ("invalid character encountered while parsing JSON string");
561	}	1062	}
562	}	1063	}
563	else if (ch >= 0x20 && ch <= 0x7f)	1064	while (cur < buf + SHORT_STRING_LEN);
564	APPEND_CH (*dec->cur++);	1065
565	else if (ch >= 0x80)	1066	{
		1067	STRLEN len = cur - buf;
		1068
		1069	if (sv)
566	{	1070	{
567	STRLEN clen;	1071	STRLEN cur = SvCUR (sv);
568	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
569	if (clen < 0)
570	ERR ("malformed UTF-8 character in string, cannot convert to JSON");
571		1072
572	APPEND_GROW (clen);	1073	if (SvLEN (sv) <= cur + len)
573	memcpy (cur, dec->cur, clen);	1074	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
574	cur += clen;	1075
575	dec->cur += clen;	1076	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		1077	SvCUR_set (sv, SvCUR (sv) + len);
576	}	1078	}
577	else	1079	else
578	ERR ("invalid character encountered");	1080	sv = newSVpvn (buf, len);
579	}	1081	}
		1082	}
		1083	while (*dec_cur != '"');
580		1084
581	++dec->cur;	1085	++dec_cur;
582		1086
583	SvCUR_set (sv, cur - SvPVX (sv));	1087	if (sv)
584		1088	{
585	SvPOK_only (sv);	1089	SvPOK_only (sv);
586	*SvEND (sv) = 0;	1090	*SvEND (sv) = 0;
587		1091
588	if (utf8)	1092	if (utf8)
589	SvUTF8_on (sv);	1093	SvUTF8_on (sv);
		1094	}
		1095	else
		1096	sv = newSVpvn ("", 0);
590		1097
		1098	dec->cur = dec_cur;
591	return sv;	1099	return sv;
592		1100
593	fail:	1101	fail:
594	SvREFCNT_dec (sv);	1102	dec->cur = dec_cur;
595	return 0;	1103	return 0;
596	}	1104	}
597		1105
598	static SV *	1106	static SV *
599	decode_num (dec_t *dec)	1107	decode_num (dec_t *dec)
…		…
609	{	1117	{
610	++dec->cur;	1118	++dec->cur;
611	if (*dec->cur >= '0' && *dec->cur <= '9')	1119	if (*dec->cur >= '0' && *dec->cur <= '9')
612	ERR ("malformed number (leading zero must not be followed by another digit)");	1120	ERR ("malformed number (leading zero must not be followed by another digit)");
613	}	1121	}
614		1122	else if (*dec->cur < '0' || *dec->cur > '9')
615	// int	1123	ERR ("malformed number (no digits after initial minus)");
		1124	else
		1125	do
		1126	{
		1127	++dec->cur;
		1128	}
616	while (*dec->cur >= '0' && *dec->cur <= '9')	1129	while (*dec->cur >= '0' && *dec->cur <= '9');
617	++dec->cur;
618		1130
619	// [frac]	1131	// [frac]
620	if (*dec->cur == '.')	1132	if (*dec->cur == '.')
621	{	1133	{
622	is_nv = 1;	1134	++dec->cur;
		1135
		1136	if (*dec->cur < '0' || *dec->cur > '9')
		1137	ERR ("malformed number (no digits after decimal point)");
623		1138
624	do	1139	do
625	{	1140	{
626	++dec->cur;	1141	++dec->cur;
627	}	1142	}
628	while (*dec->cur >= '0' && *dec->cur <= '9');	1143	while (*dec->cur >= '0' && *dec->cur <= '9');
		1144
		1145	is_nv = 1;
629	}	1146	}
630		1147
631	// [exp]	1148	// [exp]
632	if (*dec->cur == 'e' || *dec->cur == 'E')	1149	if (*dec->cur == 'e' || *dec->cur == 'E')
633	{	1150	{
634	is_nv = 1;
635
636	++dec->cur;	1151	++dec->cur;
		1152
637	if (*dec->cur == '-' || *dec->cur == '+')	1153	if (*dec->cur == '-' || *dec->cur == '+')
638	++dec->cur;	1154	++dec->cur;
639		1155
		1156	if (*dec->cur < '0' || *dec->cur > '9')
		1157	ERR ("malformed number (no digits after exp sign)");
		1158
		1159	do
		1160	{
		1161	++dec->cur;
		1162	}
640	while (*dec->cur >= '0' && *dec->cur <= '9')	1163	while (*dec->cur >= '0' && *dec->cur <= '9');
641	++dec->cur;	1164
		1165	is_nv = 1;
642	}	1166	}
643		1167
644	if (!is_nv)	1168	if (!is_nv)
645	{	1169	{
646	UV uv;	1170	int len = dec->cur - start;
647	int numtype = grok_number (start, dec->cur - start, &uv);	1171
648	if (numtype & IS_NUMBER_IN_UV)	1172	// special case the rather common 1..5-digit-int case
649	if (numtype & IS_NUMBER_NEG)	1173	if (*start == '-')
		1174	switch (len)
650	{	1175	{
651	if (uv < (UV)IV_MIN)	1176	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
652	return newSViv (-(IV)uv);	1177	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
		1178	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		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));
653	}	1181	}
		1182	else
		1183	switch (len)
		1184	{
		1185	case 1: return newSViv ( start [0] - '0' * 1);
		1186	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1187	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		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);
		1190	}
		1191
		1192	{
		1193	UV uv;
		1194	int numtype = grok_number (start, len, &uv);
		1195	if (numtype & IS_NUMBER_IN_UV)
		1196	if (numtype & IS_NUMBER_NEG)
		1197	{
		1198	if (uv < (UV)IV_MIN)
		1199	return newSViv (-(IV)uv);
		1200	}
654	else	1201	else
655	return newSVuv (uv);	1202	return newSVuv (uv);
656	}	1203	}
657		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
658	return newSVnv (Atof (start));	1217	return newSVnv (json_atof (start));
659		1218
660	fail:	1219	fail:
661	return 0;	1220	return 0;
662	}	1221	}
663		1222
664	static SV *	1223	static SV *
665	decode_av (dec_t *dec)	1224	decode_av (dec_t *dec)
666	{	1225	{
667	AV *av = newAV ();	1226	AV *av = newAV ();
668		1227
		1228	DEC_INC_DEPTH;
		1229	decode_ws (dec);
		1230
		1231	if (*dec->cur == ']')
		1232	++dec->cur;
		1233	else
669	for (;;)	1234	for (;;)
670	{	1235	{
671	SV *value;	1236	SV *value;
672		1237
673	value = decode_sv (dec);	1238	value = decode_sv (dec);
674	if (!value)	1239	if (!value)
675	goto fail;	1240	goto fail;
676		1241
677	av_push (av, value);	1242	av_push (av, value);
678		1243
679	WS;	1244	decode_ws (dec);
680		1245
681	if (*dec->cur == ']')	1246	if (*dec->cur == ']')
682	{	1247	{
683	++dec->cur;	1248	++dec->cur;
684	break;	1249	break;
		1250	}
685	}	1251
686
687	if (*dec->cur != ',')	1252	if (*dec->cur != ',')
688	ERR (", or ] expected while parsing array");	1253	ERR (", or ] expected while parsing array");
689		1254
690	++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	}
691	}	1264	}
692		1265
		1266	DEC_DEC_DEPTH;
693	return newRV_noinc ((SV *)av);	1267	return newRV_noinc ((SV *)av);
694		1268
695	fail:	1269	fail:
696	SvREFCNT_dec (av);	1270	SvREFCNT_dec (av);
		1271	DEC_DEC_DEPTH;
697	return 0;	1272	return 0;
698	}	1273	}
699		1274
700	static SV *	1275	static SV *
701	decode_hv (dec_t *dec)	1276	decode_hv (dec_t *dec)
702	{	1277	{
		1278	SV *sv;
703	HV *hv = newHV ();	1279	HV *hv = newHV ();
704		1280
		1281	DEC_INC_DEPTH;
		1282	decode_ws (dec);
		1283
		1284	if (*dec->cur == '}')
		1285	++dec->cur;
		1286	else
705	for (;;)	1287	for (;;)
706	{	1288	{
707	SV *key, *value;
708
709	WS; EXPECT_CH ('"');	1289	EXPECT_CH ('"');
710		1290
711	key = decode_str (dec);	1291	// heuristic: assume that
712	if (!key)	1292	// a) decode_str + hv_store_ent are abysmally slow.
713	goto fail;	1293	// b) most hash keys are short, simple ascii text.
714		1294	// => try to "fast-match" such strings to avoid
715	WS; EXPECT_CH (':');	1295	// the overhead of decode_str + hv_store_ent.
716
717	value = decode_sv (dec);
718	if (!value)
719	{	1296	{
		1297	SV *value;
		1298	char *p = dec->cur;
		1299	char *e = p + 24; // only try up to 24 bytes
		1300
		1301	for (;;)
		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);
720	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)
721	goto fail;	1338	goto fail;
		1339
		1340	hv_store (hv, key, len, value, 0);
		1341
		1342	break;
		1343	}
		1344
		1345	++p;
		1346	}
722	}	1347	}
723		1348
724	//TODO: optimise	1349	decode_ws (dec);
725	hv_store_ent (hv, key, value, 0);
726		1350
727	WS;
728
729	if (*dec->cur == '}')	1351	if (*dec->cur == '}')
		1352	{
		1353	++dec->cur;
		1354	break;
		1355	}
		1356
		1357	if (*dec->cur != ',')
		1358	ERR (", or } expected while parsing object/hash");
		1359
		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	}
		1369	}
		1370
		1371	DEC_DEC_DEPTH;
		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)
730	{	1378	{
731	++dec->cur;	1379	HE *cb, *he;
732	break;	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	}
733	}	1408	}
734		1409
735	if (*dec->cur != ',')	1410	if (dec->json.cb_object)
736	ERR (", or } expected while parsing object/hash");	1411	{
		1412	dSP;
		1413	int count;
737		1414
738	++dec->cur;	1415	ENTER; SAVETMPS; PUSHMARK (SP);
739	}	1416	XPUSHs (sv_2mortal (sv));
740		1417
741	return newRV_noinc ((SV *)hv);	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;
742		1433
743	fail:	1434	fail:
744	SvREFCNT_dec (hv);	1435	SvREFCNT_dec (hv);
		1436	DEC_DEC_DEPTH;
745	return 0;	1437	return 0;
746	}	1438	}
747		1439
748	static SV *	1440	static SV *
749	decode_sv (dec_t *dec)	1441	decode_sv (dec_t *dec)
750	{	1442	{
751	WS;	1443	// the beauty of JSON: you need exactly one character lookahead
		1444	// to parse everything.
752	switch (*dec->cur)	1445	switch (*dec->cur)
753	{	1446	{
754	case '"': ++dec->cur; return decode_str (dec);	1447	case '"': ++dec->cur; return decode_str (dec);
755	case '[': ++dec->cur; return decode_av (dec);	1448	case '[': ++dec->cur; return decode_av (dec);
756	case '{': ++dec->cur; return decode_hv (dec);	1449	case '{': ++dec->cur; return decode_hv (dec);
757		1450
758	case '-':	1451	case '-':
759	case '0': case '1': case '2': case '3': case '4':	1452	case '0': case '1': case '2': case '3': case '4':
760	case '5': case '6': case '7': case '8': case '9':	1453	case '5': case '6': case '7': case '8': case '9':
761	return decode_num (dec);	1454	return decode_num (dec);
762		1455
763	case 't':	1456	case 't':
764	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1457	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
765	{	1458	{
766	dec->cur += 4;	1459	dec->cur += 4;
		1460	#if JSON_SLOW
		1461	json_true = get_bool ("JSON::XS::true");
		1462	#endif
767	return newSViv (1);	1463	return newSVsv (json_true);
768	}	1464	}
769	else	1465	else
770	ERR ("'true' expected");	1466	ERR ("'true' expected");
771		1467
772	break;	1468	break;
773		1469
774	case 'f':	1470	case 'f':
775	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1471	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
776	{	1472	{
777	dec->cur += 5;	1473	dec->cur += 5;
		1474	#if JSON_SLOW
		1475	json_false = get_bool ("JSON::XS::false");
		1476	#endif
778	return newSViv (0);	1477	return newSVsv (json_false);
779	}	1478	}
780	else	1479	else
781	ERR ("'false' expected");	1480	ERR ("'false' expected");
782		1481
783	break;	1482	break;
784		1483
785	case 'n':	1484	case 'n':
786	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))	1485	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
787	{	1486	{
788	dec->cur += 4;	1487	dec->cur += 4;
789	return newSViv (1);	1488	return newSVsv (&PL_sv_undef);
790	}	1489	}
791	else	1490	else
792	ERR ("'null' expected");	1491	ERR ("'null' expected");
793		1492
794	break;	1493	break;
795		1494
796	default:	1495	default:
797	ERR ("malformed json string");	1496	ERR ("malformed JSON string, neither array, object, number, string or atom");
798	break;	1497	break;
799	}	1498	}
800		1499
801	fail:	1500	fail:
802	return 0;	1501	return 0;
803	}	1502	}
804		1503
805	static SV *	1504	static SV *
806	decode_json (SV *string, UV flags)	1505	decode_json (SV *string, JSON *json, char **offset_return)
807	{	1506	{
		1507	dec_t dec;
808	SV *sv;	1508	SV *sv;
809		1509
		1510	/* work around bugs in 5.10 where manipulating magic values
		1511	* will perl ignore the magic in subsequent accesses
		1512	*/
		1513	/*SvGETMAGIC (string);*/
		1514	if (SvMAGICAL (string))
		1515	string = sv_2mortal (newSVsv (string));
		1516
		1517	SvUPGRADE (string, SVt_PV);
		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
810	if (!(flags & F_UTF8))	1541	if (json->flags & F_UTF8)
		1542	sv_utf8_downgrade (string, 0);
		1543	else
811	sv_utf8_upgrade (string);	1544	sv_utf8_upgrade (string);
812		1545
813	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1546	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
814		1547
815	dec_t dec;	1548	dec.json = *json;
816	dec.flags = flags;
817	dec.cur = SvPVX (string);	1549	dec.cur = SvPVX (string);
818	dec.end = SvEND (string);	1550	dec.end = SvEND (string);
819	dec.err = 0;	1551	dec.err = 0;
		1552	dec.depth = 0;
820		1553
821	*dec.end = 1; // invalid anywhere	1554	if (dec.json.cb_object || dec.json.cb_sk_object)
		1555	dec.json.flags |= F_HOOK;
		1556
		1557	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1558
		1559	decode_ws (&dec);
822	sv = decode_sv (&dec);	1560	sv = decode_sv (&dec);
823	*dec.end = 0;	1561
		1562	if (offset_return)
		1563	*offset_return = dec.cur;
		1564
		1565	if (!(offset_return || !sv))
		1566	{
		1567	// check for trailing garbage
		1568	decode_ws (&dec);
		1569
		1570	if (*dec.cur)
		1571	{
		1572	dec.err = "garbage after JSON object";
		1573	SvREFCNT_dec (sv);
		1574	sv = 0;
		1575	}
		1576	}
824		1577
825	if (!sv)	1578	if (!sv)
826	{	1579	{
827	IV offset = utf8_distance (dec.cur, SvPVX (string));
828	SV *uni = sv_newmortal ();	1580	SV *uni = sv_newmortal ();
829		1581
		1582	// horrible hack to silence warning inside pv_uni_display
		1583	COP cop = *PL_curcop;
		1584	cop.cop_warnings = pWARN_NONE;
		1585	ENTER;
		1586	SAVEVPTR (PL_curcop);
		1587	PL_curcop = &cop;
830	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);
		1589	LEAVE;
		1590
831	croak ("%s, at character %d (%s)",	1591	croak ("%s, at character offset %d (before \"%s\")",
832	dec.err,	1592	dec.err,
833	(int)offset,	1593	ptr_to_index (string, dec.cur),
834	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1594	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
835	}	1595	}
836		1596
837	sv = sv_2mortal (sv);	1597	sv = sv_2mortal (sv);
838		1598
839	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1599	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
840	croak ("JSON object or array expected (but number, string, true, false or null found, 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)");
841		1601
842	return sv;	1602	return sv;
843	}	1603	}
844		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
		1756	}
		1757
		1758	/////////////////////////////////////////////////////////////////////////////
		1759	// XS interface functions
		1760
845	MODULE = JSON::XS PACKAGE = JSON::XS	1761	MODULE = JSON::XS PACKAGE = JSON::XS
846		1762
847	BOOT:	1763	BOOT:
848	{	1764	{
849	int i;	1765	int i;
850		1766
851	memset (decode_hexdigit, 0xff, 256);
852	for (i = 10; i--; )	1767	for (i = 0; i < 256; ++i)
853	decode_hexdigit ['0' + i] = i;	1768	decode_hexdigit [i] =
		1769	i >= '0' && i <= '9' ? i - '0'
		1770	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1771	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1772	: -1;
854		1773
855	for (i = 7; i--; )
856	{
857	decode_hexdigit ['a' + i] = 10 + i;
858	decode_hexdigit ['A' + i] = 10 + i;
859	}
860
861	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 */
862	}	1781	}
863		1782
864	PROTOTYPES: DISABLE	1783	PROTOTYPES: DISABLE
865		1784
866	SV *new (char *dummy)	1785	void CLONE (...)
867	CODE:	1786	CODE:
868	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;
869	OUTPUT:	1854	OUTPUT:
870	RETVAL	1855	RETVAL
871		1856
872	SV *ascii (SV *self, int enable)	1857	void max_size (JSON *self, U32 max_size = 0)
873	ALIAS:	1858	PPCODE:
874	ascii = F_ASCII	1859	self->max_size = max_size;
875	utf8 = F_UTF8	1860	XPUSHs (ST (0));
876	indent = F_INDENT	1861
877	canonical = F_CANONICAL	1862	int get_max_size (JSON *self)
878	space_before = F_SPACE_BEFORE
879	space_after = F_SPACE_AFTER
880	json_rpc = F_JSON_RPC
881	pretty = F_PRETTY
882	allow_nonref = F_ALLOW_NONREF
883	CODE:	1863	CODE:
884	{	1864	RETVAL = self->max_size;
885	UV *uv = SvJSON (self);
886	if (enable)
887	*uv |= ix;
888	else
889	*uv &= ~ix;
890
891	RETVAL = newSVsv (self);
892	}
893	OUTPUT:	1865	OUTPUT:
894	RETVAL	1866	RETVAL
895		1867
896	void encode (SV *self, SV *scalar)	1868	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
897	PPCODE:	1869	PPCODE:
898	XPUSHs (encode_json (scalar, *SvJSON (self)));	1870	{
		1871	SvREFCNT_dec (self->cb_object);
		1872	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
899		1873
900	void decode (SV *self, SV *jsonstr)	1874	XPUSHs (ST (0));
		1875	}
		1876
		1877	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
901	PPCODE:	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:
902	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	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
		1984	CODE:
		1985	{
		1986	if (self->incr_pos)
		1987	croak ("incr_text can not be called when the incremental parser already started parsing");
		1988
		1989	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		1990	}
		1991	OUTPUT:
		1992	RETVAL
		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);
903		2021
904	PROTOTYPES: ENABLE	2022	PROTOTYPES: ENABLE
905		2023
906	void to_json (SV *scalar)	2024	void encode_json (SV *scalar)
		2025	ALIAS:
		2026	to_json_ = 0
		2027	encode_json = F_UTF8
907	PPCODE:	2028	PPCODE:
		2029	{
		2030	JSON json;
		2031	json_init (&json);
		2032	json.flags |= ix;
908	XPUSHs (encode_json (scalar, F_UTF8));	2033	XPUSHs (encode_json (scalar, &json));
		2034	}
909		2035
910	void from_json (SV *jsonstr)	2036	void decode_json (SV *jsonstr)
		2037	ALIAS:
		2038	from_json_ = 0
		2039	decode_json = F_UTF8
911	PPCODE:	2040	PPCODE:
		2041	{
		2042	JSON json;
		2043	json_init (&json);
		2044	json.flags |= ix;
912	XPUSHs (decode_json (jsonstr, F_UTF8));	2045	XPUSHs (decode_json (jsonstr, &json, 0));
		2046	}
913		2047

Diff Legend

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