ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.10 by root, Sat Mar 24 01:15:22 2007 UTC vs.
Revision 1.110 by root, Wed Jun 1 13:01:09 2011 UTC

1	#include "EXTERN.h"	1	#include "EXTERN.h"
2	#include "perl.h"	2	#include "perl.h"
3	#include "XSUB.h"	3	#include "XSUB.h"
4		4
5	#include "assert.h"	5	#include <assert.h>
6	#include "string.h"	6	#include <string.h>
7	#include "stdlib.h"	7	#include <stdlib.h>
		8	#include <stdio.h>
		9	#include <limits.h>
		10	#include <float.h>
8		11
		12	#if defined(__BORLANDC__) || defined(_MSC_VER)
		13	# define snprintf _snprintf // C compilers have this in stdio.h
		14	#endif
		15
		16	// some old perls do not have this, try to make it work, no
		17	// guarantees, though. if it breaks, you get to keep the pieces.
		18	#ifndef UTF8_MAXBYTES
		19	# define UTF8_MAXBYTES 13
		20	#endif
		21
		22	// three extra for rounding, sign, and end of string
		23	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
		24
9	#define F_ASCII 0x00000001	25	#define F_ASCII 0x00000001UL
		26	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	27	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	28	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	29	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	30	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	31	#define F_SPACE_AFTER 0x00000040UL
15	#define F_ALLOW_NONREF 0x00000080	32	#define F_ALLOW_NONREF 0x00000100UL
16	#define F_SHRINK 0x00000100	33	#define F_SHRINK 0x00000200UL
		34	#define F_ALLOW_BLESSED 0x00000400UL
		35	#define F_CONV_BLESSED 0x00000800UL
		36	#define F_RELAXED 0x00001000UL
		37	#define F_ALLOW_UNKNOWN 0x00002000UL
		38	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
17		39
18	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	40	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
19	#define F_DEFAULT 0
20		41
21	#define INIT_SIZE 32 // initial scalar size to be allocated	42	#define INIT_SIZE 32 // initial scalar size to be allocated
		43	#define INDENT_STEP 3 // spaces per indentation level
		44
		45	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
		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	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		78	static SV *json_true, *json_false;
		79
		80	enum {
		81	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		82	INCR_M_STR, // inside string
		83	INCR_M_BS, // inside backslash
		84	INCR_M_C0, // inside comment in initial whitespace sequence
		85	INCR_M_C1, // inside comment in other places
		86	INCR_M_JSON // outside anything, count nesting
		87	};
		88
		89	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		90
		91	typedef struct {
		92	U32 flags;
		93	U32 max_depth;
		94	STRLEN max_size;
		95
		96	SV *cb_object;
		97	HV *cb_sk_object;
		98
		99	// for the incremental parser
		100	SV *incr_text; // the source text so far
		101	STRLEN incr_pos; // the current offset into the text
		102	int incr_nest; // {[]}-nesting level
		103	unsigned char incr_mode;
		104	} JSON;
		105
		106	INLINE void
		107	json_init (JSON *json)
		108	{
		109	Zero (json, 1, JSON);
		110	json->max_depth = 512;
		111	}
		112
		113	/////////////////////////////////////////////////////////////////////////////
		114	// utility functions
		115
		116	INLINE SV *
		117	get_bool (const char *name)
		118	{
		119	SV *sv = get_sv (name, 1);
		120
		121	SvREADONLY_on (sv);
		122	SvREADONLY_on (SvRV (sv));
		123
		124	return sv;
		125	}
		126
		127	INLINE void
		128	shrink (SV *sv)
		129	{
		130	sv_utf8_downgrade (sv, 1);
		131
		132	if (SvLEN (sv) > SvCUR (sv) + 1)
		133	{
		134	#ifdef SvPV_shrink_to_cur
		135	SvPV_shrink_to_cur (sv);
		136	#elif defined (SvPV_renew)
		137	SvPV_renew (sv, SvCUR (sv) + 1);
		138	#endif
		139	}
		140	}
		141
		142	// decode an utf-8 character and return it, or (UV)-1 in
		143	// case of an error.
		144	// we special-case "safe" characters from U+80 .. U+7FF,
		145	// but use the very good perl function to parse anything else.
		146	// note that we never call this function for a ascii codepoints
		147	INLINE UV
		148	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		149	{
		150	if (expect_true (len >= 2
		151	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
		152	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
		153	{
		154	*clen = 2;
		155	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		156	}
		157	else
		158	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		159	}
		160
		161	// likewise for encoding, also never called for ascii codepoints
		162	// this function takes advantage of this fact, although current gccs
		163	// seem to optimise the check for >= 0x80 away anyways
		164	INLINE unsigned char *
		165	encode_utf8 (unsigned char *s, UV ch)
		166	{
		167	if (expect_false (ch < 0x000080))
		168	*s++ = ch;
		169	else if (expect_true (ch < 0x000800))
		170	*s++ = 0xc0 | ( ch >> 6),
		171	*s++ = 0x80 | ( ch & 0x3f);
		172	else if ( ch < 0x010000)
		173	*s++ = 0xe0 | ( ch >> 12),
		174	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		175	*s++ = 0x80 | ( ch & 0x3f);
		176	else if ( ch < 0x110000)
		177	*s++ = 0xf0 | ( ch >> 18),
		178	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		179	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		180	*s++ = 0x80 | ( ch & 0x3f);
		181
		182	return s;
		183	}
		184
		185	// convert offset pointer to character index, sv must be string
		186	static STRLEN
		187	ptr_to_index (SV *sv, char *offset)
		188	{
		189	return SvUTF8 (sv)
		190	? utf8_distance (offset, SvPVX (sv))
		191	: offset - SvPVX (sv);
		192	}
		193
		194	/////////////////////////////////////////////////////////////////////////////
		195	// fp hell
		196
		197	// scan a group of digits, and a trailing exponent
		198	static void
		199	json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
		200	{
		201	UV uaccum = 0;
		202	int eaccum = 0;
		203
		204	// if we recurse too deep, skip all remaining digits
		205	// to avoid a stack overflow attack
		206	if (expect_false (--maxdepth <= 0))
		207	while (((U8)*s - '0') < 10)
		208	++s;
		209
		210	for (;;)
		211	{
		212	U8 dig = (U8)*s - '0';
		213
		214	if (expect_false (dig >= 10))
		215	{
		216	if (dig == (U8)((U8)'.' - (U8)'0'))
		217	{
		218	++s;
		219	json_atof_scan1 (s, accum, expo, 1, maxdepth);
		220	}
		221	else if ((dig | ' ') == 'e' - '0')
		222	{
		223	int exp2 = 0;
		224	int neg = 0;
		225
		226	++s;
		227
		228	if (*s == '-')
		229	{
		230	++s;
		231	neg = 1;
		232	}
		233	else if (*s == '+')
		234	++s;
		235
		236	while ((dig = (U8)*s - '0') < 10)
		237	exp2 = exp2 * 10 + *s++ - '0';
		238
		239	*expo += neg ? -exp2 : exp2;
		240	}
		241
		242	break;
		243	}
		244
		245	++s;
		246
		247	uaccum = uaccum * 10 + dig;
		248	++eaccum;
		249
		250	// if we have too many digits, then recurse for more
		251	// we actually do this for rather few digits
		252	if (uaccum >= (UV_MAX - 9) / 10)
		253	{
		254	if (postdp) *expo -= eaccum;
		255	json_atof_scan1 (s, accum, expo, postdp, maxdepth);
		256	if (postdp) *expo += eaccum;
		257
		258	break;
		259	}
		260	}
		261
		262	// this relies greatly on the quality of the pow ()
		263	// implementation of the platform, but a good
		264	// implementation is hard to beat.
		265	// (IEEE 754 conformant ones are required to be exact)
		266	if (postdp) *expo -= eaccum;
		267	*accum += uaccum * Perl_pow (10., *expo);
		268	*expo += eaccum;
		269	}
		270
		271	static NV
		272	json_atof (const char *s)
		273	{
		274	NV accum = 0.;
		275	int expo = 0;
		276	int neg = 0;
		277
		278	if (*s == '-')
		279	{
		280	++s;
		281	neg = 1;
		282	}
		283
		284	// a recursion depth of ten gives us >>500 bits
		285	json_atof_scan1 (s, &accum, &expo, 0, 10);
		286
		287	return neg ? -accum : accum;
		288	}
		289	/////////////////////////////////////////////////////////////////////////////
		290	// encoder
27		291
28	// structure used for encoding JSON	292	// structure used for encoding JSON
29	typedef struct	293	typedef struct
30	{	294	{
31	char *cur;	295	char *cur; // SvPVX (sv) + current output position
32	STRLEN len; // SvLEN (sv)
33	char *end; // SvEND (sv)	296	char *end; // SvEND (sv)
34	SV *sv;	297	SV *sv; // result scalar
35	UV flags;	298	JSON json;
36	int max_recurse;	299	U32 indent; // indentation level
37	int indent;	300	UV limit; // escape character values >= this value when encoding
38	} enc_t;	301	} enc_t;
39		302
40	// structure used for decoding JSON	303	INLINE void
41	typedef struct
42	{
43	char *cur;
44	char *end;
45	const char *err;
46	UV flags;
47	} dec_t;
48
49	static UV *
50	SvJSON (SV *sv)
51	{
52	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
53	croak ("object is not of type JSON::XS");
54
55	return &SvUVX (SvRV (sv));
56	}
57
58	static void
59	shrink (SV *sv)
60	{
61	sv_utf8_downgrade (sv, 1);
62	#ifdef SvPV_shrink_to_cur
63	SvPV_shrink_to_cur (sv);
64	#endif
65	}
66
67	/////////////////////////////////////////////////////////////////////////////
68
69	static void
70	need (enc_t *enc, STRLEN len)	304	need (enc_t *enc, STRLEN len)
71	{	305	{
72	if (enc->cur + len >= enc->end)	306	if (expect_false (enc->cur + len >= enc->end))
73	{	307	{
74	STRLEN cur = enc->cur - SvPVX (enc->sv);	308	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
75	SvGROW (enc->sv, cur + len + 1);	309	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
76	enc->cur = SvPVX (enc->sv) + cur;	310	enc->cur = SvPVX (enc->sv) + cur;
77	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	311	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
78	}	312	}
79	}	313	}
80		314
81	static void	315	INLINE void
82	encode_ch (enc_t *enc, char ch)	316	encode_ch (enc_t *enc, char ch)
83	{	317	{
84	need (enc, 1);	318	need (enc, 1);
85	*enc->cur++ = ch;	319	*enc->cur++ = ch;
86	}	320	}
…		…
94		328
95	while (str < end)	329	while (str < end)
96	{	330	{
97	unsigned char ch = *(unsigned char *)str;	331	unsigned char ch = *(unsigned char *)str;
98		332
99	if (ch >= 0x20 && ch < 0x80) // most common case	333	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
100	{	334	{
101	if (ch == '"') // but with slow exceptions	335	if (expect_false (ch == '"')) // but with slow exceptions
102	{	336	{
103	need (enc, len += 1);	337	need (enc, len += 1);
104	*enc->cur++ = '\\';	338	*enc->cur++ = '\\';
105	*enc->cur++ = '"';	339	*enc->cur++ = '"';
106	}	340	}
107	else if (ch == '\\')	341	else if (expect_false (ch == '\\'))
108	{	342	{
109	need (enc, len += 1);	343	need (enc, len += 1);
110	*enc->cur++ = '\\';	344	*enc->cur++ = '\\';
111	*enc->cur++ = '\\';	345	*enc->cur++ = '\\';
112	}	346	}
…		…
130	STRLEN clen;	364	STRLEN clen;
131	UV uch;	365	UV uch;
132		366
133	if (is_utf8)	367	if (is_utf8)
134	{	368	{
135	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	369	uch = decode_utf8 (str, end - str, &clen);
136	if (clen == (STRLEN)-1)	370	if (clen == (STRLEN)-1)
137	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	371	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
138	}	372	}
139	else	373	else
140	{	374	{
141	uch = ch;	375	uch = ch;
142	clen = 1;	376	clen = 1;
143	}	377	}
144		378
145	if (uch > 0x10FFFFUL)	379	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
146	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
147
148	if (uch < 0x80 || enc->flags & F_ASCII)
149	{	380	{
150	if (uch > 0xFFFFUL)	381	if (uch >= 0x10000UL)
151	{	382	{
		383	if (uch >= 0x110000UL)
		384	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		385
152	need (enc, len += 11);	386	need (enc, len += 11);
153	sprintf (enc->cur, "\\u%04x\\u%04x",	387	sprintf (enc->cur, "\\u%04x\\u%04x",
154	(int)((uch - 0x10000) / 0x400 + 0xD800),	388	(int)((uch - 0x10000) / 0x400 + 0xD800),
155	(int)((uch - 0x10000) % 0x400 + 0xDC00));	389	(int)((uch - 0x10000) % 0x400 + 0xDC00));
156	enc->cur += 12;	390	enc->cur += 12;
157	}	391	}
158	else	392	else
159	{	393	{
160	static char hexdigit [16] = "0123456789abcdef";
161	need (enc, len += 5);	394	need (enc, len += 5);
162	*enc->cur++ = '\\';	395	*enc->cur++ = '\\';
163	*enc->cur++ = 'u';	396	*enc->cur++ = 'u';
164	*enc->cur++ = hexdigit [ uch >> 12 ];	397	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
165	*enc->cur++ = hexdigit [(uch >> 8) & 15];	398	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
166	*enc->cur++ = hexdigit [(uch >> 4) & 15];	399	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
167	*enc->cur++ = hexdigit [(uch >> 0) & 15];	400	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
168	}	401	}
169		402
		403	str += clen;
		404	}
		405	else if (enc->json.flags & F_LATIN1)
		406	{
		407	*enc->cur++ = uch;
170	str += clen;	408	str += clen;
171	}	409	}
172	else if (is_utf8)	410	else if (is_utf8)
173	{	411	{
174	need (enc, len += clen);	412	need (enc, len += clen);
…		…
178	}	416	}
179	while (--clen);	417	while (--clen);
180	}	418	}
181	else	419	else
182	{	420	{
183	need (enc, len += 10); // never more than 11 bytes needed	421	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
184	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	422	enc->cur = encode_utf8 (enc->cur, uch);
185	++str;	423	++str;
186	}	424	}
187	}	425	}
188	}	426	}
189	}	427	}
190		428
191	--len;	429	--len;
192	}	430	}
193	}	431	}
194		432
195	#define INDENT SB \	433	INLINE void
		434	encode_indent (enc_t *enc)
		435	{
196	if (enc->flags & F_INDENT) \	436	if (enc->json.flags & F_INDENT)
197	{ \	437	{
198	int i_; \	438	int spaces = enc->indent * INDENT_STEP;
199	need (enc, enc->indent); \	439
200	for (i_ = enc->indent * 3; i_--; )\	440	need (enc, spaces);
		441	memset (enc->cur, ' ', spaces);
		442	enc->cur += spaces;
		443	}
		444	}
		445
		446	INLINE void
		447	encode_space (enc_t *enc)
		448	{
		449	need (enc, 1);
		450	encode_ch (enc, ' ');
		451	}
		452
		453	INLINE void
		454	encode_nl (enc_t *enc)
		455	{
		456	if (enc->json.flags & F_INDENT)
		457	{
		458	need (enc, 1);
201	encode_ch (enc, ' '); \	459	encode_ch (enc, '\n');
202	} \	460	}
203	SE	461	}
204		462
205	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	463	INLINE void
206	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	464	encode_comma (enc_t *enc)
207	#define COMMA SB \	465	{
208	encode_ch (enc, ','); \	466	encode_ch (enc, ',');
		467
209	if (enc->flags & F_INDENT) \	468	if (enc->json.flags & F_INDENT)
210	NL; \	469	encode_nl (enc);
211	else if (enc->flags & F_SPACE_AFTER) \	470	else if (enc->json.flags & F_SPACE_AFTER)
212	SPACE; \	471	encode_space (enc);
213	SE	472	}
214		473
215	static void encode_sv (enc_t *enc, SV *sv);	474	static void encode_sv (enc_t *enc, SV *sv);
216		475
217	static void	476	static void
218	encode_av (enc_t *enc, AV *av)	477	encode_av (enc_t *enc, AV *av)
219	{	478	{
220	int i, len = av_len (av);	479	int i, len = av_len (av);
221		480
		481	if (enc->indent >= enc->json.max_depth)
		482	croak (ERR_NESTING_EXCEEDED);
		483
222	encode_ch (enc, '['); NL;	484	encode_ch (enc, '[');
223	++enc->indent;	485
		486	if (len >= 0)
		487	{
		488	encode_nl (enc); ++enc->indent;
224		489
225	for (i = 0; i <= len; ++i)	490	for (i = 0; i <= len; ++i)
226	{	491	{
227	INDENT;	492	SV **svp = av_fetch (av, i, 0);
228	encode_sv (enc, *av_fetch (av, i, 0));
229		493
		494	encode_indent (enc);
		495
		496	if (svp)
		497	encode_sv (enc, *svp);
		498	else
		499	encode_str (enc, "null", 4, 0);
		500
230	if (i < len)	501	if (i < len)
231	COMMA;	502	encode_comma (enc);
232	}	503	}
233		504
234	NL;	505	encode_nl (enc); --enc->indent; encode_indent (enc);
235		506	}
236	--enc->indent;	507
237	INDENT; encode_ch (enc, ']');	508	encode_ch (enc, ']');
238	}	509	}
239		510
240	static void	511	static void
241	encode_he (enc_t *enc, HE *he)	512	encode_hk (enc_t *enc, HE *he)
242	{	513	{
243	encode_ch (enc, '"');	514	encode_ch (enc, '"');
244		515
245	if (HeKLEN (he) == HEf_SVKEY)	516	if (HeKLEN (he) == HEf_SVKEY)
246	{	517	{
…		…
256	else	527	else
257	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	528	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
258		529
259	encode_ch (enc, '"');	530	encode_ch (enc, '"');
260		531
261	if (enc->flags & F_SPACE_BEFORE) SPACE;	532	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
262	encode_ch (enc, ':');	533	encode_ch (enc, ':');
263	if (enc->flags & F_SPACE_AFTER ) SPACE;	534	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
264	encode_sv (enc, HeVAL (he));
265	}	535	}
266		536
267	// compare hash entries, used when all keys are bytestrings	537	// compare hash entries, used when all keys are bytestrings
268	static int	538	static int
269	he_cmp_fast (const void *a_, const void *b_)	539	he_cmp_fast (const void *a_, const void *b_)
…		…
274	HE *b = *(HE **)b_;	544	HE *b = *(HE **)b_;
275		545
276	STRLEN la = HeKLEN (a);	546	STRLEN la = HeKLEN (a);
277	STRLEN lb = HeKLEN (b);	547	STRLEN lb = HeKLEN (b);
278		548
279	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	549	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
280	cmp = la < lb ? -1 : la == lb ? 0 : 1;	550	cmp = lb - la;
281		551
282	return cmp;	552	return cmp;
283	}	553	}
284		554
285	// compare hash entries, used when some keys are sv's or utf-x	555	// compare hash entries, used when some keys are sv's or utf-x
286	static int	556	static int
287	he_cmp_slow (const void *a, const void *b)	557	he_cmp_slow (const void *a, const void *b)
288	{	558	{
289	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	559	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
290	}	560	}
291		561
292	static void	562	static void
293	encode_hv (enc_t *enc, HV *hv)	563	encode_hv (enc_t *enc, HV *hv)
294	{	564	{
295	int count, i;	565	HE *he;
296		566
297	encode_ch (enc, '{'); NL; ++enc->indent;	567	if (enc->indent >= enc->json.max_depth)
		568	croak (ERR_NESTING_EXCEEDED);
298		569
299	if ((count = hv_iterinit (hv)))	570	encode_ch (enc, '{');
300	{	571
301	// for canonical output we have to sort by keys first	572	// for canonical output we have to sort by keys first
302	// actually, this is mostly due to the stupid so-called	573	// actually, this is mostly due to the stupid so-called
303	// security workaround added somewhere in 5.8.x.	574	// security workaround added somewhere in 5.8.x
304	// that randomises hash orderings	575	// that randomises hash orderings
305	if (enc->flags & F_CANONICAL)	576	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
		577	{
		578	int count = hv_iterinit (hv);
		579
		580	if (SvMAGICAL (hv))
306	{	581	{
307	HE *he, *hes [count];	582	// need to count by iterating. could improve by dynamically building the vector below
		583	// but I don't care for the speed of this special case.
		584	// note also that we will run into undefined behaviour when the two iterations
		585	// do not result in the same count, something I might care for in some later release.
		586
		587	count = 0;
		588	while (hv_iternext (hv))
		589	++count;
		590
		591	hv_iterinit (hv);
		592	}
		593
		594	if (count)
		595	{
308	int fast = 1;	596	int i, fast = 1;
		597	#if defined(__BORLANDC__) || defined(_MSC_VER)
		598	HE **hes = _alloca (count * sizeof (HE));
		599	#else
		600	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		601	#endif
309		602
310	i = 0;	603	i = 0;
311	while ((he = hv_iternext (hv)))	604	while ((he = hv_iternext (hv)))
312	{	605	{
313	hes [i++] = he;	606	hes [i++] = he;
…		…
335		628
336	FREETMPS;	629	FREETMPS;
337	LEAVE;	630	LEAVE;
338	}	631	}
339		632
340	for (i = 0; i < count; ++i)	633	encode_nl (enc); ++enc->indent;
		634
		635	while (count--)
		636	{
		637	encode_indent (enc);
		638	he = hes [count];
		639	encode_hk (enc, he);
		640	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		641
		642	if (count)
		643	encode_comma (enc);
		644	}
		645
		646	encode_nl (enc); --enc->indent; encode_indent (enc);
		647	}
		648	}
		649	else
		650	{
		651	if (hv_iterinit (hv) || SvMAGICAL (hv))
		652	if ((he = hv_iternext (hv)))
		653	{
		654	encode_nl (enc); ++enc->indent;
		655
		656	for (;;)
341	{	657	{
342	INDENT;	658	encode_indent (enc);
343	encode_he (enc, hes [i]);	659	encode_hk (enc, he);
		660	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
344		661
345	if (i < count - 1)	662	if (!(he = hv_iternext (hv)))
346	COMMA;	663	break;
		664
		665	encode_comma (enc);
347	}	666	}
348		667
		668	encode_nl (enc); --enc->indent; encode_indent (enc);
349	NL;	669	}
		670	}
		671
		672	encode_ch (enc, '}');
		673	}
		674
		675	// encode objects, arrays and special \0=false and \1=true values.
		676	static void
		677	encode_rv (enc_t *enc, SV *sv)
		678	{
		679	svtype svt;
		680
		681	SvGETMAGIC (sv);
		682	svt = SvTYPE (sv);
		683
		684	if (expect_false (SvOBJECT (sv)))
		685	{
		686	HV *stash = !JSON_SLOW || json_boolean_stash
		687	? json_boolean_stash
		688	: gv_stashpv ("JSON::XS::Boolean", 1);
		689
		690	if (SvSTASH (sv) == stash)
		691	{
		692	if (SvIV (sv))
		693	encode_str (enc, "true", 4, 0);
		694	else
		695	encode_str (enc, "false", 5, 0);
350	}	696	}
351	else	697	else
352	{	698	{
353	SV *sv;	699	#if 0
354	HE *he = hv_iternext (hv);	700	if (0 && sv_derived_from (rv, "JSON::Literal"))
355
356	for (;;)
357	{
358	INDENT;
359	encode_he (enc, he);
360
361	if (!(he = hv_iternext (hv)))
362	break;
363
364	COMMA;
365	}	701	{
366		702	// not yet
367	NL;	703	}
		704	#endif
		705	if (enc->json.flags & F_CONV_BLESSED)
		706	{
		707	// we re-bless the reference to get overload and other niceties right
		708	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
		709
		710	if (to_json)
		711	{
		712	dSP;
		713
		714	ENTER; SAVETMPS; PUSHMARK (SP);
		715	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		716
		717	// calling with G_SCALAR ensures that we always get a 1 return value
		718	PUTBACK;
		719	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		720	SPAGAIN;
		721
		722	// catch this surprisingly common error
		723	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		724	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		725
		726	sv = POPs;
		727	PUTBACK;
		728
		729	encode_sv (enc, sv);
		730
		731	FREETMPS; LEAVE;
		732	}
		733	else if (enc->json.flags & F_ALLOW_BLESSED)
		734	encode_str (enc, "null", 4, 0);
		735	else
		736	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		737	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		738	}
		739	else if (enc->json.flags & F_ALLOW_BLESSED)
		740	encode_str (enc, "null", 4, 0);
		741	else
		742	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		743	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
368	}	744	}
369	}	745	}
		746	else if (svt == SVt_PVHV)
		747	encode_hv (enc, (HV *)sv);
		748	else if (svt == SVt_PVAV)
		749	encode_av (enc, (AV *)sv);
		750	else if (svt < SVt_PVAV)
		751	{
		752	STRLEN len = 0;
		753	char *pv = svt ? SvPV (sv, len) : 0;
370		754
371	--enc->indent; INDENT; encode_ch (enc, '}');	755	if (len == 1 && *pv == '1')
		756	encode_str (enc, "true", 4, 0);
		757	else if (len == 1 && *pv == '0')
		758	encode_str (enc, "false", 5, 0);
		759	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		760	encode_str (enc, "null", 4, 0);
		761	else
		762	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		763	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		764	}
		765	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		766	encode_str (enc, "null", 4, 0);
		767	else
		768	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		769	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
372	}	770	}
373		771
374	static void	772	static void
375	encode_sv (enc_t *enc, SV *sv)	773	encode_sv (enc_t *enc, SV *sv)
376	{	774	{
…		…
384	encode_str (enc, str, len, SvUTF8 (sv));	782	encode_str (enc, str, len, SvUTF8 (sv));
385	encode_ch (enc, '"');	783	encode_ch (enc, '"');
386	}	784	}
387	else if (SvNOKp (sv))	785	else if (SvNOKp (sv))
388	{	786	{
		787	// trust that perl will do the right thing w.r.t. JSON syntax.
389	need (enc, NV_DIG + 32);	788	need (enc, NV_DIG + 32);
390	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	789	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
391	enc->cur += strlen (enc->cur);	790	enc->cur += strlen (enc->cur);
392	}	791	}
393	else if (SvIOKp (sv))	792	else if (SvIOKp (sv))
394	{	793	{
		794	// we assume we can always read an IV as a UV and vice versa
		795	// we assume two's complement
		796	// we assume no aliasing issues in the union
		797	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		798	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		799	{
		800	// optimise the "small number case"
		801	// code will likely be branchless and use only a single multiplication
		802	// works for numbers up to 59074
		803	I32 i = SvIVX (sv);
		804	U32 u;
		805	char digit, nz = 0;
		806
395	need (enc, 64);	807	need (enc, 6);
		808
		809	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		810	u = i < 0 ? -i : i;
		811
		812	// convert to 4.28 fixed-point representation
		813	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		814
		815	// now output digit by digit, each time masking out the integer part
		816	// and multiplying by 5 while moving the decimal point one to the right,
		817	// resulting in a net multiplication by 10.
		818	// we always write the digit to memory but conditionally increment
		819	// the pointer, to enable the use of conditional move instructions.
		820	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		821	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		822	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		823	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		824	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		825	}
		826	else
		827	{
		828	// large integer, use the (rather slow) snprintf way.
		829	need (enc, IVUV_MAXCHARS);
396	enc->cur +=	830	enc->cur +=
397	SvIsUV(sv)	831	SvIsUV(sv)
398	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	832	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
399	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	833	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		834	}
400	}	835	}
401	else if (SvROK (sv))	836	else if (SvROK (sv))
402	{	837	encode_rv (enc, SvRV (sv));
403	SV *rv = SvRV (sv);	838	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
404
405	if (!--enc->max_recurse)
406	croak ("data structure too deep (hit recursion limit)");
407
408	switch (SvTYPE (rv))
409	{
410	case SVt_PVAV: encode_av (enc, (AV *)rv); break;
411	case SVt_PVHV: encode_hv (enc, (HV *)rv); break;
412
413	default:
414	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
415	SvPV_nolen (sv));
416	}
417	}
418	else if (!SvOK (sv))
419	encode_str (enc, "null", 4, 0);	839	encode_str (enc, "null", 4, 0);
420	else	840	else
421	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	841	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
422	SvPV_nolen (sv), SvFLAGS (sv));	842	SvPV_nolen (sv), SvFLAGS (sv));
423	}	843	}
424		844
425	static SV *	845	static SV *
426	encode_json (SV *scalar, UV flags)	846	encode_json (SV *scalar, JSON *json)
427	{	847	{
		848	enc_t enc;
		849
428	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	850	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
429	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	851	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
430		852
431	enc_t enc;	853	enc.json = *json;
432	enc.flags = flags;
433	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	854	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
434	enc.cur = SvPVX (enc.sv);	855	enc.cur = SvPVX (enc.sv);
435	enc.end = SvEND (enc.sv);	856	enc.end = SvEND (enc.sv);
436	enc.max_recurse = 0;
437	enc.indent = 0;	857	enc.indent = 0;
		858	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		859	: enc.json.flags & F_LATIN1 ? 0x000100UL
		860	: 0x110000UL;
438		861
439	SvPOK_only (enc.sv);	862	SvPOK_only (enc.sv);
440	encode_sv (&enc, scalar);	863	encode_sv (&enc, scalar);
		864	encode_nl (&enc);
441		865
		866	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		867	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		868
442	if (!(flags & (F_ASCII | F_UTF8)))	869	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
443	SvUTF8_on (enc.sv);	870	SvUTF8_on (enc.sv);
444		871
445	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
446
447	if (enc.flags & F_SHRINK)	872	if (enc.json.flags & F_SHRINK)
448	shrink (enc.sv);	873	shrink (enc.sv);
449		874
450	return enc.sv;	875	return enc.sv;
451	}	876	}
452		877
453	/////////////////////////////////////////////////////////////////////////////	878	/////////////////////////////////////////////////////////////////////////////
		879	// decoder
454		880
455	#define WS \	881	// structure used for decoding JSON
		882	typedef struct
		883	{
		884	char *cur; // current parser pointer
		885	char *end; // end of input string
		886	const char *err; // parse error, if != 0
		887	JSON json;
		888	U32 depth; // recursion depth
		889	U32 maxdepth; // recursion depth limit
		890	} dec_t;
		891
		892	INLINE void
		893	decode_comment (dec_t *dec)
		894	{
		895	// only '#'-style comments allowed a.t.m.
		896
		897	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		898	++dec->cur;
		899	}
		900
		901	INLINE void
		902	decode_ws (dec_t *dec)
		903	{
456	for (;;) \	904	for (;;)
457	{ \	905	{
458	char ch = *dec->cur; \	906	char ch = *dec->cur;
		907
459	if (ch > 0x20 \	908	if (ch > 0x20)
		909	{
		910	if (expect_false (ch == '#'))
		911	{
		912	if (dec->json.flags & F_RELAXED)
		913	decode_comment (dec);
		914	else
		915	break;
		916	}
		917	else
		918	break;
		919	}
460	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	920	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
461	break; \	921	break; // parse error, but let higher level handle it, gives better error messages
		922
462	++dec->cur; \	923	++dec->cur;
463	}	924	}
		925	}
464		926
465	#define ERR(reason) SB dec->err = reason; goto fail; SE	927	#define ERR(reason) SB dec->err = reason; goto fail; SE
		928
466	#define EXPECT_CH(ch) SB \	929	#define EXPECT_CH(ch) SB \
467	if (*dec->cur != ch) \	930	if (*dec->cur != ch) \
468	ERR (# ch " expected"); \	931	ERR (# ch " expected"); \
469	++dec->cur; \	932	++dec->cur; \
470	SE	933	SE
471		934
		935	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
		936	#define DEC_DEC_DEPTH --dec->depth
		937
472	static SV *decode_sv (dec_t *dec);	938	static SV *decode_sv (dec_t *dec);
473		939
474	static signed char decode_hexdigit[256];	940	static signed char decode_hexdigit[256];
475		941
476	static UV	942	static UV
477	decode_4hex (dec_t *dec)	943	decode_4hex (dec_t *dec)
478	{	944	{
479	signed char d1, d2, d3, d4;	945	signed char d1, d2, d3, d4;
		946	unsigned char *cur = (unsigned char *)dec->cur;
480		947
481	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	948	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
482	if (d1 < 0) ERR ("four hexadecimal digits expected");	949	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
483	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	950	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
484	if (d2 < 0) ERR ("four hexadecimal digits expected");	951	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
485	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
486	if (d3 < 0) ERR ("four hexadecimal digits expected");
487	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
488	if (d4 < 0) ERR ("four hexadecimal digits expected");
489		952
490	dec->cur += 4;	953	dec->cur += 4;
491		954
492	return ((UV)d1) << 12	955	return ((UV)d1) << 12
493	| ((UV)d2) << 8	956	| ((UV)d2) << 8
…		…
496		959
497	fail:	960	fail:
498	return (UV)-1;	961	return (UV)-1;
499	}	962	}
500		963
501	#define APPEND_GROW(n) SB \
502	if (cur + (n) >= end) \
503	{ \
504	STRLEN ofs = cur - SvPVX (sv); \
505	SvGROW (sv, ofs + (n) + 1); \
506	cur = SvPVX (sv) + ofs; \
507	end = SvEND (sv); \
508	} \
509	SE
510
511	#define APPEND_CH(ch) SB \
512	APPEND_GROW (1); \
513	*cur++ = (ch); \
514	SE
515
516	static SV *	964	static SV *
517	decode_str (dec_t *dec)	965	decode_str (dec_t *dec)
518	{	966	{
519	SV *sv = NEWSV (0,2);	967	SV *sv = 0;
520	int utf8 = 0;	968	int utf8 = 0;
521	char *cur = SvPVX (sv);	969	char *dec_cur = dec->cur;
522	char *end = SvEND (sv);
523		970
524	for (;;)	971	do
525	{	972	{
526	unsigned char ch = *(unsigned char *)dec->cur;	973	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		974	char *cur = buf;
527		975
528	if (ch == '"')	976	do
529	break;
530	else if (ch == '\\')
531	{	977	{
532	switch (*++dec->cur)	978	unsigned char ch = *(unsigned char *)dec_cur++;
		979
		980	if (expect_false (ch == '"'))
		981	{
		982	--dec_cur;
		983	break;
533	{	984	}
534	case '\\':	985	else if (expect_false (ch == '\\'))
535	case '/':	986	{
536	case '"': APPEND_CH (*dec->cur++); break;	987	switch (*dec_cur)
537
538	case 'b': APPEND_CH ('\010'); ++dec->cur; break;
539	case 't': APPEND_CH ('\011'); ++dec->cur; break;
540	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
541	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
542	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
543
544	case 'u':
545	{	988	{
546	UV lo, hi;	989	case '\\':
547	++dec->cur;	990	case '/':
		991	case '"': *cur++ = *dec_cur++; break;
548		992
549	hi = decode_4hex (dec);	993	case 'b': ++dec_cur; *cur++ = '\010'; break;
550	if (hi == (UV)-1)	994	case 't': ++dec_cur; *cur++ = '\011'; break;
551	goto fail;	995	case 'n': ++dec_cur; *cur++ = '\012'; break;
		996	case 'f': ++dec_cur; *cur++ = '\014'; break;
		997	case 'r': ++dec_cur; *cur++ = '\015'; break;
552		998
553	// possibly a surrogate pair	999	case 'u':
554	if (hi >= 0xd800 && hi < 0xdc00)
555	{	1000	{
556	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	1001	UV lo, hi;
557	ERR ("missing low surrogate character in surrogate pair");	1002	++dec_cur;
558		1003
559	dec->cur += 2;	1004	dec->cur = dec_cur;
560
561	lo = decode_4hex (dec);	1005	hi = decode_4hex (dec);
		1006	dec_cur = dec->cur;
562	if (lo == (UV)-1)	1007	if (hi == (UV)-1)
563	goto fail;	1008	goto fail;
564		1009
		1010	// possibly a surrogate pair
		1011	if (hi >= 0xd800)
		1012	if (hi < 0xdc00)
		1013	{
		1014	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		1015	ERR ("missing low surrogate character in surrogate pair");
		1016
		1017	dec_cur += 2;
		1018
		1019	dec->cur = dec_cur;
		1020	lo = decode_4hex (dec);
		1021	dec_cur = dec->cur;
		1022	if (lo == (UV)-1)
		1023	goto fail;
		1024
565	if (lo < 0xdc00 || lo >= 0xe000)	1025	if (lo < 0xdc00 || lo >= 0xe000)
566	ERR ("surrogate pair expected");	1026	ERR ("surrogate pair expected");
567		1027
568	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	1028	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		1029	}
		1030	else if (hi < 0xe000)
		1031	ERR ("missing high surrogate character in surrogate pair");
		1032
		1033	if (hi >= 0x80)
		1034	{
		1035	utf8 = 1;
		1036
		1037	cur = encode_utf8 (cur, hi);
		1038	}
		1039	else
		1040	*cur++ = hi;
569	}	1041	}
570	else if (hi >= 0xdc00 && hi < 0xe000)
571	ERR ("missing high surrogate character in surrogate pair");
572
573	if (hi >= 0x80)
574	{	1042	break;
575	utf8 = 1;
576		1043
577	APPEND_GROW (4); // at most 4 bytes for 21 bits
578	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
579	}
580	else	1044	default:
581	APPEND_CH (hi);	1045	--dec_cur;
		1046	ERR ("illegal backslash escape sequence in string");
582	}	1047	}
583	break;	1048	}
		1049	else if (expect_true (ch >= 0x20 && ch < 0x80))
		1050	*cur++ = ch;
		1051	else if (ch >= 0x80)
		1052	{
		1053	STRLEN clen;
584		1054
585	default:
586	--dec->cur;	1055	--dec_cur;
587	ERR ("illegal backslash escape sequence in string");	1056
		1057	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		1058	if (clen == (STRLEN)-1)
		1059	ERR ("malformed UTF-8 character in JSON string");
		1060
		1061	do
		1062	*cur++ = *dec_cur++;
		1063	while (--clen);
		1064
		1065	utf8 = 1;
		1066	}
		1067	else
		1068	{
		1069	--dec_cur;
		1070
		1071	if (!ch)
		1072	ERR ("unexpected end of string while parsing JSON string");
		1073	else
		1074	ERR ("invalid character encountered while parsing JSON string");
588	}	1075	}
589	}	1076	}
590	else if (ch >= 0x20 && ch <= 0x7f)	1077	while (cur < buf + SHORT_STRING_LEN);
591	APPEND_CH (*dec->cur++);	1078
592	else if (ch >= 0x80)
593	{	1079	{
594	STRLEN clen;	1080	STRLEN len = cur - buf;
595	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
596	if (clen == (STRLEN)-1)
597	ERR ("malformed UTF-8 character in JSON string");
598		1081
599	APPEND_GROW (clen);	1082	if (sv)
600	do
601	{	1083	{
602	*cur++ = *dec->cur++;	1084	STRLEN cur = SvCUR (sv);
		1085
		1086	if (SvLEN (sv) <= cur + len)
		1087	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1088
		1089	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		1090	SvCUR_set (sv, SvCUR (sv) + len);
603	}	1091	}
604	while (--clen);
605
606	utf8 = 1;
607	}
608	else if (dec->cur == dec->end)
609	ERR ("unexpected end of string while parsing json string");
610	else	1092	else
611	ERR ("invalid character encountered");	1093	sv = newSVpvn (buf, len);
612	}	1094	}
		1095	}
		1096	while (*dec_cur != '"');
613		1097
614	++dec->cur;	1098	++dec_cur;
615		1099
616	SvCUR_set (sv, cur - SvPVX (sv));	1100	if (sv)
617		1101	{
618	SvPOK_only (sv);	1102	SvPOK_only (sv);
619	*SvEND (sv) = 0;	1103	*SvEND (sv) = 0;
620		1104
621	if (utf8)	1105	if (utf8)
622	SvUTF8_on (sv);	1106	SvUTF8_on (sv);
		1107	}
		1108	else
		1109	sv = newSVpvn ("", 0);
623		1110
624	if (dec->flags & F_SHRINK)	1111	dec->cur = dec_cur;
625	shrink (sv);
626
627	return sv;	1112	return sv;
628		1113
629	fail:	1114	fail:
630	SvREFCNT_dec (sv);	1115	dec->cur = dec_cur;
631	return 0;	1116	return 0;
632	}	1117	}
633		1118
634	static SV *	1119	static SV *
635	decode_num (dec_t *dec)	1120	decode_num (dec_t *dec)
…		…
693	is_nv = 1;	1178	is_nv = 1;
694	}	1179	}
695		1180
696	if (!is_nv)	1181	if (!is_nv)
697	{	1182	{
698	UV uv;	1183	int len = dec->cur - start;
699	int numtype = grok_number (start, dec->cur - start, &uv);	1184
700	if (numtype & IS_NUMBER_IN_UV)	1185	// special case the rather common 1..5-digit-int case
701	if (numtype & IS_NUMBER_NEG)	1186	if (*start == '-')
		1187	switch (len)
702	{	1188	{
703	if (uv < (UV)IV_MIN)	1189	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
704	return newSViv (-(IV)uv);	1190	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
		1191	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1192	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1193	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
705	}	1194	}
		1195	else
		1196	switch (len)
		1197	{
		1198	case 1: return newSViv ( start [0] - '0' * 1);
		1199	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1200	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1201	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1202	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1203	}
		1204
		1205	{
		1206	UV uv;
		1207	int numtype = grok_number (start, len, &uv);
		1208	if (numtype & IS_NUMBER_IN_UV)
		1209	if (numtype & IS_NUMBER_NEG)
		1210	{
		1211	if (uv < (UV)IV_MIN)
		1212	return newSViv (-(IV)uv);
		1213	}
706	else	1214	else
707	return newSVuv (uv);	1215	return newSVuv (uv);
708	}	1216	}
709		1217
		1218	len -= *start == '-' ? 1 : 0;
		1219
		1220	// does not fit into IV or UV, try NV
		1221	if (len <= NV_DIG)
		1222	// fits into NV without loss of precision
		1223	return newSVnv (json_atof (start));
		1224
		1225	// everything else fails, convert it to a string
		1226	return newSVpvn (start, dec->cur - start);
		1227	}
		1228
		1229	// loss of precision here
710	return newSVnv (Atof (start));	1230	return newSVnv (json_atof (start));
711		1231
712	fail:	1232	fail:
713	return 0;	1233	return 0;
714	}	1234	}
715		1235
716	static SV *	1236	static SV *
717	decode_av (dec_t *dec)	1237	decode_av (dec_t *dec)
718	{	1238	{
719	AV *av = newAV ();	1239	AV *av = newAV ();
720		1240
721	WS;	1241	DEC_INC_DEPTH;
		1242	decode_ws (dec);
		1243
722	if (*dec->cur == ']')	1244	if (*dec->cur == ']')
723	++dec->cur;	1245	++dec->cur;
724	else	1246	else
725	for (;;)	1247	for (;;)
726	{	1248	{
…		…
730	if (!value)	1252	if (!value)
731	goto fail;	1253	goto fail;
732		1254
733	av_push (av, value);	1255	av_push (av, value);
734		1256
735	WS;	1257	decode_ws (dec);
736		1258
737	if (*dec->cur == ']')	1259	if (*dec->cur == ']')
738	{	1260	{
739	++dec->cur;	1261	++dec->cur;
740	break;	1262	break;
…		…
742		1264
743	if (*dec->cur != ',')	1265	if (*dec->cur != ',')
744	ERR (", or ] expected while parsing array");	1266	ERR (", or ] expected while parsing array");
745		1267
746	++dec->cur;	1268	++dec->cur;
		1269
		1270	decode_ws (dec);
		1271
		1272	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1273	{
		1274	++dec->cur;
		1275	break;
		1276	}
747	}	1277	}
748		1278
		1279	DEC_DEC_DEPTH;
749	return newRV_noinc ((SV *)av);	1280	return newRV_noinc ((SV *)av);
750		1281
751	fail:	1282	fail:
752	SvREFCNT_dec (av);	1283	SvREFCNT_dec (av);
		1284	DEC_DEC_DEPTH;
753	return 0;	1285	return 0;
754	}	1286	}
755		1287
756	static SV *	1288	static SV *
757	decode_hv (dec_t *dec)	1289	decode_hv (dec_t *dec)
758	{	1290	{
		1291	SV *sv;
759	HV *hv = newHV ();	1292	HV *hv = newHV ();
760		1293
761	WS;	1294	DEC_INC_DEPTH;
		1295	decode_ws (dec);
		1296
762	if (*dec->cur == '}')	1297	if (*dec->cur == '}')
763	++dec->cur;	1298	++dec->cur;
764	else	1299	else
765	for (;;)	1300	for (;;)
766	{	1301	{
767	SV *key, *value;
768
769	WS; EXPECT_CH ('"');	1302	EXPECT_CH ('"');
770		1303
771	key = decode_str (dec);	1304	// heuristic: assume that
772	if (!key)	1305	// a) decode_str + hv_store_ent are abysmally slow.
773	goto fail;	1306	// b) most hash keys are short, simple ascii text.
		1307	// => try to "fast-match" such strings to avoid
		1308	// the overhead of decode_str + hv_store_ent.
		1309	{
		1310	SV *value;
		1311	char *p = dec->cur;
		1312	char *e = p + 24; // only try up to 24 bytes
774		1313
775	WS; EXPECT_CH (':');	1314	for (;;)
776
777	value = decode_sv (dec);
778	if (!value)
779	{	1315	{
		1316	// the >= 0x80 is false on most architectures
		1317	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1318	{
		1319	// slow path, back up and use decode_str
		1320	SV *key = decode_str (dec);
		1321	if (!key)
		1322	goto fail;
		1323
		1324	decode_ws (dec); EXPECT_CH (':');
		1325
		1326	decode_ws (dec);
		1327	value = decode_sv (dec);
		1328	if (!value)
		1329	{
		1330	SvREFCNT_dec (key);
		1331	goto fail;
		1332	}
		1333
		1334	hv_store_ent (hv, key, value, 0);
780	SvREFCNT_dec (key);	1335	SvREFCNT_dec (key);
		1336
		1337	break;
		1338	}
		1339	else if (*p == '"')
		1340	{
		1341	// fast path, got a simple key
		1342	char *key = dec->cur;
		1343	int len = p - key;
		1344	dec->cur = p + 1;
		1345
		1346	decode_ws (dec); EXPECT_CH (':');
		1347
		1348	decode_ws (dec);
		1349	value = decode_sv (dec);
		1350	if (!value)
781	goto fail;	1351	goto fail;
		1352
		1353	hv_store (hv, key, len, value, 0);
		1354
		1355	break;
		1356	}
		1357
		1358	++p;
782	}	1359	}
783
784	//TODO: optimise
785	hv_store_ent (hv, key, value, 0);
786
787	WS;	1360	}
		1361
		1362	decode_ws (dec);
788		1363
789	if (*dec->cur == '}')	1364	if (*dec->cur == '}')
790	{	1365	{
791	++dec->cur;	1366	++dec->cur;
792	break;	1367	break;
…		…
794		1369
795	if (*dec->cur != ',')	1370	if (*dec->cur != ',')
796	ERR (", or } expected while parsing object/hash");	1371	ERR (", or } expected while parsing object/hash");
797		1372
798	++dec->cur;	1373	++dec->cur;
		1374
		1375	decode_ws (dec);
		1376
		1377	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1378	{
		1379	++dec->cur;
		1380	break;
		1381	}
799	}	1382	}
800		1383
		1384	DEC_DEC_DEPTH;
801	return newRV_noinc ((SV *)hv);	1385	sv = newRV_noinc ((SV *)hv);
		1386
		1387	// check filter callbacks
		1388	if (dec->json.flags & F_HOOK)
		1389	{
		1390	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1391	{
		1392	HE *cb, *he;
		1393
		1394	hv_iterinit (hv);
		1395	he = hv_iternext (hv);
		1396	hv_iterinit (hv);
		1397
		1398	// the next line creates a mortal sv each time its called.
		1399	// might want to optimise this for common cases.
		1400	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1401
		1402	if (cb)
		1403	{
		1404	dSP;
		1405	int count;
		1406
		1407	ENTER; SAVETMPS; PUSHMARK (SP);
		1408	XPUSHs (HeVAL (he));
		1409	sv_2mortal (sv);
		1410
		1411	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1412
		1413	if (count == 1)
		1414	{
		1415	sv = newSVsv (POPs);
		1416	FREETMPS; LEAVE;
		1417	return sv;
		1418	}
		1419
		1420	SvREFCNT_inc (sv);
		1421	FREETMPS; LEAVE;
		1422	}
		1423	}
		1424
		1425	if (dec->json.cb_object)
		1426	{
		1427	dSP;
		1428	int count;
		1429
		1430	ENTER; SAVETMPS; PUSHMARK (SP);
		1431	XPUSHs (sv_2mortal (sv));
		1432
		1433	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1434
		1435	if (count == 1)
		1436	{
		1437	sv = newSVsv (POPs);
		1438	FREETMPS; LEAVE;
		1439	return sv;
		1440	}
		1441
		1442	SvREFCNT_inc (sv);
		1443	FREETMPS; LEAVE;
		1444	}
		1445	}
		1446
		1447	return sv;
802		1448
803	fail:	1449	fail:
804	SvREFCNT_dec (hv);	1450	SvREFCNT_dec (hv);
		1451	DEC_DEC_DEPTH;
805	return 0;	1452	return 0;
806	}	1453	}
807		1454
808	static SV *	1455	static SV *
809	decode_sv (dec_t *dec)	1456	decode_sv (dec_t *dec)
810	{	1457	{
811	WS;	1458	// the beauty of JSON: you need exactly one character lookahead
		1459	// to parse everything.
812	switch (*dec->cur)	1460	switch (*dec->cur)
813	{	1461	{
814	case '"': ++dec->cur; return decode_str (dec);	1462	case '"': ++dec->cur; return decode_str (dec);
815	case '[': ++dec->cur; return decode_av (dec);	1463	case '[': ++dec->cur; return decode_av (dec);
816	case '{': ++dec->cur; return decode_hv (dec);	1464	case '{': ++dec->cur; return decode_hv (dec);
817		1465
818	case '-':	1466	case '-':
819	case '0': case '1': case '2': case '3': case '4':	1467	case '0': case '1': case '2': case '3': case '4':
820	case '5': case '6': case '7': case '8': case '9':	1468	case '5': case '6': case '7': case '8': case '9':
821	return decode_num (dec);	1469	return decode_num (dec);
822		1470
823	case 't':	1471	case 't':
824	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1472	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
825	{	1473	{
826	dec->cur += 4;	1474	dec->cur += 4;
		1475	#if JSON_SLOW
		1476	json_true = get_bool ("JSON::XS::true");
		1477	#endif
827	return newSViv (1);	1478	return newSVsv (json_true);
828	}	1479	}
829	else	1480	else
830	ERR ("'true' expected");	1481	ERR ("'true' expected");
831		1482
832	break;	1483	break;
833		1484
834	case 'f':	1485	case 'f':
835	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1486	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
836	{	1487	{
837	dec->cur += 5;	1488	dec->cur += 5;
		1489	#if JSON_SLOW
		1490	json_false = get_bool ("JSON::XS::false");
		1491	#endif
838	return newSViv (0);	1492	return newSVsv (json_false);
839	}	1493	}
840	else	1494	else
841	ERR ("'false' expected");	1495	ERR ("'false' expected");
842		1496
843	break;	1497	break;
…		…
852	ERR ("'null' expected");	1506	ERR ("'null' expected");
853		1507
854	break;	1508	break;
855		1509
856	default:	1510	default:
857	ERR ("malformed json string, neither array, object, number, string or atom");	1511	ERR ("malformed JSON string, neither array, object, number, string or atom");
858	break;	1512	break;
859	}	1513	}
860		1514
861	fail:	1515	fail:
862	return 0;	1516	return 0;
863	}	1517	}
864		1518
865	static SV *	1519	static SV *
866	decode_json (SV *string, UV flags)	1520	decode_json (SV *string, JSON *json, char **offset_return)
867	{	1521	{
		1522	dec_t dec;
868	SV *sv;	1523	SV *sv;
869		1524
870	if (flags & F_UTF8)	1525	/* work around bugs in 5.10 where manipulating magic values
		1526	* will perl ignore the magic in subsequent accesses.
		1527	* also make a copy of non-PV values, to get them into a clean
		1528	* state (SvPV should do that, but it's buggy, see below).
		1529	*/
		1530	/*SvGETMAGIC (string);*/
		1531	if (SvMAGICAL (string) || !SvPOK (string))
		1532	string = sv_2mortal (newSVsv (string));
		1533
		1534	SvUPGRADE (string, SVt_PV);
		1535
		1536	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1537	* assertion with -DDEBUGGING, although SvCUR is documented to
		1538	* return the xpv_cur field which certainly exists after upgrading.
		1539	* according to nicholas clark, calling SvPOK fixes this.
		1540	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1541	* and hope for the best.
		1542	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1543	* assertion business is seriously broken, try yet another workaround
		1544	* for the broken -DDEBUGGING.
		1545	*/
		1546	{
		1547	#ifdef DEBUGGING
		1548	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1549	#else
		1550	STRLEN offset = SvCUR (string);
		1551	#endif
		1552
		1553	if (offset > json->max_size && json->max_size)
		1554	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1555	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1556	}
		1557
		1558	if (DECODE_WANTS_OCTETS (json))
871	sv_utf8_downgrade (string, 0);	1559	sv_utf8_downgrade (string, 0);
872	else	1560	else
873	sv_utf8_upgrade (string);	1561	sv_utf8_upgrade (string);
874		1562
875	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1563	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
876		1564
877	dec_t dec;	1565	dec.json = *json;
878	dec.flags = flags;
879	dec.cur = SvPVX (string);	1566	dec.cur = SvPVX (string);
880	dec.end = SvEND (string);	1567	dec.end = SvEND (string);
881	dec.err = 0;	1568	dec.err = 0;
		1569	dec.depth = 0;
882		1570
		1571	if (dec.json.cb_object || dec.json.cb_sk_object)
		1572	dec.json.flags |= F_HOOK;
		1573
		1574	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1575
		1576	decode_ws (&dec);
883	sv = decode_sv (&dec);	1577	sv = decode_sv (&dec);
884		1578
		1579	if (offset_return)
		1580	*offset_return = dec.cur;
		1581
		1582	if (!(offset_return || !sv))
		1583	{
		1584	// check for trailing garbage
		1585	decode_ws (&dec);
		1586
		1587	if (*dec.cur)
		1588	{
		1589	dec.err = "garbage after JSON object";
		1590	SvREFCNT_dec (sv);
		1591	sv = 0;
		1592	}
		1593	}
		1594
885	if (!sv)	1595	if (!sv)
886	{	1596	{
887	IV offset = dec.flags & F_UTF8
888	? dec.cur - SvPVX (string)
889	: utf8_distance (dec.cur, SvPVX (string));
890	SV *uni = sv_newmortal ();	1597	SV *uni = sv_newmortal ();
891		1598
892	// horrible hack to silence warning inside pv_uni_display	1599	// horrible hack to silence warning inside pv_uni_display
893	COP cop = *PL_curcop;	1600	COP cop = *PL_curcop;
894	cop.cop_warnings = pWARN_NONE;	1601	cop.cop_warnings = pWARN_NONE;
…		…
896	SAVEVPTR (PL_curcop);	1603	SAVEVPTR (PL_curcop);
897	PL_curcop = &cop;	1604	PL_curcop = &cop;
898	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1605	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
899	LEAVE;	1606	LEAVE;
900		1607
901	croak ("%s, at character offset %d (%s)",	1608	croak ("%s, at character offset %d (before \"%s\")",
902	dec.err,	1609	dec.err,
903	(int)offset,	1610	ptr_to_index (string, dec.cur),
904	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1611	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
905	}	1612	}
906		1613
907	sv = sv_2mortal (sv);	1614	sv = sv_2mortal (sv);
908		1615
909	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1616	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
910	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1617	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
911		1618
912	return sv;	1619	return sv;
913	}	1620	}
914		1621
		1622	/////////////////////////////////////////////////////////////////////////////
		1623	// incremental parser
		1624
		1625	static void
		1626	incr_parse (JSON *self)
		1627	{
		1628	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1629
		1630	// the state machine here is a bit convoluted and could be simplified a lot
		1631	// but this would make it slower, so...
		1632
		1633	for (;;)
		1634	{
		1635	//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
		1636	switch (self->incr_mode)
		1637	{
		1638	// only used for initial whitespace skipping
		1639	case INCR_M_WS:
		1640	for (;;)
		1641	{
		1642	if (*p > 0x20)
		1643	{
		1644	if (*p == '#')
		1645	{
		1646	self->incr_mode = INCR_M_C0;
		1647	goto incr_m_c;
		1648	}
		1649	else
		1650	{
		1651	self->incr_mode = INCR_M_JSON;
		1652	goto incr_m_json;
		1653	}
		1654	}
		1655	else if (!*p)
		1656	goto interrupt;
		1657
		1658	++p;
		1659	}
		1660
		1661	// skip a single char inside a string (for \\-processing)
		1662	case INCR_M_BS:
		1663	if (!*p)
		1664	goto interrupt;
		1665
		1666	++p;
		1667	self->incr_mode = INCR_M_STR;
		1668	goto incr_m_str;
		1669
		1670	// inside #-style comments
		1671	case INCR_M_C0:
		1672	case INCR_M_C1:
		1673	incr_m_c:
		1674	for (;;)
		1675	{
		1676	if (*p == '\n')
		1677	{
		1678	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1679	break;
		1680	}
		1681	else if (!*p)
		1682	goto interrupt;
		1683
		1684	++p;
		1685	}
		1686
		1687	break;
		1688
		1689	// inside a string
		1690	case INCR_M_STR:
		1691	incr_m_str:
		1692	for (;;)
		1693	{
		1694	if (*p == '"')
		1695	{
		1696	++p;
		1697	self->incr_mode = INCR_M_JSON;
		1698
		1699	if (!self->incr_nest)
		1700	goto interrupt;
		1701
		1702	goto incr_m_json;
		1703	}
		1704	else if (*p == '\\')
		1705	{
		1706	++p; // "virtually" consumes character after \
		1707
		1708	if (!*p) // if at end of string we have to switch modes
		1709	{
		1710	self->incr_mode = INCR_M_BS;
		1711	goto interrupt;
		1712	}
		1713	}
		1714	else if (!*p)
		1715	goto interrupt;
		1716
		1717	++p;
		1718	}
		1719
		1720	// after initial ws, outside string
		1721	case INCR_M_JSON:
		1722	incr_m_json:
		1723	for (;;)
		1724	{
		1725	switch (*p++)
		1726	{
		1727	case 0:
		1728	--p;
		1729	goto interrupt;
		1730
		1731	case 0x09:
		1732	case 0x0a:
		1733	case 0x0d:
		1734	case 0x20:
		1735	if (!self->incr_nest)
		1736	{
		1737	--p; // do not eat the whitespace, let the next round do it
		1738	goto interrupt;
		1739	}
		1740	break;
		1741
		1742	case '"':
		1743	self->incr_mode = INCR_M_STR;
		1744	goto incr_m_str;
		1745
		1746	case '[':
		1747	case '{':
		1748	if (++self->incr_nest > self->max_depth)
		1749	croak (ERR_NESTING_EXCEEDED);
		1750	break;
		1751
		1752	case ']':
		1753	case '}':
		1754	if (--self->incr_nest <= 0)
		1755	goto interrupt;
		1756	break;
		1757
		1758	case '#':
		1759	self->incr_mode = INCR_M_C1;
		1760	goto incr_m_c;
		1761	}
		1762	}
		1763	}
		1764
		1765	modechange:
		1766	;
		1767	}
		1768
		1769	interrupt:
		1770	self->incr_pos = p - SvPVX (self->incr_text);
		1771	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1772	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
		1773	}
		1774
		1775	/////////////////////////////////////////////////////////////////////////////
		1776	// XS interface functions
		1777
915	MODULE = JSON::XS PACKAGE = JSON::XS	1778	MODULE = JSON::XS PACKAGE = JSON::XS
916		1779
917	BOOT:	1780	BOOT:
918	{	1781	{
919	int i;	1782	int i;
920		1783
921	memset (decode_hexdigit, 0xff, 256);
922	for (i = 10; i--; )	1784	for (i = 0; i < 256; ++i)
923	decode_hexdigit ['0' + i] = i;	1785	decode_hexdigit [i] =
		1786	i >= '0' && i <= '9' ? i - '0'
		1787	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1788	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1789	: -1;
924		1790
925	for (i = 7; i--; )
926	{
927	decode_hexdigit ['a' + i] = 10 + i;
928	decode_hexdigit ['A' + i] = 10 + i;
929	}
930
931	json_stash = gv_stashpv ("JSON::XS", 1);	1791	json_stash = gv_stashpv ("JSON::XS" , 1);
		1792	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1793
		1794	json_true = get_bool ("JSON::XS::true");
		1795	json_false = get_bool ("JSON::XS::false");
		1796
		1797	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
932	}	1798	}
933		1799
934	PROTOTYPES: DISABLE	1800	PROTOTYPES: DISABLE
935		1801
936	SV *new (char *dummy)	1802	void CLONE (...)
937	CODE:	1803	CODE:
938	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1804	json_stash = 0;
		1805	json_boolean_stash = 0;
		1806
		1807	void new (char *klass)
		1808	PPCODE:
		1809	{
		1810	SV *pv = NEWSV (0, sizeof (JSON));
		1811	SvPOK_only (pv);
		1812	json_init ((JSON *)SvPVX (pv));
		1813	XPUSHs (sv_2mortal (sv_bless (
		1814	newRV_noinc (pv),
		1815	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1816	)));
		1817	}
		1818
		1819	void ascii (JSON *self, int enable = 1)
		1820	ALIAS:
		1821	ascii = F_ASCII
		1822	latin1 = F_LATIN1
		1823	utf8 = F_UTF8
		1824	indent = F_INDENT
		1825	canonical = F_CANONICAL
		1826	space_before = F_SPACE_BEFORE
		1827	space_after = F_SPACE_AFTER
		1828	pretty = F_PRETTY
		1829	allow_nonref = F_ALLOW_NONREF
		1830	shrink = F_SHRINK
		1831	allow_blessed = F_ALLOW_BLESSED
		1832	convert_blessed = F_CONV_BLESSED
		1833	relaxed = F_RELAXED
		1834	allow_unknown = F_ALLOW_UNKNOWN
		1835	PPCODE:
		1836	{
		1837	if (enable)
		1838	self->flags |= ix;
		1839	else
		1840	self->flags &= ~ix;
		1841
		1842	XPUSHs (ST (0));
		1843	}
		1844
		1845	void get_ascii (JSON *self)
		1846	ALIAS:
		1847	get_ascii = F_ASCII
		1848	get_latin1 = F_LATIN1
		1849	get_utf8 = F_UTF8
		1850	get_indent = F_INDENT
		1851	get_canonical = F_CANONICAL
		1852	get_space_before = F_SPACE_BEFORE
		1853	get_space_after = F_SPACE_AFTER
		1854	get_allow_nonref = F_ALLOW_NONREF
		1855	get_shrink = F_SHRINK
		1856	get_allow_blessed = F_ALLOW_BLESSED
		1857	get_convert_blessed = F_CONV_BLESSED
		1858	get_relaxed = F_RELAXED
		1859	get_allow_unknown = F_ALLOW_UNKNOWN
		1860	PPCODE:
		1861	XPUSHs (boolSV (self->flags & ix));
		1862
		1863	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		1864	PPCODE:
		1865	self->max_depth = max_depth;
		1866	XPUSHs (ST (0));
		1867
		1868	U32 get_max_depth (JSON *self)
		1869	CODE:
		1870	RETVAL = self->max_depth;
939	OUTPUT:	1871	OUTPUT:
940	RETVAL	1872	RETVAL
941		1873
942	SV *ascii (SV *self, int enable = 1)	1874	void max_size (JSON *self, U32 max_size = 0)
943	ALIAS:	1875	PPCODE:
944	ascii = F_ASCII	1876	self->max_size = max_size;
945	utf8 = F_UTF8	1877	XPUSHs (ST (0));
946	indent = F_INDENT	1878
947	canonical = F_CANONICAL	1879	int get_max_size (JSON *self)
948	space_before = F_SPACE_BEFORE
949	space_after = F_SPACE_AFTER
950	pretty = F_PRETTY
951	allow_nonref = F_ALLOW_NONREF
952	shrink = F_SHRINK
953	CODE:	1880	CODE:
954	{	1881	RETVAL = self->max_size;
955	UV *uv = SvJSON (self);
956	if (enable)
957	*uv |= ix;
958	else
959	*uv &= ~ix;
960
961	RETVAL = newSVsv (self);
962	}
963	OUTPUT:	1882	OUTPUT:
964	RETVAL	1883	RETVAL
965		1884
966	void encode (SV *self, SV *scalar)	1885	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
967	PPCODE:	1886	PPCODE:
968	XPUSHs (encode_json (scalar, *SvJSON (self)));	1887	{
		1888	SvREFCNT_dec (self->cb_object);
		1889	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
969		1890
970	void decode (SV *self, SV *jsonstr)	1891	XPUSHs (ST (0));
		1892	}
		1893
		1894	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
971	PPCODE:	1895	PPCODE:
		1896	{
		1897	if (!self->cb_sk_object)
		1898	self->cb_sk_object = newHV ();
		1899
		1900	if (SvOK (cb))
		1901	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1902	else
		1903	{
		1904	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1905
		1906	if (!HvKEYS (self->cb_sk_object))
		1907	{
		1908	SvREFCNT_dec (self->cb_sk_object);
		1909	self->cb_sk_object = 0;
		1910	}
		1911	}
		1912
		1913	XPUSHs (ST (0));
		1914	}
		1915
		1916	void encode (JSON *self, SV *scalar)
		1917	PPCODE:
		1918	XPUSHs (encode_json (scalar, self));
		1919
		1920	void decode (JSON *self, SV *jsonstr)
		1921	PPCODE:
972	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1922	XPUSHs (decode_json (jsonstr, self, 0));
		1923
		1924	void decode_prefix (JSON *self, SV *jsonstr)
		1925	PPCODE:
		1926	{
		1927	char *offset;
		1928	EXTEND (SP, 2);
		1929	PUSHs (decode_json (jsonstr, self, &offset));
		1930	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		1931	}
		1932
		1933	void incr_parse (JSON *self, SV *jsonstr = 0)
		1934	PPCODE:
		1935	{
		1936	if (!self->incr_text)
		1937	self->incr_text = newSVpvn ("", 0);
		1938
		1939	/* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
		1940	if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
		1941	if (DECODE_WANTS_OCTETS (self))
		1942	{
		1943	if (self->incr_pos)
		1944	self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
		1945	(U8 *)SvPVX (self->incr_text) + self->incr_pos);
		1946
		1947	sv_utf8_downgrade (self->incr_text, 0);
		1948	}
		1949	else
		1950	{
		1951	sv_utf8_upgrade (self->incr_text);
		1952
		1953	if (self->incr_pos)
		1954	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1955	- (U8 *)SvPVX (self->incr_text);
		1956	}
		1957
		1958	// append data, if any
		1959	if (jsonstr)
		1960	{
		1961	/* make sure both strings have same encoding */
		1962	if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
		1963	if (SvUTF8 (jsonstr))
		1964	sv_utf8_downgrade (jsonstr, 0);
		1965	else
		1966	sv_utf8_upgrade (jsonstr);
		1967
		1968	/* and then just blindly append */
		1969	{
		1970	STRLEN len;
		1971	const char *str = SvPV (jsonstr, len);
		1972	STRLEN cur = SvCUR (self->incr_text);
		1973
		1974	if (SvLEN (self->incr_text) <= cur + len)
		1975	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1976
		1977	Move (str, SvEND (self->incr_text), len, char);
		1978	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1979	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1980	}
		1981	}
		1982
		1983	if (GIMME_V != G_VOID)
		1984	do
		1985	{
		1986	char *offset;
		1987
		1988	if (!INCR_DONE (self))
		1989	{
		1990	incr_parse (self);
		1991
		1992	if (self->incr_pos > self->max_size && self->max_size)
		1993	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1994	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1995
		1996	if (!INCR_DONE (self))
		1997	break;
		1998	}
		1999
		2000	XPUSHs (decode_json (self->incr_text, self, &offset));
		2001
		2002	self->incr_pos -= offset - SvPVX (self->incr_text);
		2003	self->incr_nest = 0;
		2004	self->incr_mode = 0;
		2005
		2006	sv_chop (self->incr_text, offset);
		2007	}
		2008	while (GIMME_V == G_ARRAY);
		2009	}
		2010
		2011	SV *incr_text (JSON *self)
		2012	ATTRS: lvalue
		2013	CODE:
		2014	{
		2015	if (self->incr_pos)
		2016	croak ("incr_text can not be called when the incremental parser already started parsing");
		2017
		2018	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		2019	}
		2020	OUTPUT:
		2021	RETVAL
		2022
		2023	void incr_skip (JSON *self)
		2024	CODE:
		2025	{
		2026	if (self->incr_pos)
		2027	{
		2028	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2029	self->incr_pos = 0;
		2030	self->incr_nest = 0;
		2031	self->incr_mode = 0;
		2032	}
		2033	}
		2034
		2035	void incr_reset (JSON *self)
		2036	CODE:
		2037	{
		2038	SvREFCNT_dec (self->incr_text);
		2039	self->incr_text = 0;
		2040	self->incr_pos = 0;
		2041	self->incr_nest = 0;
		2042	self->incr_mode = 0;
		2043	}
		2044
		2045	void DESTROY (JSON *self)
		2046	CODE:
		2047	SvREFCNT_dec (self->cb_sk_object);
		2048	SvREFCNT_dec (self->cb_object);
		2049	SvREFCNT_dec (self->incr_text);
973		2050
974	PROTOTYPES: ENABLE	2051	PROTOTYPES: ENABLE
975		2052
976	void to_json (SV *scalar)	2053	void encode_json (SV *scalar)
		2054	ALIAS:
		2055	to_json_ = 0
		2056	encode_json = F_UTF8
977	PPCODE:	2057	PPCODE:
		2058	{
		2059	JSON json;
		2060	json_init (&json);
		2061	json.flags |= ix;
978	XPUSHs (encode_json (scalar, F_UTF8));	2062	XPUSHs (encode_json (scalar, &json));
		2063	}
979		2064
980	void from_json (SV *jsonstr)	2065	void decode_json (SV *jsonstr)
		2066	ALIAS:
		2067	from_json_ = 0
		2068	decode_json = F_UTF8
981	PPCODE:	2069	PPCODE:
		2070	{
		2071	JSON json;
		2072	json_init (&json);
		2073	json.flags |= ix;
982	XPUSHs (decode_json (jsonstr, F_UTF8));	2074	XPUSHs (decode_json (jsonstr, &json, 0));
		2075	}
983		2076

Diff Legend

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