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 1.115 by root, Fri Dec 21 05:16:19 2012 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
		46
		47	#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
22		48
23	#define SB do {	49	#define SB do {
24	#define SE } while (0)	50	#define SE } while (0)
25		51
26	static HV *json_stash;	52	#if __GNUC__ >= 3
		53	# define expect(expr,value) __builtin_expect ((expr), (value))
		54	# define INLINE static inline
		55	#else
		56	# define expect(expr,value) (expr)
		57	# define INLINE static
		58	#endif
		59
		60	#define expect_false(expr) expect ((expr) != 0, 0)
		61	#define expect_true(expr) expect ((expr) != 0, 1)
		62
		63	#define IN_RANGE_INC(type,val,beg,end) \
		64	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		65	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		66
		67	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
		68
		69	#ifdef USE_ITHREADS
		70	# define JSON_SLOW 1
		71	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		72	#else
		73	# define JSON_SLOW 0
		74	# define JSON_STASH json_stash
		75	#endif
		76
		77	// the amount of HEs to allocate on the stack, when sorting keys
		78	#define STACK_HES 64
		79
		80	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		81	static SV *json_true, *json_false;
		82
		83	enum {
		84	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		85	INCR_M_STR, // inside string
		86	INCR_M_BS, // inside backslash
		87	INCR_M_C0, // inside comment in initial whitespace sequence
		88	INCR_M_C1, // inside comment in other places
		89	INCR_M_JSON // outside anything, count nesting
		90	};
		91
		92	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		93
		94	typedef struct {
		95	U32 flags;
		96	U32 max_depth;
		97	STRLEN max_size;
		98
		99	SV *cb_object;
		100	HV *cb_sk_object;
		101
		102	// for the incremental parser
		103	SV *incr_text; // the source text so far
		104	STRLEN incr_pos; // the current offset into the text
		105	int incr_nest; // {[]}-nesting level
		106	unsigned char incr_mode;
		107	} JSON;
		108
		109	INLINE void
		110	json_init (JSON *json)
		111	{
		112	Zero (json, 1, JSON);
		113	json->max_depth = 512;
		114	}
		115
		116	/////////////////////////////////////////////////////////////////////////////
		117	// utility functions
		118
		119	INLINE SV *
		120	get_bool (const char *name)
		121	{
		122	SV *sv = get_sv (name, 1);
		123
		124	SvREADONLY_on (sv);
		125	SvREADONLY_on (SvRV (sv));
		126
		127	return sv;
		128	}
		129
		130	INLINE void
		131	shrink (SV *sv)
		132	{
		133	sv_utf8_downgrade (sv, 1);
		134
		135	if (SvLEN (sv) > SvCUR (sv) + 1)
		136	{
		137	#ifdef SvPV_shrink_to_cur
		138	SvPV_shrink_to_cur (sv);
		139	#elif defined (SvPV_renew)
		140	SvPV_renew (sv, SvCUR (sv) + 1);
		141	#endif
		142	}
		143	}
		144
		145	// decode an utf-8 character and return it, or (UV)-1 in
		146	// case of an error.
		147	// we special-case "safe" characters from U+80 .. U+7FF,
		148	// but use the very good perl function to parse anything else.
		149	// note that we never call this function for a ascii codepoints
		150	INLINE UV
		151	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		152	{
		153	if (expect_true (len >= 2
		154	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
		155	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
		156	{
		157	*clen = 2;
		158	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		159	}
		160	else
		161	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		162	}
		163
		164	// likewise for encoding, also never called for ascii codepoints
		165	// this function takes advantage of this fact, although current gccs
		166	// seem to optimise the check for >= 0x80 away anyways
		167	INLINE unsigned char *
		168	encode_utf8 (unsigned char *s, UV ch)
		169	{
		170	if (expect_false (ch < 0x000080))
		171	*s++ = ch;
		172	else if (expect_true (ch < 0x000800))
		173	*s++ = 0xc0 | ( ch >> 6),
		174	*s++ = 0x80 | ( ch & 0x3f);
		175	else if ( ch < 0x010000)
		176	*s++ = 0xe0 | ( ch >> 12),
		177	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		178	*s++ = 0x80 | ( ch & 0x3f);
		179	else if ( ch < 0x110000)
		180	*s++ = 0xf0 | ( ch >> 18),
		181	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		182	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		183	*s++ = 0x80 | ( ch & 0x3f);
		184
		185	return s;
		186	}
		187
		188	// convert offset pointer to character index, sv must be string
		189	static STRLEN
		190	ptr_to_index (SV *sv, char *offset)
		191	{
		192	return SvUTF8 (sv)
		193	? utf8_distance (offset, SvPVX (sv))
		194	: offset - SvPVX (sv);
		195	}
		196
		197	/////////////////////////////////////////////////////////////////////////////
		198	// fp hell
		199
		200	// scan a group of digits, and a trailing exponent
		201	static void
		202	json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
		203	{
		204	UV uaccum = 0;
		205	int eaccum = 0;
		206
		207	// if we recurse too deep, skip all remaining digits
		208	// to avoid a stack overflow attack
		209	if (expect_false (--maxdepth <= 0))
		210	while (((U8)*s - '0') < 10)
		211	++s;
		212
		213	for (;;)
		214	{
		215	U8 dig = (U8)*s - '0';
		216
		217	if (expect_false (dig >= 10))
		218	{
		219	if (dig == (U8)((U8)'.' - (U8)'0'))
		220	{
		221	++s;
		222	json_atof_scan1 (s, accum, expo, 1, maxdepth);
		223	}
		224	else if ((dig | ' ') == 'e' - '0')
		225	{
		226	int exp2 = 0;
		227	int neg = 0;
		228
		229	++s;
		230
		231	if (*s == '-')
		232	{
		233	++s;
		234	neg = 1;
		235	}
		236	else if (*s == '+')
		237	++s;
		238
		239	while ((dig = (U8)*s - '0') < 10)
		240	exp2 = exp2 * 10 + *s++ - '0';
		241
		242	*expo += neg ? -exp2 : exp2;
		243	}
		244
		245	break;
		246	}
		247
		248	++s;
		249
		250	uaccum = uaccum * 10 + dig;
		251	++eaccum;
		252
		253	// if we have too many digits, then recurse for more
		254	// we actually do this for rather few digits
		255	if (uaccum >= (UV_MAX - 9) / 10)
		256	{
		257	if (postdp) *expo -= eaccum;
		258	json_atof_scan1 (s, accum, expo, postdp, maxdepth);
		259	if (postdp) *expo += eaccum;
		260
		261	break;
		262	}
		263	}
		264
		265	// this relies greatly on the quality of the pow ()
		266	// implementation of the platform, but a good
		267	// implementation is hard to beat.
		268	// (IEEE 754 conformant ones are required to be exact)
		269	if (postdp) *expo -= eaccum;
		270	*accum += uaccum * Perl_pow (10., *expo);
		271	*expo += eaccum;
		272	}
		273
		274	static NV
		275	json_atof (const char *s)
		276	{
		277	NV accum = 0.;
		278	int expo = 0;
		279	int neg = 0;
		280
		281	if (*s == '-')
		282	{
		283	++s;
		284	neg = 1;
		285	}
		286
		287	// a recursion depth of ten gives us >>500 bits
		288	json_atof_scan1 (s, &accum, &expo, 0, 10);
		289
		290	return neg ? -accum : accum;
		291	}
		292	/////////////////////////////////////////////////////////////////////////////
		293	// encoder
27		294
28	// structure used for encoding JSON	295	// structure used for encoding JSON
29	typedef struct	296	typedef struct
30	{	297	{
31	char *cur;	298	char *cur; // SvPVX (sv) + current output position
32	STRLEN len; // SvLEN (sv)
33	char *end; // SvEND (sv)	299	char *end; // SvEND (sv)
34	SV *sv;	300	SV *sv; // result scalar
35	UV flags;	301	JSON json;
36	int max_recurse;	302	U32 indent; // indentation level
37	int indent;	303	UV limit; // escape character values >= this value when encoding
38	} enc_t;	304	} enc_t;
39		305
40	// structure used for decoding JSON	306	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)	307	need (enc_t *enc, STRLEN len)
62	{	308	{
63	if (enc->cur + len >= enc->end)	309	if (expect_false (enc->cur + len >= enc->end))
64	{	310	{
65	STRLEN cur = enc->cur - SvPVX (enc->sv);	311	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
66	SvGROW (enc->sv, cur + len + 1);	312	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
67	enc->cur = SvPVX (enc->sv) + cur;	313	enc->cur = SvPVX (enc->sv) + cur;
68	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	314	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
69	}	315	}
70	}	316	}
71		317
72	static void	318	INLINE void
73	encode_ch (enc_t *enc, char ch)	319	encode_ch (enc_t *enc, char ch)
74	{	320	{
75	need (enc, 1);	321	need (enc, 1);
76	*enc->cur++ = ch;	322	*enc->cur++ = ch;
77	}	323	}
…		…
85		331
86	while (str < end)	332	while (str < end)
87	{	333	{
88	unsigned char ch = *(unsigned char *)str;	334	unsigned char ch = *(unsigned char *)str;
89		335
90	if (ch == '"')	336	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
91	{	337	{
		338	if (expect_false (ch == '"')) // but with slow exceptions
		339	{
92	need (enc, len += 1);	340	need (enc, len += 1);
93	*enc->cur++ = '\\';	341	*enc->cur++ = '\\';
94	*enc->cur++ = '"';	342	*enc->cur++ = '"';
95	++str;
96	}	343	}
97	else if (ch == '\\')	344	else if (expect_false (ch == '\\'))
98	{	345	{
99	need (enc, len += 1);	346	need (enc, len += 1);
100	*enc->cur++ = '\\';	347	*enc->cur++ = '\\';
101	*enc->cur++ = '\\';	348	*enc->cur++ = '\\';
102	++str;
103	}	349	}
104	else if (ch >= 0x20 && ch < 0x80) // most common case	350	else
105	{
106	*enc->cur++ = ch;	351	*enc->cur++ = ch;
107	++str;	352
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;	353	++str;
122	}	354	}
123	else	355	else
124	{	356	{
125	STRLEN clen;	357	switch (ch)
126	UV uch;
127
128	if (is_utf8)
129	{
130	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
131	if (clen < 0)
132	croak ("malformed UTF-8 character in string, cannot convert to JSON");
133	}	358	{
134	else	359	case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
135	{	360	case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
136	uch = ch;	361	case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
137	clen = 1;	362	case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
138	}	363	case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
139		364
140	if (uch < 0x80 || enc->flags & F_ASCII)	365	default:
141	{
142	if (uch > 0xFFFFUL)
143	{	366	{
		367	STRLEN clen;
		368	UV uch;
		369
		370	if (is_utf8)
		371	{
		372	uch = decode_utf8 (str, end - str, &clen);
		373	if (clen == (STRLEN)-1)
		374	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
		375	}
		376	else
		377	{
		378	uch = ch;
		379	clen = 1;
		380	}
		381
		382	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
		383	{
		384	if (uch >= 0x10000UL)
		385	{
		386	if (uch >= 0x110000UL)
		387	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		388
144	need (enc, len += 11);	389	need (enc, len += 11);
145	sprintf (enc->cur, "\\u%04x\\u%04x",	390	sprintf (enc->cur, "\\u%04x\\u%04x",
146	(uch - 0x10000) / 0x400 + 0xD800,	391	(int)((uch - 0x10000) / 0x400 + 0xD800),
147	(uch - 0x10000) % 0x400 + 0xDC00);	392	(int)((uch - 0x10000) % 0x400 + 0xDC00));
148	enc->cur += 12;	393	enc->cur += 12;
		394	}
		395	else
		396	{
		397	need (enc, len += 5);
		398	*enc->cur++ = '\\';
		399	*enc->cur++ = 'u';
		400	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
		401	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
		402	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
		403	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
		404	}
		405
		406	str += clen;
		407	}
		408	else if (enc->json.flags & F_LATIN1)
		409	{
		410	*enc->cur++ = uch;
		411	str += clen;
		412	}
		413	else if (is_utf8)
		414	{
		415	need (enc, len += clen);
		416	do
		417	{
		418	*enc->cur++ = *str++;
		419	}
		420	while (--clen);
		421	}
		422	else
		423	{
		424	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
		425	enc->cur = encode_utf8 (enc->cur, uch);
		426	++str;
		427	}
149	}	428	}
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	}	429	}
176	}	430	}
177		431
178	--len;	432	--len;
179	}	433	}
180	}	434	}
181		435
182	#define INDENT SB \	436	INLINE void
		437	encode_indent (enc_t *enc)
		438	{
183	if (enc->flags & F_INDENT) \	439	if (enc->json.flags & F_INDENT)
184	{ \	440	{
185	int i_; \	441	int spaces = enc->indent * INDENT_STEP;
186	need (enc, enc->indent); \	442
187	for (i_ = enc->indent * 3; i_--; )\	443	need (enc, spaces);
		444	memset (enc->cur, ' ', spaces);
		445	enc->cur += spaces;
		446	}
		447	}
		448
		449	INLINE void
		450	encode_space (enc_t *enc)
		451	{
		452	need (enc, 1);
		453	encode_ch (enc, ' ');
		454	}
		455
		456	INLINE void
		457	encode_nl (enc_t *enc)
		458	{
		459	if (enc->json.flags & F_INDENT)
		460	{
		461	need (enc, 1);
188	encode_ch (enc, ' '); \	462	encode_ch (enc, '\n');
189	} \	463	}
190	SE	464	}
191		465
192	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	466	INLINE void
193	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	467	encode_comma (enc_t *enc)
194	#define COMMA SB \	468	{
195	encode_ch (enc, ','); \	469	encode_ch (enc, ',');
		470
196	if (enc->flags & F_INDENT) \	471	if (enc->json.flags & F_INDENT)
197	NL; \	472	encode_nl (enc);
198	else if (enc->flags & F_SPACE_AFTER) \	473	else if (enc->json.flags & F_SPACE_AFTER)
199	SPACE; \	474	encode_space (enc);
200	SE	475	}
201		476
202	static void encode_sv (enc_t *enc, SV *sv);	477	static void encode_sv (enc_t *enc, SV *sv);
203		478
204	static void	479	static void
205	encode_av (enc_t *enc, AV *av)	480	encode_av (enc_t *enc, AV *av)
206	{	481	{
207	int i, len = av_len (av);	482	int i, len = av_len (av);
208		483
		484	if (enc->indent >= enc->json.max_depth)
		485	croak (ERR_NESTING_EXCEEDED);
		486
209	encode_ch (enc, '['); NL;	487	encode_ch (enc, '[');
210	++enc->indent;	488
		489	if (len >= 0)
		490	{
		491	encode_nl (enc); ++enc->indent;
211		492
212	for (i = 0; i <= len; ++i)	493	for (i = 0; i <= len; ++i)
213	{	494	{
214	INDENT;	495	SV **svp = av_fetch (av, i, 0);
215	encode_sv (enc, *av_fetch (av, i, 0));
216		496
		497	encode_indent (enc);
		498
		499	if (svp)
		500	encode_sv (enc, *svp);
		501	else
		502	encode_str (enc, "null", 4, 0);
		503
217	if (i < len)	504	if (i < len)
218	COMMA;	505	encode_comma (enc);
219	}	506	}
220		507
221	NL;	508	encode_nl (enc); --enc->indent; encode_indent (enc);
222		509	}
223	--enc->indent;	510
224	INDENT; encode_ch (enc, ']');	511	encode_ch (enc, ']');
225	}	512	}
226		513
227	static void	514	static void
228	encode_he (enc_t *enc, HE *he)	515	encode_hk (enc_t *enc, HE *he)
229	{	516	{
230	encode_ch (enc, '"');	517	encode_ch (enc, '"');
231		518
232	if (HeKLEN (he) == HEf_SVKEY)	519	if (HeKLEN (he) == HEf_SVKEY)
233	{	520	{
…		…
243	else	530	else
244	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	531	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
245		532
246	encode_ch (enc, '"');	533	encode_ch (enc, '"');
247		534
248	if (enc->flags & F_SPACE_BEFORE) SPACE;	535	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
249	encode_ch (enc, ':');	536	encode_ch (enc, ':');
250	if (enc->flags & F_SPACE_AFTER ) SPACE;	537	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
251	encode_sv (enc, HeVAL (he));
252	}	538	}
253		539
254	// compare hash entries, used when all keys are bytestrings	540	// compare hash entries, used when all keys are bytestrings
255	static int	541	static int
256	he_cmp_fast (const void *a_, const void *b_)	542	he_cmp_fast (const void *a_, const void *b_)
…		…
261	HE *b = *(HE **)b_;	547	HE *b = *(HE **)b_;
262		548
263	STRLEN la = HeKLEN (a);	549	STRLEN la = HeKLEN (a);
264	STRLEN lb = HeKLEN (b);	550	STRLEN lb = HeKLEN (b);
265		551
266	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	552	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
267	cmp = la < lb ? -1 : la == lb ? 0 : 1;	553	cmp = lb - la;
268		554
269	return cmp;	555	return cmp;
270	}	556	}
271		557
272	// compare hash entries, used when some keys are sv's or utf-x	558	// compare hash entries, used when some keys are sv's or utf-x
273	static int	559	static int
274	he_cmp_slow (const void *a, const void *b)	560	he_cmp_slow (const void *a, const void *b)
275	{	561	{
276	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	562	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
277	}	563	}
278		564
279	static void	565	static void
280	encode_hv (enc_t *enc, HV *hv)	566	encode_hv (enc_t *enc, HV *hv)
281	{	567	{
282	int count, i;	568	HE *he;
283		569
284	encode_ch (enc, '{'); NL; ++enc->indent;	570	if (enc->indent >= enc->json.max_depth)
		571	croak (ERR_NESTING_EXCEEDED);
285		572
286	if ((count = hv_iterinit (hv)))	573	encode_ch (enc, '{');
287	{	574
288	// for canonical output we have to sort by keys first	575	// for canonical output we have to sort by keys first
289	// actually, this is mostly due to the stupid so-called	576	// actually, this is mostly due to the stupid so-called
290	// security workaround added somewhere in 5.8.x.	577	// security workaround added somewhere in 5.8.x
291	// that randomises hash orderings	578	// that randomises hash orderings
292	if (enc->flags & F_CANONICAL)	579	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
		580	{
		581	int count = hv_iterinit (hv);
		582
		583	if (SvMAGICAL (hv))
293	{	584	{
294	HE *he, *hes [count];	585	// need to count by iterating. could improve by dynamically building the vector below
		586	// but I don't care for the speed of this special case.
		587	// note also that we will run into undefined behaviour when the two iterations
		588	// do not result in the same count, something I might care for in some later release.
		589
		590	count = 0;
		591	while (hv_iternext (hv))
		592	++count;
		593
		594	hv_iterinit (hv);
		595	}
		596
		597	if (count)
		598	{
295	int fast = 1;	599	int i, fast = 1;
		600	HE *hes_stack [STACK_HES];
		601	HE **hes = hes_stack;
		602
		603	// allocate larger arrays on the heap
		604	if (count > STACK_HES)
		605	{
		606	SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
		607	hes = (HE **)SvPVX (sv);
		608	}
296		609
297	i = 0;	610	i = 0;
298	while ((he = hv_iternext (hv)))	611	while ((he = hv_iternext (hv)))
299	{	612	{
300	hes [i++] = he;	613	hes [i++] = he;
…		…
306		619
307	if (fast)	620	if (fast)
308	qsort (hes, count, sizeof (HE *), he_cmp_fast);	621	qsort (hes, count, sizeof (HE *), he_cmp_fast);
309	else	622	else
310	{	623	{
311	// hack to disable "use bytes"	624	// hack to forcefully disable "use bytes"
312	COP *oldcop = PL_curcop, cop;	625	COP cop = *PL_curcop;
313	cop.op_private = 0;	626	cop.op_private = 0;
		627
		628	ENTER;
		629	SAVETMPS;
		630
		631	SAVEVPTR (PL_curcop);
314	PL_curcop = &cop;	632	PL_curcop = &cop;
315		633
316	SAVETMPS;
317	qsort (hes, count, sizeof (HE *), he_cmp_slow);	634	qsort (hes, count, sizeof (HE *), he_cmp_slow);
		635
318	FREETMPS;	636	FREETMPS;
		637	LEAVE;
		638	}
319		639
320	PL_curcop = oldcop;	640	encode_nl (enc); ++enc->indent;
		641
		642	while (count--)
321	}	643	{
		644	encode_indent (enc);
		645	he = hes [count];
		646	encode_hk (enc, he);
		647	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
322		648
323	for (i = 0; i < count; ++i)	649	if (count)
		650	encode_comma (enc);
		651	}
		652
		653	encode_nl (enc); --enc->indent; encode_indent (enc);
		654	}
		655	}
		656	else
		657	{
		658	if (hv_iterinit (hv) || SvMAGICAL (hv))
		659	if ((he = hv_iternext (hv)))
		660	{
		661	encode_nl (enc); ++enc->indent;
		662
		663	for (;;)
324	{	664	{
325	INDENT;	665	encode_indent (enc);
326	encode_he (enc, hes [i]);	666	encode_hk (enc, he);
		667	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
327		668
328	if (i < count - 1)	669	if (!(he = hv_iternext (hv)))
329	COMMA;	670	break;
		671
		672	encode_comma (enc);
330	}	673	}
331		674
		675	encode_nl (enc); --enc->indent; encode_indent (enc);
332	NL;	676	}
		677	}
		678
		679	encode_ch (enc, '}');
		680	}
		681
		682	// encode objects, arrays and special \0=false and \1=true values.
		683	static void
		684	encode_rv (enc_t *enc, SV *sv)
		685	{
		686	svtype svt;
		687
		688	SvGETMAGIC (sv);
		689	svt = SvTYPE (sv);
		690
		691	if (expect_false (SvOBJECT (sv)))
		692	{
		693	HV *stash = !JSON_SLOW || json_boolean_stash
		694	? json_boolean_stash
		695	: gv_stashpv ("JSON::XS::Boolean", 1);
		696
		697	if (SvSTASH (sv) == stash)
		698	{
		699	if (SvIV (sv))
		700	encode_str (enc, "true", 4, 0);
		701	else
		702	encode_str (enc, "false", 5, 0);
333	}	703	}
334	else	704	else
335	{	705	{
336	SV *sv;	706	#if 0
337	HE *he = hv_iternext (hv);	707	if (0 && sv_derived_from (rv, "JSON::Literal"))
338
339	for (;;)
340	{
341	INDENT;
342	encode_he (enc, he);
343
344	if (!(he = hv_iternext (hv)))
345	break;
346
347	COMMA;
348	}	708	{
349		709	// not yet
350	NL;	710	}
		711	#endif
		712	if (enc->json.flags & F_CONV_BLESSED)
		713	{
		714	// we re-bless the reference to get overload and other niceties right
		715	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
		716
		717	if (to_json)
		718	{
		719	dSP;
		720
		721	ENTER; SAVETMPS; PUSHMARK (SP);
		722	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		723
		724	// calling with G_SCALAR ensures that we always get a 1 return value
		725	PUTBACK;
		726	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		727	SPAGAIN;
		728
		729	// catch this surprisingly common error
		730	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		731	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		732
		733	sv = POPs;
		734	PUTBACK;
		735
		736	encode_sv (enc, sv);
		737
		738	FREETMPS; LEAVE;
		739	}
		740	else if (enc->json.flags & F_ALLOW_BLESSED)
		741	encode_str (enc, "null", 4, 0);
		742	else
		743	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		744	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		745	}
		746	else if (enc->json.flags & F_ALLOW_BLESSED)
		747	encode_str (enc, "null", 4, 0);
		748	else
		749	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		750	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
351	}	751	}
352	}	752	}
		753	else if (svt == SVt_PVHV)
		754	encode_hv (enc, (HV *)sv);
		755	else if (svt == SVt_PVAV)
		756	encode_av (enc, (AV *)sv);
		757	else if (svt < SVt_PVAV)
		758	{
		759	STRLEN len = 0;
		760	char *pv = svt ? SvPV (sv, len) : 0;
353		761
354	--enc->indent; INDENT; encode_ch (enc, '}');	762	if (len == 1 && *pv == '1')
		763	encode_str (enc, "true", 4, 0);
		764	else if (len == 1 && *pv == '0')
		765	encode_str (enc, "false", 5, 0);
		766	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		767	encode_str (enc, "null", 4, 0);
		768	else
		769	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		770	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		771	}
		772	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		773	encode_str (enc, "null", 4, 0);
		774	else
		775	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		776	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
355	}	777	}
356		778
357	static void	779	static void
358	encode_sv (enc_t *enc, SV *sv)	780	encode_sv (enc_t *enc, SV *sv)
359	{	781	{
…		…
367	encode_str (enc, str, len, SvUTF8 (sv));	789	encode_str (enc, str, len, SvUTF8 (sv));
368	encode_ch (enc, '"');	790	encode_ch (enc, '"');
369	}	791	}
370	else if (SvNOKp (sv))	792	else if (SvNOKp (sv))
371	{	793	{
		794	// trust that perl will do the right thing w.r.t. JSON syntax.
372	need (enc, NV_DIG + 32);	795	need (enc, NV_DIG + 32);
373	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	796	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
374	enc->cur += strlen (enc->cur);	797	enc->cur += strlen (enc->cur);
375	}	798	}
376	else if (SvIOKp (sv))	799	else if (SvIOKp (sv))
377	{	800	{
		801	// we assume we can always read an IV as a UV and vice versa
		802	// we assume two's complement
		803	// we assume no aliasing issues in the union
		804	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		805	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		806	{
		807	// optimise the "small number case"
		808	// code will likely be branchless and use only a single multiplication
		809	// works for numbers up to 59074
		810	I32 i = SvIVX (sv);
		811	U32 u;
		812	char digit, nz = 0;
		813
378	need (enc, 64);	814	need (enc, 6);
		815
		816	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		817	u = i < 0 ? -i : i;
		818
		819	// convert to 4.28 fixed-point representation
		820	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		821
		822	// now output digit by digit, each time masking out the integer part
		823	// and multiplying by 5 while moving the decimal point one to the right,
		824	// resulting in a net multiplication by 10.
		825	// we always write the digit to memory but conditionally increment
		826	// the pointer, to enable the use of conditional move instructions.
		827	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		828	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		829	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		830	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		831	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		832	}
		833	else
		834	{
		835	// large integer, use the (rather slow) snprintf way.
		836	need (enc, IVUV_MAXCHARS);
379	enc->cur +=	837	enc->cur +=
380	SvIsUV(sv)	838	SvIsUV(sv)
381	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	839	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
382	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	840	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		841	}
383	}	842	}
384	else if (SvROK (sv))	843	else if (SvROK (sv))
385	{	844	encode_rv (enc, SvRV (sv));
386	if (!--enc->max_recurse)	845	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);	846	encode_str (enc, "null", 4, 0);
402	else	847	else
403	croak ("encountered perl type that JSON cannot handle");	848	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
		849	SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
404	}	850	}
405		851
406	static SV *	852	static SV *
407	encode_json (SV *scalar, UV flags)	853	encode_json (SV *scalar, JSON *json)
408	{	854	{
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;	855	enc_t enc;
413	enc.flags = flags;	856
		857	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
		858	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
		859
		860	enc.json = *json;
414	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	861	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
415	enc.cur = SvPVX (enc.sv);	862	enc.cur = SvPVX (enc.sv);
416	enc.end = SvEND (enc.sv);	863	enc.end = SvEND (enc.sv);
417	enc.max_recurse = 0;
418	enc.indent = 0;	864	enc.indent = 0;
		865	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		866	: enc.json.flags & F_LATIN1 ? 0x000100UL
		867	: 0x110000UL;
419		868
420	SvPOK_only (enc.sv);	869	SvPOK_only (enc.sv);
421	encode_sv (&enc, scalar);	870	encode_sv (&enc, scalar);
		871	encode_nl (&enc);
422		872
		873	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		874	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		875
423	if (!(flags & (F_ASCII | F_UTF8)))	876	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
424	SvUTF8_on (enc.sv);	877	SvUTF8_on (enc.sv);
425		878
426	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	879	if (enc.json.flags & F_SHRINK)
		880	shrink (enc.sv);
		881
427	return enc.sv;	882	return enc.sv;
428	}	883	}
429		884
430	/////////////////////////////////////////////////////////////////////////////	885	/////////////////////////////////////////////////////////////////////////////
		886	// decoder
431		887
432	#define WS \	888	// structure used for decoding JSON
		889	typedef struct
		890	{
		891	char *cur; // current parser pointer
		892	char *end; // end of input string
		893	const char *err; // parse error, if != 0
		894	JSON json;
		895	U32 depth; // recursion depth
		896	U32 maxdepth; // recursion depth limit
		897	} dec_t;
		898
		899	INLINE void
		900	decode_comment (dec_t *dec)
		901	{
		902	// only '#'-style comments allowed a.t.m.
		903
		904	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		905	++dec->cur;
		906	}
		907
		908	INLINE void
		909	decode_ws (dec_t *dec)
		910	{
433	for (;;) \	911	for (;;)
434	{ \	912	{
435	char ch = *dec->cur; \	913	char ch = *dec->cur;
		914
436	if (ch > 0x20 \	915	if (ch > 0x20)
		916	{
		917	if (expect_false (ch == '#'))
		918	{
		919	if (dec->json.flags & F_RELAXED)
		920	decode_comment (dec);
		921	else
		922	break;
		923	}
		924	else
		925	break;
		926	}
437	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	927	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
438	break; \	928	break; // parse error, but let higher level handle it, gives better error messages
		929
439	++dec->cur; \	930	++dec->cur;
440	}	931	}
		932	}
441		933
442	#define ERR(reason) SB dec->err = reason; goto fail; SE	934	#define ERR(reason) SB dec->err = reason; goto fail; SE
		935
443	#define EXPECT_CH(ch) SB \	936	#define EXPECT_CH(ch) SB \
444	if (*dec->cur != ch) \	937	if (*dec->cur != ch) \
445	ERR (# ch " expected"); \	938	ERR (# ch " expected"); \
446	++dec->cur; \	939	++dec->cur; \
447	SE	940	SE
448		941
		942	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
		943	#define DEC_DEC_DEPTH --dec->depth
		944
449	static SV *decode_sv (dec_t *dec);	945	static SV *decode_sv (dec_t *dec);
450		946
451	static signed char decode_hexdigit[256];	947	static signed char decode_hexdigit[256];
452		948
453	static UV	949	static UV
454	decode_4hex (dec_t *dec)	950	decode_4hex (dec_t *dec)
455	{	951	{
456	signed char d1, d2, d3, d4;	952	signed char d1, d2, d3, d4;
		953	unsigned char *cur = (unsigned char *)dec->cur;
457		954
458	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	955	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");	956	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]];	957	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");	958	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		959
467	dec->cur += 4;	960	dec->cur += 4;
468		961
469	return ((UV)d1) << 12	962	return ((UV)d1) << 12
470	| ((UV)d2) << 8	963	| ((UV)d2) << 8
…		…
473		966
474	fail:	967	fail:
475	return (UV)-1;	968	return (UV)-1;
476	}	969	}
477		970
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 *	971	static SV *
494	decode_str (dec_t *dec)	972	decode_str (dec_t *dec)
495	{	973	{
496	SV *sv = NEWSV (0,2);	974	SV *sv = 0;
497	int utf8 = 0;	975	int utf8 = 0;
498	char *cur = SvPVX (sv);	976	char *dec_cur = dec->cur;
499	char *end = SvEND (sv);
500		977
501	for (;;)	978	do
502	{	979	{
503	unsigned char ch = *(unsigned char *)dec->cur;	980	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		981	char *cur = buf;
504		982
505	if (ch == '"')	983	do
506	break;
507	else if (ch == '\\')
508	{	984	{
509	switch (*++dec->cur)	985	unsigned char ch = *(unsigned char *)dec_cur++;
		986
		987	if (expect_false (ch == '"'))
		988	{
		989	--dec_cur;
		990	break;
510	{	991	}
511	case '\\':	992	else if (expect_false (ch == '\\'))
512	case '/':	993	{
513	case '"': APPEND_CH (*dec->cur++); break;	994	switch (*dec_cur)
514
515	case 'b': APPEND_CH ('\010'); ++dec->cur; break;
516	case 't': APPEND_CH ('\011'); ++dec->cur; break;
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	{	995	{
523	UV lo, hi;	996	case '\\':
524	++dec->cur;	997	case '/':
		998	case '"': *cur++ = *dec_cur++; break;
525		999
526	hi = decode_4hex (dec);	1000	case 'b': ++dec_cur; *cur++ = '\010'; break;
527	if (hi == (UV)-1)	1001	case 't': ++dec_cur; *cur++ = '\011'; break;
528	goto fail;	1002	case 'n': ++dec_cur; *cur++ = '\012'; break;
		1003	case 'f': ++dec_cur; *cur++ = '\014'; break;
		1004	case 'r': ++dec_cur; *cur++ = '\015'; break;
529		1005
530	// possibly a surrogate pair	1006	case 'u':
531	if (hi >= 0xd800 && hi < 0xdc00)
532	{	1007	{
533	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	1008	UV lo, hi;
534	ERR ("illegal surrogate character");	1009	++dec_cur;
535		1010
536	dec->cur += 2;	1011	dec->cur = dec_cur;
537
538	lo = decode_4hex (dec);	1012	hi = decode_4hex (dec);
		1013	dec_cur = dec->cur;
539	if (lo == (UV)-1)	1014	if (hi == (UV)-1)
540	goto fail;	1015	goto fail;
541		1016
		1017	// possibly a surrogate pair
		1018	if (hi >= 0xd800)
		1019	if (hi < 0xdc00)
		1020	{
		1021	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		1022	ERR ("missing low surrogate character in surrogate pair");
		1023
		1024	dec_cur += 2;
		1025
		1026	dec->cur = dec_cur;
		1027	lo = decode_4hex (dec);
		1028	dec_cur = dec->cur;
		1029	if (lo == (UV)-1)
		1030	goto fail;
		1031
542	if (lo < 0xdc00 || lo >= 0xe000)	1032	if (lo < 0xdc00 || lo >= 0xe000)
543	ERR ("surrogate pair expected");	1033	ERR ("surrogate pair expected");
544		1034
545	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	1035	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		1036	}
		1037	else if (hi < 0xe000)
		1038	ERR ("missing high surrogate character in surrogate pair");
		1039
		1040	if (hi >= 0x80)
		1041	{
		1042	utf8 = 1;
		1043
		1044	cur = encode_utf8 (cur, hi);
		1045	}
		1046	else
		1047	*cur++ = hi;
546	}	1048	}
547	else if (lo >= 0xdc00 && lo < 0xe000)
548	ERR ("illegal surrogate character");
549
550	if (hi >= 0x80)
551	{	1049	break;
552	utf8 = 1;
553		1050
554	APPEND_GROW (4); // at most 4 bytes for 21 bits
555	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
556	}
557	else	1051	default:
558	APPEND_CH (hi);	1052	--dec_cur;
		1053	ERR ("illegal backslash escape sequence in string");
559	}	1054	}
		1055	}
		1056	else if (expect_true (ch >= 0x20 && ch < 0x80))
		1057	*cur++ = ch;
		1058	else if (ch >= 0x80)
		1059	{
		1060	STRLEN clen;
		1061
		1062	--dec_cur;
		1063
		1064	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		1065	if (clen == (STRLEN)-1)
		1066	ERR ("malformed UTF-8 character in JSON string");
		1067
		1068	do
		1069	*cur++ = *dec_cur++;
		1070	while (--clen);
		1071
		1072	utf8 = 1;
		1073	}
		1074	else
		1075	{
		1076	--dec_cur;
		1077
		1078	if (!ch)
		1079	ERR ("unexpected end of string while parsing JSON string");
560	break;	1080	else
		1081	ERR ("invalid character encountered while parsing JSON string");
561	}	1082	}
562	}	1083	}
563	else if (ch >= 0x20 && ch <= 0x7f)	1084	while (cur < buf + SHORT_STRING_LEN);
564	APPEND_CH (*dec->cur++);	1085
565	else if (ch >= 0x80)	1086	{
		1087	STRLEN len = cur - buf;
		1088
		1089	if (sv)
566	{	1090	{
567	STRLEN clen;	1091	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		1092
572	APPEND_GROW (clen);	1093	if (SvLEN (sv) <= cur + len)
573	memcpy (cur, dec->cur, clen);	1094	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
574	cur += clen;	1095
575	dec->cur += clen;	1096	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		1097	SvCUR_set (sv, SvCUR (sv) + len);
576	}	1098	}
577	else	1099	else
578	ERR ("invalid character encountered");	1100	sv = newSVpvn (buf, len);
579	}	1101	}
		1102	}
		1103	while (*dec_cur != '"');
580		1104
581	++dec->cur;	1105	++dec_cur;
582		1106
583	SvCUR_set (sv, cur - SvPVX (sv));	1107	if (sv)
584		1108	{
585	SvPOK_only (sv);	1109	SvPOK_only (sv);
586	*SvEND (sv) = 0;	1110	*SvEND (sv) = 0;
587		1111
588	if (utf8)	1112	if (utf8)
589	SvUTF8_on (sv);	1113	SvUTF8_on (sv);
		1114	}
		1115	else
		1116	sv = newSVpvn ("", 0);
590		1117
		1118	dec->cur = dec_cur;
591	return sv;	1119	return sv;
592		1120
593	fail:	1121	fail:
594	SvREFCNT_dec (sv);	1122	dec->cur = dec_cur;
595	return 0;	1123	return 0;
596	}	1124	}
597		1125
598	static SV *	1126	static SV *
599	decode_num (dec_t *dec)	1127	decode_num (dec_t *dec)
…		…
609	{	1137	{
610	++dec->cur;	1138	++dec->cur;
611	if (*dec->cur >= '0' && *dec->cur <= '9')	1139	if (*dec->cur >= '0' && *dec->cur <= '9')
612	ERR ("malformed number (leading zero must not be followed by another digit)");	1140	ERR ("malformed number (leading zero must not be followed by another digit)");
613	}	1141	}
614		1142	else if (*dec->cur < '0' || *dec->cur > '9')
615	// int	1143	ERR ("malformed number (no digits after initial minus)");
		1144	else
		1145	do
		1146	{
		1147	++dec->cur;
		1148	}
616	while (*dec->cur >= '0' && *dec->cur <= '9')	1149	while (*dec->cur >= '0' && *dec->cur <= '9');
617	++dec->cur;
618		1150
619	// [frac]	1151	// [frac]
620	if (*dec->cur == '.')	1152	if (*dec->cur == '.')
621	{	1153	{
622	is_nv = 1;	1154	++dec->cur;
		1155
		1156	if (*dec->cur < '0' || *dec->cur > '9')
		1157	ERR ("malformed number (no digits after decimal point)");
623		1158
624	do	1159	do
625	{	1160	{
626	++dec->cur;	1161	++dec->cur;
627	}	1162	}
628	while (*dec->cur >= '0' && *dec->cur <= '9');	1163	while (*dec->cur >= '0' && *dec->cur <= '9');
		1164
		1165	is_nv = 1;
629	}	1166	}
630		1167
631	// [exp]	1168	// [exp]
632	if (*dec->cur == 'e' || *dec->cur == 'E')	1169	if (*dec->cur == 'e' || *dec->cur == 'E')
633	{	1170	{
634	is_nv = 1;
635
636	++dec->cur;	1171	++dec->cur;
		1172
637	if (*dec->cur == '-' || *dec->cur == '+')	1173	if (*dec->cur == '-' || *dec->cur == '+')
638	++dec->cur;	1174	++dec->cur;
639		1175
		1176	if (*dec->cur < '0' || *dec->cur > '9')
		1177	ERR ("malformed number (no digits after exp sign)");
		1178
		1179	do
		1180	{
		1181	++dec->cur;
		1182	}
640	while (*dec->cur >= '0' && *dec->cur <= '9')	1183	while (*dec->cur >= '0' && *dec->cur <= '9');
641	++dec->cur;	1184
		1185	is_nv = 1;
642	}	1186	}
643		1187
644	if (!is_nv)	1188	if (!is_nv)
645	{	1189	{
646	UV uv;	1190	int len = dec->cur - start;
647	int numtype = grok_number (start, dec->cur - start, &uv);	1191
648	if (numtype & IS_NUMBER_IN_UV)	1192	// special case the rather common 1..5-digit-int case
649	if (numtype & IS_NUMBER_NEG)	1193	if (*start == '-')
		1194	switch (len)
650	{	1195	{
651	if (uv < (UV)IV_MIN)	1196	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
652	return newSViv (-(IV)uv);	1197	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
		1198	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1199	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1200	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
653	}	1201	}
		1202	else
		1203	switch (len)
		1204	{
		1205	case 1: return newSViv ( start [0] - '0' * 1);
		1206	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1207	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1208	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1209	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1210	}
		1211
		1212	{
		1213	UV uv;
		1214	int numtype = grok_number (start, len, &uv);
		1215	if (numtype & IS_NUMBER_IN_UV)
		1216	if (numtype & IS_NUMBER_NEG)
		1217	{
		1218	if (uv < (UV)IV_MIN)
		1219	return newSViv (-(IV)uv);
		1220	}
654	else	1221	else
655	return newSVuv (uv);	1222	return newSVuv (uv);
656	}	1223	}
657		1224
		1225	len -= *start == '-' ? 1 : 0;
		1226
		1227	// does not fit into IV or UV, try NV
		1228	if (len <= NV_DIG)
		1229	// fits into NV without loss of precision
		1230	return newSVnv (json_atof (start));
		1231
		1232	// everything else fails, convert it to a string
		1233	return newSVpvn (start, dec->cur - start);
		1234	}
		1235
		1236	// loss of precision here
658	return newSVnv (Atof (start));	1237	return newSVnv (json_atof (start));
659		1238
660	fail:	1239	fail:
661	return 0;	1240	return 0;
662	}	1241	}
663		1242
664	static SV *	1243	static SV *
665	decode_av (dec_t *dec)	1244	decode_av (dec_t *dec)
666	{	1245	{
667	AV *av = newAV ();	1246	AV *av = newAV ();
668		1247
		1248	DEC_INC_DEPTH;
		1249	decode_ws (dec);
		1250
		1251	if (*dec->cur == ']')
		1252	++dec->cur;
		1253	else
669	for (;;)	1254	for (;;)
670	{	1255	{
671	SV *value;	1256	SV *value;
672		1257
673	value = decode_sv (dec);	1258	value = decode_sv (dec);
674	if (!value)	1259	if (!value)
675	goto fail;	1260	goto fail;
676		1261
677	av_push (av, value);	1262	av_push (av, value);
678		1263
679	WS;	1264	decode_ws (dec);
680		1265
681	if (*dec->cur == ']')	1266	if (*dec->cur == ']')
682	{	1267	{
683	++dec->cur;	1268	++dec->cur;
684	break;	1269	break;
		1270	}
685	}	1271
686
687	if (*dec->cur != ',')	1272	if (*dec->cur != ',')
688	ERR (", or ] expected while parsing array");	1273	ERR (", or ] expected while parsing array");
689		1274
690	++dec->cur;	1275	++dec->cur;
		1276
		1277	decode_ws (dec);
		1278
		1279	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1280	{
		1281	++dec->cur;
		1282	break;
		1283	}
691	}	1284	}
692		1285
		1286	DEC_DEC_DEPTH;
693	return newRV_noinc ((SV *)av);	1287	return newRV_noinc ((SV *)av);
694		1288
695	fail:	1289	fail:
696	SvREFCNT_dec (av);	1290	SvREFCNT_dec (av);
		1291	DEC_DEC_DEPTH;
697	return 0;	1292	return 0;
698	}	1293	}
699		1294
700	static SV *	1295	static SV *
701	decode_hv (dec_t *dec)	1296	decode_hv (dec_t *dec)
702	{	1297	{
		1298	SV *sv;
703	HV *hv = newHV ();	1299	HV *hv = newHV ();
704		1300
		1301	DEC_INC_DEPTH;
		1302	decode_ws (dec);
		1303
		1304	if (*dec->cur == '}')
		1305	++dec->cur;
		1306	else
705	for (;;)	1307	for (;;)
706	{	1308	{
707	SV *key, *value;
708
709	WS; EXPECT_CH ('"');	1309	EXPECT_CH ('"');
710		1310
711	key = decode_str (dec);	1311	// heuristic: assume that
712	if (!key)	1312	// a) decode_str + hv_store_ent are abysmally slow.
713	goto fail;	1313	// b) most hash keys are short, simple ascii text.
714		1314	// => try to "fast-match" such strings to avoid
715	WS; EXPECT_CH (':');	1315	// the overhead of decode_str + hv_store_ent.
716
717	value = decode_sv (dec);
718	if (!value)
719	{	1316	{
		1317	SV *value;
		1318	char *p = dec->cur;
		1319	char *e = p + 24; // only try up to 24 bytes
		1320
		1321	for (;;)
		1322	{
		1323	// the >= 0x80 is false on most architectures
		1324	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1325	{
		1326	// slow path, back up and use decode_str
		1327	SV *key = decode_str (dec);
		1328	if (!key)
		1329	goto fail;
		1330
		1331	decode_ws (dec); EXPECT_CH (':');
		1332
		1333	decode_ws (dec);
		1334	value = decode_sv (dec);
		1335	if (!value)
		1336	{
		1337	SvREFCNT_dec (key);
		1338	goto fail;
		1339	}
		1340
		1341	hv_store_ent (hv, key, value, 0);
720	SvREFCNT_dec (key);	1342	SvREFCNT_dec (key);
		1343
		1344	break;
		1345	}
		1346	else if (*p == '"')
		1347	{
		1348	// fast path, got a simple key
		1349	char *key = dec->cur;
		1350	int len = p - key;
		1351	dec->cur = p + 1;
		1352
		1353	decode_ws (dec); EXPECT_CH (':');
		1354
		1355	decode_ws (dec);
		1356	value = decode_sv (dec);
		1357	if (!value)
721	goto fail;	1358	goto fail;
		1359
		1360	hv_store (hv, key, len, value, 0);
		1361
		1362	break;
		1363	}
		1364
		1365	++p;
		1366	}
722	}	1367	}
723		1368
724	//TODO: optimise	1369	decode_ws (dec);
725	hv_store_ent (hv, key, value, 0);
726		1370
727	WS;
728
729	if (*dec->cur == '}')	1371	if (*dec->cur == '}')
		1372	{
		1373	++dec->cur;
		1374	break;
		1375	}
		1376
		1377	if (*dec->cur != ',')
		1378	ERR (", or } expected while parsing object/hash");
		1379
		1380	++dec->cur;
		1381
		1382	decode_ws (dec);
		1383
		1384	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1385	{
		1386	++dec->cur;
		1387	break;
		1388	}
		1389	}
		1390
		1391	DEC_DEC_DEPTH;
		1392	sv = newRV_noinc ((SV *)hv);
		1393
		1394	// check filter callbacks
		1395	if (dec->json.flags & F_HOOK)
		1396	{
		1397	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
730	{	1398	{
731	++dec->cur;	1399	HE *cb, *he;
732	break;	1400
		1401	hv_iterinit (hv);
		1402	he = hv_iternext (hv);
		1403	hv_iterinit (hv);
		1404
		1405	// the next line creates a mortal sv each time its called.
		1406	// might want to optimise this for common cases.
		1407	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1408
		1409	if (cb)
		1410	{
		1411	dSP;
		1412	int count;
		1413
		1414	ENTER; SAVETMPS; PUSHMARK (SP);
		1415	XPUSHs (HeVAL (he));
		1416	sv_2mortal (sv);
		1417
		1418	PUTBACK; count = call_sv (HeVAL (cb), 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	}
733	}	1430	}
734		1431
735	if (*dec->cur != ',')	1432	if (dec->json.cb_object)
736	ERR (", or } expected while parsing object/hash");	1433	{
		1434	dSP;
		1435	int count;
737		1436
738	++dec->cur;	1437	ENTER; SAVETMPS; PUSHMARK (SP);
739	}	1438	XPUSHs (sv_2mortal (sv));
740		1439
741	return newRV_noinc ((SV *)hv);	1440	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1441
		1442	if (count == 1)
		1443	{
		1444	sv = newSVsv (POPs);
		1445	FREETMPS; LEAVE;
		1446	return sv;
		1447	}
		1448
		1449	SvREFCNT_inc (sv);
		1450	FREETMPS; LEAVE;
		1451	}
		1452	}
		1453
		1454	return sv;
742		1455
743	fail:	1456	fail:
744	SvREFCNT_dec (hv);	1457	SvREFCNT_dec (hv);
		1458	DEC_DEC_DEPTH;
745	return 0;	1459	return 0;
746	}	1460	}
747		1461
748	static SV *	1462	static SV *
749	decode_sv (dec_t *dec)	1463	decode_sv (dec_t *dec)
750	{	1464	{
751	WS;	1465	// the beauty of JSON: you need exactly one character lookahead
		1466	// to parse everything.
752	switch (*dec->cur)	1467	switch (*dec->cur)
753	{	1468	{
754	case '"': ++dec->cur; return decode_str (dec);	1469	case '"': ++dec->cur; return decode_str (dec);
755	case '[': ++dec->cur; return decode_av (dec);	1470	case '[': ++dec->cur; return decode_av (dec);
756	case '{': ++dec->cur; return decode_hv (dec);	1471	case '{': ++dec->cur; return decode_hv (dec);
757		1472
758	case '-':	1473	case '-':
759	case '0': case '1': case '2': case '3': case '4':	1474	case '0': case '1': case '2': case '3': case '4':
760	case '5': case '6': case '7': case '8': case '9':	1475	case '5': case '6': case '7': case '8': case '9':
761	return decode_num (dec);	1476	return decode_num (dec);
762		1477
763	case 't':	1478	case 't':
764	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1479	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
765	{	1480	{
766	dec->cur += 4;	1481	dec->cur += 4;
		1482	#if JSON_SLOW
		1483	json_true = get_bool ("JSON::XS::true");
		1484	#endif
767	return newSViv (1);	1485	return newSVsv (json_true);
768	}	1486	}
769	else	1487	else
770	ERR ("'true' expected");	1488	ERR ("'true' expected");
771		1489
772	break;	1490	break;
773		1491
774	case 'f':	1492	case 'f':
775	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1493	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
776	{	1494	{
777	dec->cur += 5;	1495	dec->cur += 5;
		1496	#if JSON_SLOW
		1497	json_false = get_bool ("JSON::XS::false");
		1498	#endif
778	return newSViv (0);	1499	return newSVsv (json_false);
779	}	1500	}
780	else	1501	else
781	ERR ("'false' expected");	1502	ERR ("'false' expected");
782		1503
783	break;	1504	break;
784		1505
785	case 'n':	1506	case 'n':
786	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))	1507	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
787	{	1508	{
788	dec->cur += 4;	1509	dec->cur += 4;
789	return newSViv (1);	1510	return newSVsv (&PL_sv_undef);
790	}	1511	}
791	else	1512	else
792	ERR ("'null' expected");	1513	ERR ("'null' expected");
793		1514
794	break;	1515	break;
795		1516
796	default:	1517	default:
797	ERR ("malformed json string");	1518	ERR ("malformed JSON string, neither array, object, number, string or atom");
798	break;	1519	break;
799	}	1520	}
800		1521
801	fail:	1522	fail:
802	return 0;	1523	return 0;
803	}	1524	}
804		1525
805	static SV *	1526	static SV *
806	decode_json (SV *string, UV flags)	1527	decode_json (SV *string, JSON *json, char **offset_return)
807	{	1528	{
		1529	dec_t dec;
808	SV *sv;	1530	SV *sv;
809		1531
810	if (!(flags & F_UTF8))	1532	/* work around bugs in 5.10 where manipulating magic values
		1533	* will perl ignore the magic in subsequent accesses.
		1534	* also make a copy of non-PV values, to get them into a clean
		1535	* state (SvPV should do that, but it's buggy, see below).
		1536	*/
		1537	/*SvGETMAGIC (string);*/
		1538	if (SvMAGICAL (string) || !SvPOK (string))
		1539	string = sv_2mortal (newSVsv (string));
		1540
		1541	SvUPGRADE (string, SVt_PV);
		1542
		1543	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1544	* assertion with -DDEBUGGING, although SvCUR is documented to
		1545	* return the xpv_cur field which certainly exists after upgrading.
		1546	* according to nicholas clark, calling SvPOK fixes this.
		1547	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1548	* and hope for the best.
		1549	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1550	* assertion business is seriously broken, try yet another workaround
		1551	* for the broken -DDEBUGGING.
		1552	*/
		1553	{
		1554	#ifdef DEBUGGING
		1555	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1556	#else
		1557	STRLEN offset = SvCUR (string);
		1558	#endif
		1559
		1560	if (offset > json->max_size && json->max_size)
		1561	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1562	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1563	}
		1564
		1565	if (DECODE_WANTS_OCTETS (json))
		1566	sv_utf8_downgrade (string, 0);
		1567	else
811	sv_utf8_upgrade (string);	1568	sv_utf8_upgrade (string);
812		1569
813	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1570	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
814		1571
815	dec_t dec;	1572	dec.json = *json;
816	dec.flags = flags;
817	dec.cur = SvPVX (string);	1573	dec.cur = SvPVX (string);
818	dec.end = SvEND (string);	1574	dec.end = SvEND (string);
819	dec.err = 0;	1575	dec.err = 0;
		1576	dec.depth = 0;
820		1577
821	*dec.end = 1; // invalid anywhere	1578	if (dec.json.cb_object || dec.json.cb_sk_object)
		1579	dec.json.flags |= F_HOOK;
		1580
		1581	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1582
		1583	decode_ws (&dec);
822	sv = decode_sv (&dec);	1584	sv = decode_sv (&dec);
823	*dec.end = 0;	1585
		1586	if (offset_return)
		1587	*offset_return = dec.cur;
		1588
		1589	if (!(offset_return || !sv))
		1590	{
		1591	// check for trailing garbage
		1592	decode_ws (&dec);
		1593
		1594	if (*dec.cur)
		1595	{
		1596	dec.err = "garbage after JSON object";
		1597	SvREFCNT_dec (sv);
		1598	sv = 0;
		1599	}
		1600	}
824		1601
825	if (!sv)	1602	if (!sv)
826	{	1603	{
827	IV offset = utf8_distance (dec.cur, SvPVX (string));
828	SV *uni = sv_newmortal ();	1604	SV *uni = sv_newmortal ();
829		1605
		1606	// horrible hack to silence warning inside pv_uni_display
		1607	COP cop = *PL_curcop;
		1608	cop.cop_warnings = pWARN_NONE;
		1609	ENTER;
		1610	SAVEVPTR (PL_curcop);
		1611	PL_curcop = &cop;
830	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1612	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
		1613	LEAVE;
		1614
831	croak ("%s, at character %d (%s)",	1615	croak ("%s, at character offset %d (before \"%s\")",
832	dec.err,	1616	dec.err,
833	(int)offset,	1617	(int)ptr_to_index (string, dec.cur),
834	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1618	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
835	}	1619	}
836		1620
837	sv = sv_2mortal (sv);	1621	sv = sv_2mortal (sv);
838		1622
839	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1623	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)");	1624	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
841		1625
842	return sv;	1626	return sv;
843	}	1627	}
844		1628
		1629	/////////////////////////////////////////////////////////////////////////////
		1630	// incremental parser
		1631
		1632	static void
		1633	incr_parse (JSON *self)
		1634	{
		1635	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1636
		1637	// the state machine here is a bit convoluted and could be simplified a lot
		1638	// but this would make it slower, so...
		1639
		1640	for (;;)
		1641	{
		1642	//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
		1643	switch (self->incr_mode)
		1644	{
		1645	// only used for initial whitespace skipping
		1646	case INCR_M_WS:
		1647	for (;;)
		1648	{
		1649	if (*p > 0x20)
		1650	{
		1651	if (*p == '#')
		1652	{
		1653	self->incr_mode = INCR_M_C0;
		1654	goto incr_m_c;
		1655	}
		1656	else
		1657	{
		1658	self->incr_mode = INCR_M_JSON;
		1659	goto incr_m_json;
		1660	}
		1661	}
		1662	else if (!*p)
		1663	goto interrupt;
		1664
		1665	++p;
		1666	}
		1667
		1668	// skip a single char inside a string (for \\-processing)
		1669	case INCR_M_BS:
		1670	if (!*p)
		1671	goto interrupt;
		1672
		1673	++p;
		1674	self->incr_mode = INCR_M_STR;
		1675	goto incr_m_str;
		1676
		1677	// inside #-style comments
		1678	case INCR_M_C0:
		1679	case INCR_M_C1:
		1680	incr_m_c:
		1681	for (;;)
		1682	{
		1683	if (*p == '\n')
		1684	{
		1685	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1686	break;
		1687	}
		1688	else if (!*p)
		1689	goto interrupt;
		1690
		1691	++p;
		1692	}
		1693
		1694	break;
		1695
		1696	// inside a string
		1697	case INCR_M_STR:
		1698	incr_m_str:
		1699	for (;;)
		1700	{
		1701	if (*p == '"')
		1702	{
		1703	++p;
		1704	self->incr_mode = INCR_M_JSON;
		1705
		1706	if (!self->incr_nest)
		1707	goto interrupt;
		1708
		1709	goto incr_m_json;
		1710	}
		1711	else if (*p == '\\')
		1712	{
		1713	++p; // "virtually" consumes character after \
		1714
		1715	if (!*p) // if at end of string we have to switch modes
		1716	{
		1717	self->incr_mode = INCR_M_BS;
		1718	goto interrupt;
		1719	}
		1720	}
		1721	else if (!*p)
		1722	goto interrupt;
		1723
		1724	++p;
		1725	}
		1726
		1727	// after initial ws, outside string
		1728	case INCR_M_JSON:
		1729	incr_m_json:
		1730	for (;;)
		1731	{
		1732	switch (*p++)
		1733	{
		1734	case 0:
		1735	--p;
		1736	goto interrupt;
		1737
		1738	case 0x09:
		1739	case 0x0a:
		1740	case 0x0d:
		1741	case 0x20:
		1742	if (!self->incr_nest)
		1743	{
		1744	--p; // do not eat the whitespace, let the next round do it
		1745	goto interrupt;
		1746	}
		1747	break;
		1748
		1749	case '"':
		1750	self->incr_mode = INCR_M_STR;
		1751	goto incr_m_str;
		1752
		1753	case '[':
		1754	case '{':
		1755	if (++self->incr_nest > self->max_depth)
		1756	croak (ERR_NESTING_EXCEEDED);
		1757	break;
		1758
		1759	case ']':
		1760	case '}':
		1761	if (--self->incr_nest <= 0)
		1762	goto interrupt;
		1763	break;
		1764
		1765	case '#':
		1766	self->incr_mode = INCR_M_C1;
		1767	goto incr_m_c;
		1768	}
		1769	}
		1770	}
		1771
		1772	modechange:
		1773	;
		1774	}
		1775
		1776	interrupt:
		1777	self->incr_pos = p - SvPVX (self->incr_text);
		1778	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1779	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
		1780	}
		1781
		1782	/////////////////////////////////////////////////////////////////////////////
		1783	// XS interface functions
		1784
845	MODULE = JSON::XS PACKAGE = JSON::XS	1785	MODULE = JSON::XS PACKAGE = JSON::XS
846		1786
847	BOOT:	1787	BOOT:
848	{	1788	{
849	int i;	1789	int i;
850		1790
851	memset (decode_hexdigit, 0xff, 256);
852	for (i = 10; i--; )	1791	for (i = 0; i < 256; ++i)
853	decode_hexdigit ['0' + i] = i;	1792	decode_hexdigit [i] =
		1793	i >= '0' && i <= '9' ? i - '0'
		1794	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1795	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1796	: -1;
854		1797
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);	1798	json_stash = gv_stashpv ("JSON::XS" , 1);
		1799	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1800
		1801	json_true = get_bool ("JSON::XS::true");
		1802	json_false = get_bool ("JSON::XS::false");
		1803
		1804	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
862	}	1805	}
863		1806
864	PROTOTYPES: DISABLE	1807	PROTOTYPES: DISABLE
865		1808
866	SV *new (char *dummy)	1809	void CLONE (...)
867	CODE:	1810	CODE:
868	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1811	json_stash = 0;
		1812	json_boolean_stash = 0;
		1813
		1814	void new (char *klass)
		1815	PPCODE:
		1816	{
		1817	SV *pv = NEWSV (0, sizeof (JSON));
		1818	SvPOK_only (pv);
		1819	json_init ((JSON *)SvPVX (pv));
		1820	XPUSHs (sv_2mortal (sv_bless (
		1821	newRV_noinc (pv),
		1822	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1823	)));
		1824	}
		1825
		1826	void ascii (JSON *self, int enable = 1)
		1827	ALIAS:
		1828	ascii = F_ASCII
		1829	latin1 = F_LATIN1
		1830	utf8 = F_UTF8
		1831	indent = F_INDENT
		1832	canonical = F_CANONICAL
		1833	space_before = F_SPACE_BEFORE
		1834	space_after = F_SPACE_AFTER
		1835	pretty = F_PRETTY
		1836	allow_nonref = F_ALLOW_NONREF
		1837	shrink = F_SHRINK
		1838	allow_blessed = F_ALLOW_BLESSED
		1839	convert_blessed = F_CONV_BLESSED
		1840	relaxed = F_RELAXED
		1841	allow_unknown = F_ALLOW_UNKNOWN
		1842	PPCODE:
		1843	{
		1844	if (enable)
		1845	self->flags |= ix;
		1846	else
		1847	self->flags &= ~ix;
		1848
		1849	XPUSHs (ST (0));
		1850	}
		1851
		1852	void get_ascii (JSON *self)
		1853	ALIAS:
		1854	get_ascii = F_ASCII
		1855	get_latin1 = F_LATIN1
		1856	get_utf8 = F_UTF8
		1857	get_indent = F_INDENT
		1858	get_canonical = F_CANONICAL
		1859	get_space_before = F_SPACE_BEFORE
		1860	get_space_after = F_SPACE_AFTER
		1861	get_allow_nonref = F_ALLOW_NONREF
		1862	get_shrink = F_SHRINK
		1863	get_allow_blessed = F_ALLOW_BLESSED
		1864	get_convert_blessed = F_CONV_BLESSED
		1865	get_relaxed = F_RELAXED
		1866	get_allow_unknown = F_ALLOW_UNKNOWN
		1867	PPCODE:
		1868	XPUSHs (boolSV (self->flags & ix));
		1869
		1870	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		1871	PPCODE:
		1872	self->max_depth = max_depth;
		1873	XPUSHs (ST (0));
		1874
		1875	U32 get_max_depth (JSON *self)
		1876	CODE:
		1877	RETVAL = self->max_depth;
869	OUTPUT:	1878	OUTPUT:
870	RETVAL	1879	RETVAL
871		1880
872	SV *ascii (SV *self, int enable)	1881	void max_size (JSON *self, U32 max_size = 0)
873	ALIAS:	1882	PPCODE:
874	ascii = F_ASCII	1883	self->max_size = max_size;
875	utf8 = F_UTF8	1884	XPUSHs (ST (0));
876	indent = F_INDENT	1885
877	canonical = F_CANONICAL	1886	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:	1887	CODE:
884	{	1888	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:	1889	OUTPUT:
894	RETVAL	1890	RETVAL
895		1891
896	void encode (SV *self, SV *scalar)	1892	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
897	PPCODE:	1893	PPCODE:
898	XPUSHs (encode_json (scalar, *SvJSON (self)));	1894	{
		1895	SvREFCNT_dec (self->cb_object);
		1896	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
899		1897
900	void decode (SV *self, SV *jsonstr)	1898	XPUSHs (ST (0));
		1899	}
		1900
		1901	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
901	PPCODE:	1902	PPCODE:
902	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1903	{
		1904	if (!self->cb_sk_object)
		1905	self->cb_sk_object = newHV ();
		1906
		1907	if (SvOK (cb))
		1908	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1909	else
		1910	{
		1911	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1912
		1913	if (!HvKEYS (self->cb_sk_object))
		1914	{
		1915	SvREFCNT_dec (self->cb_sk_object);
		1916	self->cb_sk_object = 0;
		1917	}
		1918	}
		1919
		1920	XPUSHs (ST (0));
		1921	}
		1922
		1923	void encode (JSON *self, SV *scalar)
		1924	PPCODE:
		1925	PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
		1926	XPUSHs (scalar);
		1927
		1928	void decode (JSON *self, SV *jsonstr)
		1929	PPCODE:
		1930	PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
		1931	XPUSHs (jsonstr);
		1932
		1933	void decode_prefix (JSON *self, SV *jsonstr)
		1934	PPCODE:
		1935	{
		1936	SV *sv;
		1937	char *offset;
		1938	PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
		1939	EXTEND (SP, 2);
		1940	PUSHs (sv);
		1941	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		1942	}
		1943
		1944	void incr_parse (JSON *self, SV *jsonstr = 0)
		1945	PPCODE:
		1946	{
		1947	if (!self->incr_text)
		1948	self->incr_text = newSVpvn ("", 0);
		1949
		1950	/* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
		1951	if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
		1952	if (DECODE_WANTS_OCTETS (self))
		1953	{
		1954	if (self->incr_pos)
		1955	self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
		1956	(U8 *)SvPVX (self->incr_text) + self->incr_pos);
		1957
		1958	sv_utf8_downgrade (self->incr_text, 0);
		1959	}
		1960	else
		1961	{
		1962	sv_utf8_upgrade (self->incr_text);
		1963
		1964	if (self->incr_pos)
		1965	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1966	- (U8 *)SvPVX (self->incr_text);
		1967	}
		1968
		1969	// append data, if any
		1970	if (jsonstr)
		1971	{
		1972	/* make sure both strings have same encoding */
		1973	if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
		1974	if (SvUTF8 (jsonstr))
		1975	sv_utf8_downgrade (jsonstr, 0);
		1976	else
		1977	sv_utf8_upgrade (jsonstr);
		1978
		1979	/* and then just blindly append */
		1980	{
		1981	STRLEN len;
		1982	const char *str = SvPV (jsonstr, len);
		1983	STRLEN cur = SvCUR (self->incr_text);
		1984
		1985	if (SvLEN (self->incr_text) <= cur + len)
		1986	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1987
		1988	Move (str, SvEND (self->incr_text), len, char);
		1989	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1990	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1991	}
		1992	}
		1993
		1994	if (GIMME_V != G_VOID)
		1995	do
		1996	{
		1997	SV *sv;
		1998	char *offset;
		1999
		2000	if (!INCR_DONE (self))
		2001	{
		2002	incr_parse (self);
		2003
		2004	if (self->incr_pos > self->max_size && self->max_size)
		2005	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		2006	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		2007
		2008	if (!INCR_DONE (self))
		2009	{
		2010	// as an optimisation, do not accumulate white space in the incr buffer
		2011	if (self->incr_mode == INCR_M_WS && self->incr_pos)
		2012	{
		2013	self->incr_pos = 0;
		2014	SvCUR_set (self->incr_text, 0);
		2015	}
		2016
		2017	break;
		2018	}
		2019	}
		2020
		2021	PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
		2022	XPUSHs (sv);
		2023
		2024	self->incr_pos -= offset - SvPVX (self->incr_text);
		2025	self->incr_nest = 0;
		2026	self->incr_mode = 0;
		2027
		2028	sv_chop (self->incr_text, offset);
		2029	}
		2030	while (GIMME_V == G_ARRAY);
		2031	}
		2032
		2033	SV *incr_text (JSON *self)
		2034	ATTRS: lvalue
		2035	CODE:
		2036	{
		2037	if (self->incr_pos)
		2038	croak ("incr_text can not be called when the incremental parser already started parsing");
		2039
		2040	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		2041	}
		2042	OUTPUT:
		2043	RETVAL
		2044
		2045	void incr_skip (JSON *self)
		2046	CODE:
		2047	{
		2048	if (self->incr_pos)
		2049	{
		2050	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2051	self->incr_pos = 0;
		2052	self->incr_nest = 0;
		2053	self->incr_mode = 0;
		2054	}
		2055	}
		2056
		2057	void incr_reset (JSON *self)
		2058	CODE:
		2059	{
		2060	SvREFCNT_dec (self->incr_text);
		2061	self->incr_text = 0;
		2062	self->incr_pos = 0;
		2063	self->incr_nest = 0;
		2064	self->incr_mode = 0;
		2065	}
		2066
		2067	void DESTROY (JSON *self)
		2068	CODE:
		2069	SvREFCNT_dec (self->cb_sk_object);
		2070	SvREFCNT_dec (self->cb_object);
		2071	SvREFCNT_dec (self->incr_text);
903		2072
904	PROTOTYPES: ENABLE	2073	PROTOTYPES: ENABLE
905		2074
906	void to_json (SV *scalar)	2075	void encode_json (SV *scalar)
		2076	ALIAS:
		2077	to_json_ = 0
		2078	encode_json = F_UTF8
907	PPCODE:	2079	PPCODE:
908	XPUSHs (encode_json (scalar, F_UTF8));	2080	{
		2081	JSON json;
		2082	json_init (&json);
		2083	json.flags |= ix;
		2084	PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
		2085	XPUSHs (scalar);
		2086	}
909		2087
910	void from_json (SV *jsonstr)	2088	void decode_json (SV *jsonstr)
		2089	ALIAS:
		2090	from_json_ = 0
		2091	decode_json = F_UTF8
911	PPCODE:	2092	PPCODE:
912	XPUSHs (decode_json (jsonstr, F_UTF8));	2093	{
		2094	JSON json;
		2095	json_init (&json);
		2096	json.flags |= ix;
		2097	PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
		2098	XPUSHs (jsonstr);
		2099	}
913		2100

Diff Legend

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