ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.3 by root, Thu Mar 22 18:10:29 2007 UTC vs.
Revision 1.134 by root, Thu Nov 15 20:13:03 2018 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>
		11	#include <inttypes.h>
8		12
		13	#if defined(__BORLANDC__) || defined(_MSC_VER)
		14	# define snprintf _snprintf // C compilers have this in stdio.h
		15	#endif
		16
		17	// some old perls do not have this, try to make it work, no
		18	// guarantees, though. if it breaks, you get to keep the pieces.
		19	#ifndef UTF8_MAXBYTES
		20	# define UTF8_MAXBYTES 13
		21	#endif
		22
		23	// compatibility with perl <5.18
		24	#ifndef HvNAMELEN_get
		25	# define HvNAMELEN_get(hv) strlen (HvNAME (hv))
		26	#endif
		27	#ifndef HvNAMELEN
		28	# define HvNAMELEN(hv) HvNAMELEN_get (hv)
		29	#endif
		30	#ifndef HvNAMEUTF8
		31	# define HvNAMEUTF8(hv) 0
		32	#endif
		33
		34	// three extra for rounding, sign, and end of string
		35	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
		36
9	#define F_ASCII 0x00000001	37	#define F_ASCII 0x00000001UL
		38	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	39	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	40	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	41	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	42	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	43	#define F_SPACE_AFTER 0x00000040UL
15	#define F_JSON_RPC 0x00000040
16	#define F_ALLOW_NONREF 0x00000080	44	#define F_ALLOW_NONREF 0x00000100UL
		45	#define F_SHRINK 0x00000200UL
		46	#define F_ALLOW_BLESSED 0x00000400UL
		47	#define F_CONV_BLESSED 0x00000800UL
		48	#define F_RELAXED 0x00001000UL
		49	#define F_ALLOW_UNKNOWN 0x00002000UL
		50	#define F_ALLOW_TAGS 0x00004000UL
		51	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
17		52
18	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	53	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
19	#define F_DEFAULT 0
20		54
21	#define INIT_SIZE 32 // initial scalar size to be allocated	55	#define INIT_SIZE 64 // initial scalar size to be allocated
		56	#define INDENT_STEP 3 // spaces per indentation level
		57
		58	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
		59
		60	#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
22		61
23	#define SB do {	62	#define SB do {
24	#define SE } while (0)	63	#define SE } while (0)
25		64
26	static HV *json_stash;	65	#if __GNUC__ >= 3
		66	# define expect(expr,value) __builtin_expect ((expr), (value))
		67	# define INLINE static inline
		68	#else
		69	# define expect(expr,value) (expr)
		70	# define INLINE static
		71	#endif
		72
		73	#define expect_false(expr) expect ((expr) != 0, 0)
		74	#define expect_true(expr) expect ((expr) != 0, 1)
		75
		76	#define IN_RANGE_INC(type,val,beg,end) \
		77	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		78	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		79
		80	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
		81
		82	#ifdef USE_ITHREADS
		83	# define JSON_STASH (expect_true (json_stash) ? json_stash : gv_stashpv ("JSON::XS", 1))
		84	# define BOOL_STASH (expect_true (bool_stash) ? bool_stash : gv_stashpv ("Types::Serialiser::Boolean", 1))
		85	# define GET_BOOL(value) (expect_true (bool_ ## value) ? bool_ ## value : get_bool ("Types::Serialiser::" # value))
		86	#else
		87	# define JSON_STASH json_stash
		88	# define BOOL_STASH bool_stash
		89	# define GET_BOOL(value) bool_ ## value
		90	#endif
		91
		92	// the amount of HEs to allocate on the stack, when sorting keys
		93	#define STACK_HES 64
		94
		95	static HV *json_stash, *bool_stash; // JSON::XS::, Types::Serialiser::Boolean::
		96	static SV *bool_false, *bool_true;
		97	static SV *sv_json;
		98
		99	enum {
		100	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		101	INCR_M_STR, // inside string
		102	INCR_M_BS, // inside backslash
		103	INCR_M_C0, // inside comment in initial whitespace sequence
		104	INCR_M_C1, // inside comment in other places
		105	INCR_M_JSON // outside anything, count nesting
		106	};
		107
		108	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		109
		110	typedef struct {
		111	U32 flags;
		112	U32 max_depth;
		113	STRLEN max_size;
		114
		115	SV *cb_object;
		116	HV *cb_sk_object;
		117
		118	// for the incremental parser
		119	SV *incr_text; // the source text so far
		120	STRLEN incr_pos; // the current offset into the text
		121	int incr_nest; // {[]}-nesting level
		122	unsigned char incr_mode;
		123
		124	SV *v_false, *v_true;
		125	} JSON;
		126
		127	INLINE void
		128	json_init (JSON *json)
		129	{
		130	Zero (json, 1, JSON);
		131	json->max_depth = 512;
		132	}
		133
		134	/////////////////////////////////////////////////////////////////////////////
		135	// utility functions
		136
		137	INLINE SV *
		138	get_bool (const char *name)
		139	{
		140	SV *sv = get_sv (name, 1);
		141
		142	SvREADONLY_on (sv);
		143	SvREADONLY_on (SvRV (sv));
		144
		145	return sv;
		146	}
		147
		148	INLINE void
		149	shrink (SV *sv)
		150	{
		151	sv_utf8_downgrade (sv, 1);
		152
		153	if (SvLEN (sv) > SvCUR (sv) + 1)
		154	{
		155	#ifdef SvPV_shrink_to_cur
		156	SvPV_shrink_to_cur (sv);
		157	#elif defined (SvPV_renew)
		158	SvPV_renew (sv, SvCUR (sv) + 1);
		159	#endif
		160	}
		161	}
		162
		163	/* adds two STRLENs together, slow, and with paranoia */
		164	STRLEN
		165	strlen_sum (STRLEN l1, STRLEN l2)
		166	{
		167	size_t sum = l1 + l2;
		168
		169	if (sum < (size_t)l2 || sum != (size_t)(STRLEN)sum)
		170	croak ("JSON::XS: string size overflow");
		171
		172	return sum;
		173	}
		174
		175	/* similar to SvGROW, but somewhat safer and guarantees exponential realloc strategy */
		176	static char *
		177	json_sv_grow (SV *sv, size_t len1, size_t len2)
		178	{
		179	len1 = strlen_sum (len1, len2);
		180	len1 = strlen_sum (len1, len1 >> 1);
		181
		182	if (len1 > 4096 - 24)
		183	len1 = (len1 | 4095) - 24;
		184
		185	return SvGROW (sv, len1);
		186	}
		187
		188	// decode a utf-8 character and return it, or (UV)-1 in
		189	// case of an error.
		190	// we special-case "safe" characters from U+80 .. U+7FF,
		191	// but use the very good perl function to parse anything else.
		192	// note that we never call this function for a ascii codepoints
		193	INLINE UV
		194	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		195	{
		196	if (expect_true (len >= 2
		197	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
		198	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
		199	{
		200	*clen = 2;
		201	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		202	}
		203	else
		204	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		205	}
		206
		207	// likewise for encoding, also never called for ascii codepoints
		208	// this function takes advantage of this fact, although current gccs
		209	// seem to optimise the check for >= 0x80 away anyways
		210	INLINE unsigned char *
		211	encode_utf8 (unsigned char *s, UV ch)
		212	{
		213	if (expect_false (ch < 0x000080))
		214	*s++ = ch;
		215	else if (expect_true (ch < 0x000800))
		216	*s++ = 0xc0 | ( ch >> 6),
		217	*s++ = 0x80 | ( ch & 0x3f);
		218	else if ( ch < 0x010000)
		219	*s++ = 0xe0 | ( ch >> 12),
		220	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		221	*s++ = 0x80 | ( ch & 0x3f);
		222	else if ( ch < 0x110000)
		223	*s++ = 0xf0 | ( ch >> 18),
		224	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		225	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		226	*s++ = 0x80 | ( ch & 0x3f);
		227
		228	return s;
		229	}
		230
		231	// convert offset pointer to character index, sv must be string
		232	static STRLEN
		233	ptr_to_index (SV *sv, char *offset)
		234	{
		235	return SvUTF8 (sv)
		236	? utf8_distance (offset, SvPVX (sv))
		237	: offset - SvPVX (sv);
		238	}
		239
		240	/////////////////////////////////////////////////////////////////////////////
		241	// fp hell
		242
		243	// scan a group of digits, and a trailing exponent
		244	static void
		245	json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
		246	{
		247	UV uaccum = 0;
		248	int eaccum = 0;
		249
		250	// if we recurse too deep, skip all remaining digits
		251	// to avoid a stack overflow attack
		252	if (expect_false (--maxdepth <= 0))
		253	while (((U8)*s - '0') < 10)
		254	++s;
		255
		256	for (;;)
		257	{
		258	U8 dig = (U8)*s - '0';
		259
		260	if (expect_false (dig >= 10))
		261	{
		262	if (dig == (U8)((U8)'.' - (U8)'0'))
		263	{
		264	++s;
		265	json_atof_scan1 (s, accum, expo, 1, maxdepth);
		266	}
		267	else if ((dig | ' ') == 'e' - '0')
		268	{
		269	int exp2 = 0;
		270	int neg = 0;
		271
		272	++s;
		273
		274	if (*s == '-')
		275	{
		276	++s;
		277	neg = 1;
		278	}
		279	else if (*s == '+')
		280	++s;
		281
		282	while ((dig = (U8)*s - '0') < 10)
		283	exp2 = exp2 * 10 + *s++ - '0';
		284
		285	*expo += neg ? -exp2 : exp2;
		286	}
		287
		288	break;
		289	}
		290
		291	++s;
		292
		293	uaccum = uaccum * 10 + dig;
		294	++eaccum;
		295
		296	// if we have too many digits, then recurse for more
		297	// we actually do this for rather few digits
		298	if (uaccum >= (UV_MAX - 9) / 10)
		299	{
		300	if (postdp) *expo -= eaccum;
		301	json_atof_scan1 (s, accum, expo, postdp, maxdepth);
		302	if (postdp) *expo += eaccum;
		303
		304	break;
		305	}
		306	}
		307
		308	// this relies greatly on the quality of the pow ()
		309	// implementation of the platform, but a good
		310	// implementation is hard to beat.
		311	// (IEEE 754 conformant ones are required to be exact)
		312	if (postdp) *expo -= eaccum;
		313	*accum += uaccum * Perl_pow (10., *expo);
		314	*expo += eaccum;
		315	}
		316
		317	static NV
		318	json_atof (const char *s)
		319	{
		320	NV accum = 0.;
		321	int expo = 0;
		322	int neg = 0;
		323
		324	if (*s == '-')
		325	{
		326	++s;
		327	neg = 1;
		328	}
		329
		330	// a recursion depth of ten gives us >>500 bits
		331	json_atof_scan1 (s, &accum, &expo, 0, 10);
		332
		333	return neg ? -accum : accum;
		334	}
		335
		336	// target of scalar reference is bool? -1 == nope, 0 == false, 1 == true
		337	static int
		338	ref_bool_type (SV *sv)
		339	{
		340	svtype svt = SvTYPE (sv);
		341
		342	if (svt < SVt_PVAV)
		343	{
		344	STRLEN len = 0;
		345	char *pv = svt ? SvPV (sv, len) : 0;
		346
		347	if (len == 1)
		348	if (*pv == '1')
		349	return 1;
		350	else if (*pv == '0')
		351	return 0;
		352	}
		353
		354	return -1;
		355	}
		356
		357	// returns whether scalar is not a reference in the sense of allow_nonref
		358	static int
		359	json_nonref (SV *scalar)
		360	{
		361	if (!SvROK (scalar))
		362	return 1;
		363
		364	scalar = SvRV (scalar);
		365
		366	if (SvTYPE (scalar) >= SVt_PVMG)
		367	{
		368	if (SvSTASH (scalar) == bool_stash)
		369	return 1;
		370
		371	if (!SvOBJECT (scalar) && ref_bool_type (scalar) >= 0)
		372	return 1;
		373	}
		374
		375	return 0;
		376	}
		377
		378	/////////////////////////////////////////////////////////////////////////////
		379	// encoder
27		380
28	// structure used for encoding JSON	381	// structure used for encoding JSON
29	typedef struct	382	typedef struct
30	{	383	{
31	char *cur;	384	char *cur; // SvPVX (sv) + current output position
32	STRLEN len; // SvLEN (sv)
33	char *end; // SvEND (sv)	385	char *end; // SvEND (sv)
34	SV *sv;	386	SV *sv; // result scalar
35	UV flags;	387	JSON json;
36	int max_recurse;	388	U32 indent; // indentation level
37	int indent;	389	UV limit; // escape character values >= this value when encoding
38	} enc_t;	390	} enc_t;
39		391
40	// structure used for decoding JSON	392	INLINE void
41	typedef struct
42	{
43	char *cur;
44	char *end;
45	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)	393	need (enc_t *enc, STRLEN len)
62	{	394	{
63	if (enc->cur + len >= enc->end)	395	if (expect_false ((uintptr_t)(enc->end - enc->cur) < len))
64	{	396	{
65	STRLEN cur = enc->cur - SvPVX (enc->sv);	397	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
66	SvGROW (enc->sv, cur + len + 1);	398	char *buf = json_sv_grow (enc->sv, cur, len);
67	enc->cur = SvPVX (enc->sv) + cur;	399	enc->cur = buf + cur;
68	enc->end = SvEND (enc->sv);	400	enc->end = buf + SvLEN (enc->sv) - 1;
69	}	401	}
70	}	402	}
71		403
72	static void	404	INLINE void
73	encode_ch (enc_t *enc, char ch)	405	encode_ch (enc_t *enc, char ch)
74	{	406	{
75	need (enc, 1);	407	need (enc, 1);
76	*enc->cur++ = ch;	408	*enc->cur++ = ch;
77	}	409	}
…		…
79	static void	411	static void
80	encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)	412	encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)
81	{	413	{
82	char *end = str + len;	414	char *end = str + len;
83		415
		416	need (enc, len);
		417
84	while (str < end)	418	while (str < end)
85	{	419	{
86	unsigned char ch = *(unsigned char *)str;	420	unsigned char ch = *(unsigned char *)str;
		421
87	if (ch >= 0x20 && ch < 0x80) // most common case	422	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
88	{	423	{
		424	if (expect_false (ch == '"')) // but with slow exceptions
		425	{
		426	need (enc, len + 1);
		427	*enc->cur++ = '\\';
		428	*enc->cur++ = '"';
		429	}
		430	else if (expect_false (ch == '\\'))
		431	{
		432	need (enc, len + 1);
		433	*enc->cur++ = '\\';
		434	*enc->cur++ = '\\';
		435	}
		436	else
89	*enc->cur++ = ch;	437	*enc->cur++ = ch;
		438
90	str++;	439	++str;
91	}	440	}
92	else	441	else
93	{	442	{
94	STRLEN clen;	443	switch (ch)
95	UV uch;
96
97	if (is_utf8)
98	{	444	{
99	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	445	case '\010': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
100	if (clen < 0)	446	case '\011': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
101	croak ("malformed UTF-8 character in string, cannot convert to JSON");	447	case '\012': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
102	}	448	case '\014': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
103	else	449	case '\015': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
104	{
105	uch = ch;
106	clen = 1;
107	}
108		450
109	need (enc, len += 6);	451	default:
110
111	if (uch < 0xa0 || enc->flags & F_ASCII)
112	{
113	if (uch > 0xFFFFUL)
114	{	452	{
		453	STRLEN clen;
		454	UV uch;
		455
		456	if (is_utf8)
		457	{
		458	uch = decode_utf8 (str, end - str, &clen);
		459	if (clen == (STRLEN)-1)
		460	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
		461	}
		462	else
		463	{
		464	uch = ch;
115	len += 6;	465	clen = 1;
		466	}
		467
		468	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
		469	{
		470	if (uch >= 0x10000UL)
		471	{
		472	if (uch >= 0x110000UL)
		473	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		474
116	need (enc, len += 6);	475	need (enc, len + 11);
117	sprintf (enc->cur, "\\u%04x\\u%04x",	476	sprintf (enc->cur, "\\u%04x\\u%04x",
118	(uch - 0x10000) / 0x400 + 0xD800,	477	(int)((uch - 0x10000) / 0x400 + 0xD800),
119	(uch - 0x10000) % 0x400 + 0xDC00);	478	(int)((uch - 0x10000) % 0x400 + 0xDC00));
120	enc->cur += 12;	479	enc->cur += 12;
		480	}
		481	else
		482	{
		483	need (enc, len + 5);
		484	*enc->cur++ = '\\';
		485	*enc->cur++ = 'u';
		486	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
		487	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
		488	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
		489	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
		490	}
		491
		492	str += clen;
121	}	493	}
		494	else if (enc->json.flags & F_LATIN1)
		495	{
		496	*enc->cur++ = uch;
		497	str += clen;
		498	}
		499	else if (is_utf8)
		500	{
		501	need (enc, len + clen);
		502	do
		503	{
		504	*enc->cur++ = *str++;
		505	}
		506	while (--clen);
		507	}
122	else	508	else
123	{	509	{
124	sprintf (enc->cur, "\\u%04x", uch);	510	need (enc, len + UTF8_MAXBYTES - 1); // never more than 11 bytes needed
125	enc->cur += 6;	511	enc->cur = encode_utf8 (enc->cur, uch);
		512	++str;
		513	}
126	}	514	}
127	}	515	}
128	else if (is_utf8)
129	{
130	memcpy (enc->cur, str, clen);
131	enc->cur += clen;
132	}
133	else
134	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
135
136	str += clen;
137	}	516	}
138		517
139	--len;	518	--len;
140	}	519	}
141	}	520	}
142		521
143	#define INDENT SB \	522	INLINE void
		523	encode_indent (enc_t *enc)
		524	{
144	if (enc->flags & F_INDENT) \	525	if (enc->json.flags & F_INDENT)
145	{ \	526	{
146	int i_; \	527	int spaces = enc->indent * INDENT_STEP;
147	need (enc, enc->indent); \	528
148	for (i_ = enc->indent * 3; i_--; )\	529	need (enc, spaces);
		530	memset (enc->cur, ' ', spaces);
		531	enc->cur += spaces;
		532	}
		533	}
		534
		535	INLINE void
		536	encode_space (enc_t *enc)
		537	{
		538	need (enc, 1);
		539	encode_ch (enc, ' ');
		540	}
		541
		542	INLINE void
		543	encode_nl (enc_t *enc)
		544	{
		545	if (enc->json.flags & F_INDENT)
		546	{
		547	need (enc, 1);
149	encode_ch (enc, ' '); \	548	encode_ch (enc, '\n');
150	} \	549	}
151	SE	550	}
152		551
153	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	552	INLINE void
154	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	553	encode_comma (enc_t *enc)
155	#define COMMA SB \	554	{
156	encode_ch (enc, ','); \	555	encode_ch (enc, ',');
		556
157	if (enc->flags & F_INDENT) \	557	if (enc->json.flags & F_INDENT)
158	NL; \	558	encode_nl (enc);
159	else if (enc->flags & F_SPACE_AFTER) \	559	else if (enc->json.flags & F_SPACE_AFTER)
160	SPACE; \	560	encode_space (enc);
161	SE	561	}
162		562
163	static void encode_sv (enc_t *enc, SV *sv);	563	static void encode_sv (enc_t *enc, SV *sv);
164		564
165	static void	565	static void
166	encode_av (enc_t *enc, AV *av)	566	encode_av (enc_t *enc, AV *av)
167	{	567	{
168	int i, len = av_len (av);	568	int i, len = av_len (av);
169		569
		570	if (enc->indent >= enc->json.max_depth)
		571	croak (ERR_NESTING_EXCEEDED);
		572
170	encode_ch (enc, '['); NL;	573	encode_ch (enc, '[');
171	++enc->indent;
172		574
		575	if (len >= 0)
		576	{
		577	encode_nl (enc); ++enc->indent;
		578
173	for (i = 0; i <= len; ++i)	579	for (i = 0; i <= len; ++i)
174	{	580	{
175	INDENT;	581	SV **svp = av_fetch (av, i, 0);
176	encode_sv (enc, *av_fetch (av, i, 0));
177		582
		583	encode_indent (enc);
		584
		585	if (svp)
		586	encode_sv (enc, *svp);
		587	else
		588	encode_str (enc, "null", 4, 0);
		589
178	if (i < len)	590	if (i < len)
179	COMMA;	591	encode_comma (enc);
180	}	592	}
181		593
182	NL;	594	encode_nl (enc); --enc->indent; encode_indent (enc);
		595	}
183		596
184	--enc->indent;
185	INDENT; encode_ch (enc, ']');	597	encode_ch (enc, ']');
186	}	598	}
187		599
188	static void	600	static void
189	encode_he (enc_t *enc, HE *he)	601	encode_hk (enc_t *enc, HE *he)
190	{	602	{
191	encode_ch (enc, '"');	603	encode_ch (enc, '"');
192		604
193	if (HeKLEN (he) == HEf_SVKEY)	605	if (HeKLEN (he) == HEf_SVKEY)
194	{	606	{
195	SV *sv = HeSVKEY (he);	607	SV *sv = HeSVKEY (he);
196	STRLEN len;	608	STRLEN len;
		609	char *str;
		610
		611	SvGETMAGIC (sv);
197	char *str = SvPV (sv, len);	612	str = SvPV (sv, len);
198		613
199	encode_str (enc, str, len, SvUTF8 (sv));	614	encode_str (enc, str, len, SvUTF8 (sv));
200	}	615	}
201	else	616	else
202	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	617	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
203		618
204	encode_ch (enc, '"');	619	encode_ch (enc, '"');
205		620
206	if (enc->flags & F_SPACE_BEFORE) SPACE;	621	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
207	encode_ch (enc, ':');	622	encode_ch (enc, ':');
208	if (enc->flags & F_SPACE_AFTER ) SPACE;	623	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
209	encode_sv (enc, HeVAL (he));
210	}	624	}
211		625
212	// compare hash entries, used when all keys are bytestrings	626	// compare hash entries, used when all keys are bytestrings
213	static int	627	static int
214	he_cmp_fast (const void *a_, const void *b_)	628	he_cmp_fast (const void *a_, const void *b_)
…		…
219	HE *b = *(HE **)b_;	633	HE *b = *(HE **)b_;
220		634
221	STRLEN la = HeKLEN (a);	635	STRLEN la = HeKLEN (a);
222	STRLEN lb = HeKLEN (b);	636	STRLEN lb = HeKLEN (b);
223		637
224	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	638	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
225	cmp = la < lb ? -1 : la == lb ? 0 : 1;	639	cmp = lb - la;
226		640
227	return cmp;	641	return cmp;
228	}	642	}
229		643
230	// compare hash entries, used when some keys are sv's or utf-x	644	// compare hash entries, used when some keys are sv's or utf-x
231	static int	645	static int
232	he_cmp_slow (const void *a, const void *b)	646	he_cmp_slow (const void *a, const void *b)
233	{	647	{
234	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	648	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
235	}	649	}
236		650
237	static void	651	static void
238	encode_hv (enc_t *enc, HV *hv)	652	encode_hv (enc_t *enc, HV *hv)
239	{	653	{
240	int count, i;	654	HE *he;
241		655
242	encode_ch (enc, '{'); NL; ++enc->indent;	656	if (enc->indent >= enc->json.max_depth)
		657	croak (ERR_NESTING_EXCEEDED);
243		658
244	if ((count = hv_iterinit (hv)))	659	encode_ch (enc, '{');
245	{	660
246	// for canonical output we have to sort by keys first	661	// for canonical output we have to sort by keys first
247	// actually, this is mostly due to the stupid so-called	662	// actually, this is mostly due to the stupid so-called
248	// security workaround added somewhere in 5.8.x.	663	// security workaround added somewhere in 5.8.x
249	// that randomises hash orderings	664	// that randomises hash orderings
250	if (enc->flags & F_CANONICAL)	665	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
		666	{
		667	int count = hv_iterinit (hv);
		668
		669	if (SvMAGICAL (hv))
251	{	670	{
252	HE *he, *hes [count];	671	// need to count by iterating. could improve by dynamically building the vector below
		672	// but I don't care for the speed of this special case.
		673	// note also that we will run into undefined behaviour when the two iterations
		674	// do not result in the same count, something I might care for in some later release.
		675
		676	count = 0;
		677	while (hv_iternext (hv))
		678	++count;
		679
		680	hv_iterinit (hv);
		681	}
		682
		683	if (count)
		684	{
253	int fast = 1;	685	int i, fast = 1;
		686	HE *hes_stack [STACK_HES];
		687	HE **hes = hes_stack;
		688
		689	// allocate larger arrays on the heap
		690	if (count > STACK_HES)
		691	{
		692	SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
		693	hes = (HE **)SvPVX (sv);
		694	}
254		695
255	i = 0;	696	i = 0;
256	while ((he = hv_iternext (hv)))	697	while ((he = hv_iternext (hv)))
257	{	698	{
258	hes [i++] = he;	699	hes [i++] = he;
…		…
264		705
265	if (fast)	706	if (fast)
266	qsort (hes, count, sizeof (HE *), he_cmp_fast);	707	qsort (hes, count, sizeof (HE *), he_cmp_fast);
267	else	708	else
268	{	709	{
269	// hack to disable "use bytes"	710	// hack to forcefully disable "use bytes"
270	COP *oldcop = PL_curcop, cop;	711	COP cop = *PL_curcop;
271	cop.op_private = 0;	712	cop.op_private = 0;
		713
		714	ENTER;
		715	SAVETMPS;
		716
		717	SAVEVPTR (PL_curcop);
272	PL_curcop = &cop;	718	PL_curcop = &cop;
273		719
274	SAVETMPS;
275	qsort (hes, count, sizeof (HE *), he_cmp_slow);	720	qsort (hes, count, sizeof (HE *), he_cmp_slow);
		721
276	FREETMPS;	722	FREETMPS;
277		723	LEAVE;
278	PL_curcop = oldcop;
279	}	724	}
280		725
281	for (i = 0; i < count; ++i)	726	encode_nl (enc); ++enc->indent;
		727
		728	while (count--)
282	{	729	{
283	INDENT;	730	encode_indent (enc);
		731	he = hes [count];
284	encode_he (enc, hes [i]);	732	encode_hk (enc, he);
		733	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
285		734
286	if (i < count - 1)	735	if (count)
287	COMMA;	736	encode_comma (enc);
288	}	737	}
289		738
290	NL;	739	encode_nl (enc); --enc->indent; encode_indent (enc);
291	}	740	}
		741	}
		742	else
		743	{
		744	if (hv_iterinit (hv) || SvMAGICAL (hv))
		745	if ((he = hv_iternext (hv)))
		746	{
		747	encode_nl (enc); ++enc->indent;
		748
		749	for (;;)
		750	{
		751	encode_indent (enc);
		752	encode_hk (enc, he);
		753	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		754
		755	if (!(he = hv_iternext (hv)))
		756	break;
		757
		758	encode_comma (enc);
		759	}
		760
		761	encode_nl (enc); --enc->indent; encode_indent (enc);
		762	}
		763	}
		764
		765	encode_ch (enc, '}');
		766	}
		767
		768	// encode objects, arrays and special \0=false and \1=true values.
		769	static void
		770	encode_rv (enc_t *enc, SV *sv)
		771	{
		772	svtype svt;
		773	GV *method;
		774
		775	SvGETMAGIC (sv);
		776	svt = SvTYPE (sv);
		777
		778	if (expect_false (SvOBJECT (sv)))
		779	{
		780	HV *stash = SvSTASH (sv);
		781
		782	if (stash == bool_stash)
		783	{
		784	if (SvIV (sv)) encode_str (enc, "true" , 4, 0);
		785	else encode_str (enc, "false", 5, 0);
		786	}
		787	else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
		788	{
		789	int count;
		790	dSP;
		791
		792	ENTER; SAVETMPS;
		793	PUSHMARK (SP);
		794	EXTEND (SP, 2);
		795	// we re-bless the reference to get overload and other niceties right
		796	PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		797	PUSHs (sv_json);
		798
		799	PUTBACK;
		800	count = call_sv ((SV *)GvCV (method), G_ARRAY);
		801	SPAGAIN;
		802
		803	// catch this surprisingly common error
		804	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		805	croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		806
		807	encode_ch (enc, '(');
		808	encode_ch (enc, '"');
		809	encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
		810	encode_ch (enc, '"');
		811	encode_ch (enc, ')');
		812	encode_ch (enc, '[');
		813
		814	if (count)
		815	{
		816	int i;
		817
		818	for (i = 0; i < count - 1; ++i)
		819	{
		820	encode_sv (enc, SP[i + 1 - count]);
		821	encode_ch (enc, ',');
		822	}
		823
		824	encode_sv (enc, TOPs);
		825	SP -= count;
		826	}
		827
		828	encode_ch (enc, ']');
		829
		830	FREETMPS; LEAVE;
		831	}
		832	else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
		833	{
		834	dSP;
		835
		836	ENTER; SAVETMPS;
		837	PUSHMARK (SP);
		838	// we re-bless the reference to get overload and other niceties right
		839	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		840
		841	// calling with G_SCALAR ensures that we always get a 1 return value
		842	PUTBACK;
		843	call_sv ((SV *)GvCV (method), G_SCALAR);
		844	SPAGAIN;
		845
		846	// catch this surprisingly common error
		847	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		848	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		849
		850	sv = POPs;
		851	PUTBACK;
		852
		853	encode_sv (enc, sv);
		854
		855	FREETMPS; LEAVE;
		856	}
		857	else if (enc->json.flags & F_ALLOW_BLESSED)
		858	encode_str (enc, "null", 4, 0);
292	else	859	else
293	{	860	croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
294	SV *sv;	861	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
295	HE *he = hv_iternext (hv);	862	}
296		863	else if (svt == SVt_PVHV)
297	for (;;)	864	encode_hv (enc, (HV *)sv);
298	{	865	else if (svt == SVt_PVAV)
299	INDENT;	866	encode_av (enc, (AV *)sv);
300	encode_he (enc, he);	867	else if (svt < SVt_PVAV)
301
302	if (!(he = hv_iternext (hv)))
303	break;
304
305	COMMA;
306	}
307
308	NL;
309	}
310	}	868	{
		869	int bool_type = ref_bool_type (sv);
311		870
312	--enc->indent; INDENT; encode_ch (enc, '}');	871	if (bool_type == 1)
		872	encode_str (enc, "true", 4, 0);
		873	else if (bool_type == 0)
		874	encode_str (enc, "false", 5, 0);
		875	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		876	encode_str (enc, "null", 4, 0);
		877	else
		878	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		879	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		880	}
		881	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		882	encode_str (enc, "null", 4, 0);
		883	else
		884	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		885	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
313	}	886	}
314		887
315	static void	888	static void
316	encode_sv (enc_t *enc, SV *sv)	889	encode_sv (enc_t *enc, SV *sv)
317	{	890	{
		891	SvGETMAGIC (sv);
		892
318	if (SvPOKp (sv))	893	if (SvPOKp (sv))
319	{	894	{
320	STRLEN len;	895	STRLEN len;
321	char *str = SvPV (sv, len);	896	char *str = SvPV (sv, len);
322	encode_ch (enc, '"');	897	encode_ch (enc, '"');
323	encode_str (enc, str, len, SvUTF8 (sv));	898	encode_str (enc, str, len, SvUTF8 (sv));
324	encode_ch (enc, '"');	899	encode_ch (enc, '"');
325	}	900	}
326	else if (SvNOKp (sv))	901	else if (SvNOKp (sv))
327	{	902	{
		903	// trust that perl will do the right thing w.r.t. JSON syntax.
328	need (enc, NV_DIG + 32);	904	need (enc, NV_DIG + 32);
329	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	905	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
330	enc->cur += strlen (enc->cur);	906	enc->cur += strlen (enc->cur);
331	}	907	}
332	else if (SvIOKp (sv))	908	else if (SvIOKp (sv))
333	{	909	{
		910	// we assume we can always read an IV as a UV and vice versa
		911	// we assume two's complement
		912	// we assume no aliasing issues in the union
		913	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		914	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		915	{
		916	// optimise the "small number case"
		917	// code will likely be branchless and use only a single multiplication
		918	// works for numbers up to 59074
		919	I32 i = SvIVX (sv);
		920	U32 u;
		921	char digit, nz = 0;
		922
334	need (enc, 64);	923	need (enc, 6);
		924
		925	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		926	u = i < 0 ? -i : i;
		927
		928	// convert to 4.28 fixed-point representation
		929	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		930
		931	// now output digit by digit, each time masking out the integer part
		932	// and multiplying by 5 while moving the decimal point one to the right,
		933	// resulting in a net multiplication by 10.
		934	// we always write the digit to memory but conditionally increment
		935	// the pointer, to enable the use of conditional move instructions.
		936	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		937	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		938	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		939	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		940	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		941	}
		942	else
		943	{
		944	// large integer, use the (rather slow) snprintf way.
		945	need (enc, IVUV_MAXCHARS);
335	enc->cur +=	946	enc->cur +=
336	SvIsUV(sv)	947	SvIsUV(sv)
337	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	948	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
338	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	949	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		950	}
339	}	951	}
340	else if (SvROK (sv))	952	else if (SvROK (sv))
341	{	953	encode_rv (enc, SvRV (sv));
342	if (!--enc->max_recurse)	954	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
343	croak ("data structure too deep (hit recursion limit)");
344
345	sv = SvRV (sv);
346
347	switch (SvTYPE (sv))
348	{
349	case SVt_PVAV: encode_av (enc, (AV *)sv); break;
350	case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
351
352	default:
353	croak ("JSON can only represent references to arrays or hashes");
354	}
355	}
356	else if (!SvOK (sv))
357	encode_str (enc, "null", 4, 0);	955	encode_str (enc, "null", 4, 0);
358	else	956	else
359	croak ("encountered perl type that JSON cannot handle");	957	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
		958	SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
360	}	959	}
361		960
362	static SV *	961	static SV *
363	encode_json (SV *scalar, UV flags)	962	encode_json (SV *scalar, JSON *json)
364	{	963	{
365	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
366	croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");
367
368	enc_t enc;	964	enc_t enc;
369	enc.flags = flags;	965
		966	if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar))
		967	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
		968
		969	enc.json = *json;
370	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	970	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
371	enc.cur = SvPVX (enc.sv);	971	enc.cur = SvPVX (enc.sv);
372	enc.end = SvEND (enc.sv);	972	enc.end = SvEND (enc.sv);
373	enc.max_recurse = 0;
374	enc.indent = 0;	973	enc.indent = 0;
		974	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		975	: enc.json.flags & F_LATIN1 ? 0x000100UL
		976	: 0x110000UL;
375		977
376	SvPOK_only (enc.sv);	978	SvPOK_only (enc.sv);
377	encode_sv (&enc, scalar);	979	encode_sv (&enc, scalar);
		980	encode_nl (&enc);
378		981
		982	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		983	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		984
379	if (!(flags & (F_ASCII | F_UTF8)))	985	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
380	SvUTF8_on (enc.sv);	986	SvUTF8_on (enc.sv);
381		987
382	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	988	if (enc.json.flags & F_SHRINK)
		989	shrink (enc.sv);
		990
383	return enc.sv;	991	return enc.sv;
384	}	992	}
385		993
386	/////////////////////////////////////////////////////////////////////////////	994	/////////////////////////////////////////////////////////////////////////////
		995	// decoder
387		996
388	#define WS \	997	// structure used for decoding JSON
		998	typedef struct
		999	{
		1000	char *cur; // current parser pointer
		1001	char *end; // end of input string
		1002	const char *err; // parse error, if != 0
		1003	JSON json;
		1004	U32 depth; // recursion depth
		1005	U32 maxdepth; // recursion depth limit
		1006	} dec_t;
		1007
		1008	INLINE void
		1009	decode_comment (dec_t *dec)
		1010	{
		1011	// only '#'-style comments allowed a.t.m.
		1012
		1013	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		1014	++dec->cur;
		1015	}
		1016
		1017	INLINE void
		1018	decode_ws (dec_t *dec)
		1019	{
389	for (;;) \	1020	for (;;)
390	{ \	1021	{
391	char ch = *dec->cur; \	1022	char ch = *dec->cur;
		1023
392	if (ch > 0x20 \	1024	if (ch > 0x20)
		1025	{
		1026	if (expect_false (ch == '#'))
		1027	{
		1028	if (dec->json.flags & F_RELAXED)
		1029	decode_comment (dec);
		1030	else
		1031	break;
		1032	}
		1033	else
		1034	break;
		1035	}
393	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	1036	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
394	break; \	1037	break; // parse error, but let higher level handle it, gives better error messages
		1038
395	++dec->cur; \	1039	++dec->cur;
396	}	1040	}
		1041	}
397		1042
398	#define ERR(reason) SB dec->err = reason; goto fail; SE	1043	#define ERR(reason) SB dec->err = reason; goto fail; SE
		1044
399	#define EXPECT_CH(ch) SB \	1045	#define EXPECT_CH(ch) SB \
400	if (*dec->cur != ch) \	1046	if (*dec->cur != ch) \
401	ERR (# ch " expected"); \	1047	ERR (# ch " expected"); \
402	++dec->cur; \	1048	++dec->cur; \
403	SE	1049	SE
404		1050
		1051	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
		1052	#define DEC_DEC_DEPTH --dec->depth
		1053
405	static SV *decode_sv (dec_t *dec);	1054	static SV *decode_sv (dec_t *dec);
406
407	#define APPEND_CH(ch) SB \
408	SvGROW (sv, cur + 1 + 1); \
409	SvPVX (sv)[cur++] = (ch); \
410	SE
411		1055
412	static signed char decode_hexdigit[256];	1056	static signed char decode_hexdigit[256];
413		1057
414	static UV	1058	static UV
415	decode_4hex (dec_t *dec)	1059	decode_4hex (dec_t *dec)
416	{	1060	{
417	signed char d1, d2, d3, d4;	1061	signed char d1, d2, d3, d4;
		1062	unsigned char *cur = (unsigned char *)dec->cur;
418		1063
419	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	1064	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
420	if (d1 < 0) ERR ("four hexadecimal digits expected");	1065	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
421	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	1066	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
422	if (d2 < 0) ERR ("four hexadecimal digits expected");	1067	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
423	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
424	if (d3 < 0) ERR ("four hexadecimal digits expected");
425	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
426	if (d4 < 0) ERR ("four hexadecimal digits expected");
427		1068
428	dec->cur += 4;	1069	dec->cur += 4;
429		1070
430	return ((UV)d1) << 12	1071	return ((UV)d1) << 12
431	| ((UV)d2) << 8	1072	| ((UV)d2) << 8
…		…
437	}	1078	}
438		1079
439	static SV *	1080	static SV *
440	decode_str (dec_t *dec)	1081	decode_str (dec_t *dec)
441	{	1082	{
442	SV *sv = NEWSV (0,2);	1083	SV *sv = 0;
443	STRLEN cur = 0;
444	int utf8 = 0;	1084	int utf8 = 0;
		1085	char *dec_cur = dec->cur;
445		1086
446	for (;;)	1087	do
447	{	1088	{
448	unsigned char ch = *(unsigned char *)dec->cur;	1089	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		1090	char *cur = buf;
449		1091
450	if (ch == '"')	1092	do
451	break;
452	else if (ch == '\\')
453	{	1093	{
454	switch (*++dec->cur)	1094	unsigned char ch = *(unsigned char *)dec_cur++;
		1095
		1096	if (expect_false (ch == '"'))
455	{	1097	{
456	case '\\':	1098	--dec_cur;
457	case '/':	1099	break;
458	case '"': APPEND_CH (*dec->cur++); break;	1100	}
459		1101	else if (expect_false (ch == '\\'))
460	case 'b': APPEND_CH ('\010'); ++dec->cur; break;	1102	{
461	case 't': APPEND_CH ('\011'); ++dec->cur; break;	1103	switch (*dec_cur)
462	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
463	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
464	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
465
466	case 'u':
467	{	1104	{
468	UV lo, hi;	1105	case '\\':
469	++dec->cur;	1106	case '/':
		1107	case '"': *cur++ = *dec_cur++; break;
470		1108
471	hi = decode_4hex (dec);	1109	case 'b': ++dec_cur; *cur++ = '\010'; break;
472	if (hi == (UV)-1)	1110	case 't': ++dec_cur; *cur++ = '\011'; break;
473	goto fail;	1111	case 'n': ++dec_cur; *cur++ = '\012'; break;
		1112	case 'f': ++dec_cur; *cur++ = '\014'; break;
		1113	case 'r': ++dec_cur; *cur++ = '\015'; break;
474		1114
475	// possibly a surrogate pair	1115	case 'u':
476	if (hi >= 0xd800 && hi < 0xdc00)
477	{	1116	{
478	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	1117	UV lo, hi;
479	ERR ("illegal surrogate character");	1118	++dec_cur;
480		1119
481	dec->cur += 2;	1120	dec->cur = dec_cur;
482
483	lo = decode_4hex (dec);	1121	hi = decode_4hex (dec);
		1122	dec_cur = dec->cur;
484	if (lo == (UV)-1)	1123	if (hi == (UV)-1)
485	goto fail;	1124	goto fail;
486		1125
		1126	// possibly a surrogate pair
		1127	if (hi >= 0xd800)
		1128	if (hi < 0xdc00)
		1129	{
		1130	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		1131	ERR ("missing low surrogate character in surrogate pair");
		1132
		1133	dec_cur += 2;
		1134
		1135	dec->cur = dec_cur;
		1136	lo = decode_4hex (dec);
		1137	dec_cur = dec->cur;
		1138	if (lo == (UV)-1)
		1139	goto fail;
		1140
487	if (lo < 0xdc00 || lo >= 0xe000)	1141	if (lo < 0xdc00 || lo >= 0xe000)
488	ERR ("surrogate pair expected");	1142	ERR ("surrogate pair expected");
489		1143
490	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	1144	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		1145	}
		1146	else if (hi < 0xe000)
		1147	ERR ("missing high surrogate character in surrogate pair");
		1148
		1149	if (hi >= 0x80)
		1150	{
		1151	utf8 = 1;
		1152
		1153	cur = encode_utf8 (cur, hi);
		1154	}
		1155	else
		1156	*cur++ = hi;
491	}	1157	}
492	else if (lo >= 0xdc00 && lo < 0xe000)
493	ERR ("illegal surrogate character");
494
495	if (hi >= 0x80)
496	{	1158	break;
497	utf8 = 1;
498		1159
499	SvGROW (sv, cur + 4 + 1); // at most 4 bytes for 21 bits
500	cur = (char *)uvuni_to_utf8_flags (SvPVX (sv) + cur, hi, 0) - SvPVX (sv);
501	}
502	else	1160	default:
503	APPEND_CH (hi);	1161	--dec_cur;
		1162	ERR ("illegal backslash escape sequence in string");
504	}	1163	}
		1164	}
		1165	else if (expect_true (ch >= 0x20 && ch < 0x80))
		1166	*cur++ = ch;
		1167	else if (ch >= 0x80)
		1168	{
		1169	STRLEN clen;
		1170
		1171	--dec_cur;
		1172
		1173	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		1174	if (clen == (STRLEN)-1)
		1175	ERR ("malformed UTF-8 character in JSON string");
		1176
		1177	do
		1178	*cur++ = *dec_cur++;
		1179	while (--clen);
		1180
		1181	utf8 = 1;
		1182	}
		1183	else if (ch == '\t' && dec->json.flags & F_RELAXED)
		1184	*cur++ = ch;
		1185	else
		1186	{
		1187	--dec_cur;
		1188
		1189	if (!ch)
		1190	ERR ("unexpected end of string while parsing JSON string");
505	break;	1191	else
		1192	ERR ("invalid character encountered while parsing JSON string");
506	}	1193	}
507	}	1194	}
508	else if (ch >= 0x20 && ch <= 0x7f)	1195	while (cur < buf + SHORT_STRING_LEN);
509	APPEND_CH (*dec->cur++);	1196
510	else if (ch >= 0x80)	1197	{
		1198	STRLEN len = cur - buf;
		1199
		1200	if (sv)
511	{	1201	{
512	STRLEN clen;	1202	STRLEN cur = SvCUR (sv);
513	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
514	if (clen < 0)
515	ERR ("malformed UTF-8 character in string, cannot convert to JSON");
516		1203
517	SvGROW (sv, cur + clen + 1); // at most 4 bytes for 21 bits	1204	if (SvLEN (sv) - cur <= len)
518	memcpy (SvPVX (sv) + cur, dec->cur, clen);	1205	json_sv_grow (sv, cur, len);
519	dec->cur += clen;	1206
		1207	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		1208	SvCUR_set (sv, SvCUR (sv) + len);
520	}	1209	}
521	else	1210	else
522	ERR ("invalid character encountered");	1211	sv = newSVpvn (buf, len);
523	}	1212	}
		1213	}
		1214	while (*dec_cur != '"');
524		1215
525	++dec->cur;	1216	++dec_cur;
526		1217
		1218	if (sv)
		1219	{
527	SvPOK_only (sv);	1220	SvPOK_only (sv);
528
529	SvCUR_set (sv, cur);
530	*SvEND (sv) = 0;	1221	*SvEND (sv) = 0;
531		1222
532	if (utf8)	1223	if (utf8)
533	SvUTF8_on (sv);	1224	SvUTF8_on (sv);
		1225	}
		1226	else
		1227	sv = newSVpvn ("", 0);
534		1228
		1229	dec->cur = dec_cur;
535	return sv;	1230	return sv;
536		1231
537	fail:	1232	fail:
538	SvREFCNT_dec (sv);	1233	dec->cur = dec_cur;
539	return 0;	1234	return 0;
540	}	1235	}
541		1236
542	static SV *	1237	static SV *
543	decode_num (dec_t *dec)	1238	decode_num (dec_t *dec)
…		…
553	{	1248	{
554	++dec->cur;	1249	++dec->cur;
555	if (*dec->cur >= '0' && *dec->cur <= '9')	1250	if (*dec->cur >= '0' && *dec->cur <= '9')
556	ERR ("malformed number (leading zero must not be followed by another digit)");	1251	ERR ("malformed number (leading zero must not be followed by another digit)");
557	}	1252	}
558		1253	else if (*dec->cur < '0' || *dec->cur > '9')
559	// int	1254	ERR ("malformed number (no digits after initial minus)");
		1255	else
		1256	do
		1257	{
		1258	++dec->cur;
		1259	}
560	while (*dec->cur >= '0' && *dec->cur <= '9')	1260	while (*dec->cur >= '0' && *dec->cur <= '9');
561	++dec->cur;
562		1261
563	// [frac]	1262	// [frac]
564	if (*dec->cur == '.')	1263	if (*dec->cur == '.')
565	{	1264	{
566	is_nv = 1;	1265	++dec->cur;
		1266
		1267	if (*dec->cur < '0' || *dec->cur > '9')
		1268	ERR ("malformed number (no digits after decimal point)");
567		1269
568	do	1270	do
569	{	1271	{
570	++dec->cur;	1272	++dec->cur;
571	}	1273	}
572	while (*dec->cur >= '0' && *dec->cur <= '9');	1274	while (*dec->cur >= '0' && *dec->cur <= '9');
		1275
		1276	is_nv = 1;
573	}	1277	}
574		1278
575	// [exp]	1279	// [exp]
576	if (*dec->cur == 'e' || *dec->cur == 'E')	1280	if (*dec->cur == 'e' || *dec->cur == 'E')
577	{	1281	{
578	is_nv = 1;
579
580	++dec->cur;	1282	++dec->cur;
		1283
581	if (*dec->cur == '-' || *dec->cur == '+')	1284	if (*dec->cur == '-' || *dec->cur == '+')
582	++dec->cur;	1285	++dec->cur;
583		1286
		1287	if (*dec->cur < '0' || *dec->cur > '9')
		1288	ERR ("malformed number (no digits after exp sign)");
		1289
		1290	do
		1291	{
		1292	++dec->cur;
		1293	}
584	while (*dec->cur >= '0' && *dec->cur <= '9')	1294	while (*dec->cur >= '0' && *dec->cur <= '9');
585	++dec->cur;	1295
		1296	is_nv = 1;
586	}	1297	}
587		1298
588	if (!is_nv)	1299	if (!is_nv)
589	{	1300	{
590	UV uv;	1301	int len = dec->cur - start;
591	int numtype = grok_number (start, dec->cur - start, &uv);	1302
592	if (numtype & IS_NUMBER_IN_UV)	1303	// special case the rather common 1..5-digit-int case
593	if (numtype & IS_NUMBER_NEG)	1304	if (*start == '-')
		1305	switch (len)
594	{	1306	{
595	if (uv < (UV)IV_MIN)	1307	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
596	return newSViv (-(IV)uv);	1308	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
		1309	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1310	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1311	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
597	}	1312	}
		1313	else
		1314	switch (len)
		1315	{
		1316	case 1: return newSViv ( start [0] - '0' * 1);
		1317	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1318	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1319	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1320	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1321	}
		1322
		1323	{
		1324	UV uv;
		1325	int numtype = grok_number (start, len, &uv);
		1326	if (numtype & IS_NUMBER_IN_UV)
		1327	if (numtype & IS_NUMBER_NEG)
		1328	{
		1329	if (uv < (UV)IV_MIN)
		1330	return newSViv (-(IV)uv);
		1331	}
598	else	1332	else
599	return newSVuv (uv);	1333	return newSVuv (uv);
600	}	1334	}
601		1335
		1336	len -= *start == '-' ? 1 : 0;
		1337
		1338	// does not fit into IV or UV, try NV
		1339	if (len <= NV_DIG)
		1340	// fits into NV without loss of precision
		1341	return newSVnv (json_atof (start));
		1342
		1343	// everything else fails, convert it to a string
		1344	return newSVpvn (start, dec->cur - start);
		1345	}
		1346
		1347	// loss of precision here
602	return newSVnv (Atof (start));	1348	return newSVnv (json_atof (start));
603		1349
604	fail:	1350	fail:
605	return 0;	1351	return 0;
606	}	1352	}
607		1353
608	static SV *	1354	static SV *
609	decode_av (dec_t *dec)	1355	decode_av (dec_t *dec)
610	{	1356	{
611	AV *av = newAV ();	1357	AV *av = newAV ();
612		1358
		1359	DEC_INC_DEPTH;
		1360	decode_ws (dec);
		1361
		1362	if (*dec->cur == ']')
		1363	++dec->cur;
		1364	else
613	for (;;)	1365	for (;;)
614	{	1366	{
615	SV *value;	1367	SV *value;
616		1368
617	value = decode_sv (dec);	1369	value = decode_sv (dec);
618	if (!value)	1370	if (!value)
619	goto fail;	1371	goto fail;
620		1372
621	av_push (av, value);	1373	av_push (av, value);
622		1374
623	WS;	1375	decode_ws (dec);
624		1376
625	if (*dec->cur == ']')	1377	if (*dec->cur == ']')
626	{	1378	{
627	++dec->cur;	1379	++dec->cur;
628	break;	1380	break;
629	}	1381	}
630		1382
631	if (*dec->cur != ',')	1383	if (*dec->cur != ',')
632	ERR (", or ] expected while parsing array");	1384	ERR (", or ] expected while parsing array");
633		1385
634	++dec->cur;	1386	++dec->cur;
		1387
		1388	decode_ws (dec);
		1389
		1390	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1391	{
		1392	++dec->cur;
		1393	break;
		1394	}
635	}	1395	}
636		1396
		1397	DEC_DEC_DEPTH;
637	return newRV_noinc ((SV *)av);	1398	return newRV_noinc ((SV *)av);
638		1399
639	fail:	1400	fail:
640	SvREFCNT_dec (av);	1401	SvREFCNT_dec (av);
		1402	DEC_DEC_DEPTH;
641	return 0;	1403	return 0;
642	}	1404	}
643		1405
644	static SV *	1406	static SV *
645	decode_hv (dec_t *dec)	1407	decode_hv (dec_t *dec)
646	{	1408	{
		1409	SV *sv;
647	HV *hv = newHV ();	1410	HV *hv = newHV ();
648		1411
		1412	DEC_INC_DEPTH;
		1413	decode_ws (dec);
		1414
		1415	if (*dec->cur == '}')
		1416	++dec->cur;
		1417	else
649	for (;;)	1418	for (;;)
650	{	1419	{
651	SV *key, *value;
652
653	WS; EXPECT_CH ('"');	1420	EXPECT_CH ('"');
654		1421
655	key = decode_str (dec);	1422	// heuristic: assume that
656	if (!key)	1423	// a) decode_str + hv_store_ent are abysmally slow.
657	goto fail;	1424	// b) most hash keys are short, simple ascii text.
658		1425	// => try to "fast-match" such strings to avoid
659	WS; EXPECT_CH (':');	1426	// the overhead of decode_str + hv_store_ent.
660
661	value = decode_sv (dec);
662	if (!value)
663	{	1427	{
		1428	SV *value;
		1429	char *p = dec->cur;
		1430	char *e = p + 24; // only try up to 24 bytes
		1431
		1432	for (;;)
		1433	{
		1434	// the >= 0x80 is false on most architectures
		1435	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1436	{
		1437	// slow path, back up and use decode_str
		1438	SV *key = decode_str (dec);
		1439	if (!key)
		1440	goto fail;
		1441
		1442	decode_ws (dec); EXPECT_CH (':');
		1443
		1444	decode_ws (dec);
		1445	value = decode_sv (dec);
		1446	if (!value)
		1447	{
		1448	SvREFCNT_dec (key);
		1449	goto fail;
		1450	}
		1451
		1452	hv_store_ent (hv, key, value, 0);
664	SvREFCNT_dec (key);	1453	SvREFCNT_dec (key);
		1454
		1455	break;
		1456	}
		1457	else if (*p == '"')
		1458	{
		1459	// fast path, got a simple key
		1460	char *key = dec->cur;
		1461	int len = p - key;
		1462	dec->cur = p + 1;
		1463
		1464	decode_ws (dec); EXPECT_CH (':');
		1465
		1466	decode_ws (dec);
		1467	value = decode_sv (dec);
		1468	if (!value)
665	goto fail;	1469	goto fail;
		1470
		1471	hv_store (hv, key, len, value, 0);
		1472
		1473	break;
		1474	}
		1475
		1476	++p;
		1477	}
666	}	1478	}
667		1479
668	//TODO: optimise	1480	decode_ws (dec);
669	hv_store_ent (hv, key, value, 0);
670		1481
671	WS;
672
673	if (*dec->cur == '}')	1482	if (*dec->cur == '}')
		1483	{
		1484	++dec->cur;
		1485	break;
		1486	}
		1487
		1488	if (*dec->cur != ',')
		1489	ERR (", or } expected while parsing object/hash");
		1490
		1491	++dec->cur;
		1492
		1493	decode_ws (dec);
		1494
		1495	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1496	{
		1497	++dec->cur;
		1498	break;
		1499	}
		1500	}
		1501
		1502	DEC_DEC_DEPTH;
		1503	sv = newRV_noinc ((SV *)hv);
		1504
		1505	// check filter callbacks
		1506	if (dec->json.flags & F_HOOK)
		1507	{
		1508	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
674	{	1509	{
675	++dec->cur;	1510	HE *cb, *he;
676	break;	1511
		1512	hv_iterinit (hv);
		1513	he = hv_iternext (hv);
		1514	hv_iterinit (hv);
		1515
		1516	// the next line creates a mortal sv each time it's called.
		1517	// might want to optimise this for common cases.
		1518	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1519
		1520	if (cb)
		1521	{
		1522	dSP;
		1523	int count;
		1524
		1525	ENTER; SAVETMPS;
		1526	PUSHMARK (SP);
		1527	XPUSHs (HeVAL (he));
		1528	sv_2mortal (sv);
		1529
		1530	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1531
		1532	if (count == 1)
		1533	{
		1534	sv = newSVsv (POPs);
		1535	FREETMPS; LEAVE;
		1536	return sv;
		1537	}
		1538	else if (count)
		1539	croak ("filter_json_single_key_object callbacks must not return more than one scalar");
		1540
		1541	SvREFCNT_inc (sv);
		1542	FREETMPS; LEAVE;
		1543	}
677	}	1544	}
678		1545
679	if (*dec->cur != ',')	1546	if (dec->json.cb_object)
680	ERR (", or } expected while parsing object/hash");	1547	{
		1548	dSP;
		1549	int count;
681		1550
682	++dec->cur;	1551	ENTER; SAVETMPS;
683	}	1552	PUSHMARK (SP);
		1553	XPUSHs (sv_2mortal (sv));
684		1554
685	return newRV_noinc ((SV *)hv);	1555	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1556
		1557	if (count == 1)
		1558	{
		1559	sv = newSVsv (POPs);
		1560	FREETMPS; LEAVE;
		1561	return sv;
		1562	}
		1563	else if (count)
		1564	croak ("filter_json_object callbacks must not return more than one scalar");
		1565
		1566	SvREFCNT_inc (sv);
		1567	FREETMPS; LEAVE;
		1568	}
		1569	}
		1570
		1571	return sv;
686		1572
687	fail:	1573	fail:
688	SvREFCNT_dec (hv);	1574	SvREFCNT_dec (hv);
		1575	DEC_DEC_DEPTH;
		1576	return 0;
		1577	}
		1578
		1579	static SV *
		1580	decode_tag (dec_t *dec)
		1581	{
		1582	SV *tag = 0;
		1583	SV *val = 0;
		1584
		1585	if (!(dec->json.flags & F_ALLOW_TAGS))
		1586	ERR ("malformed JSON string, neither array, object, number, string or atom");
		1587
		1588	++dec->cur;
		1589
		1590	decode_ws (dec);
		1591
		1592	tag = decode_sv (dec);
		1593	if (!tag)
		1594	goto fail;
		1595
		1596	if (!SvPOK (tag))
		1597	ERR ("malformed JSON string, (tag) must be a string");
		1598
		1599	decode_ws (dec);
		1600
		1601	if (*dec->cur != ')')
		1602	ERR (") expected after tag");
		1603
		1604	++dec->cur;
		1605
		1606	decode_ws (dec);
		1607
		1608	val = decode_sv (dec);
		1609	if (!val)
		1610	goto fail;
		1611
		1612	if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
		1613	ERR ("malformed JSON string, tag value must be an array");
		1614
		1615	{
		1616	AV *av = (AV *)SvRV (val);
		1617	int i, len = av_len (av) + 1;
		1618	HV *stash = gv_stashsv (tag, 0);
		1619	SV *sv;
		1620
		1621	if (!stash)
		1622	ERR ("cannot decode perl-object (package does not exist)");
		1623
		1624	GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
		1625
		1626	if (!method)
		1627	ERR ("cannot decode perl-object (package does not have a THAW method)");
		1628
		1629	dSP;
		1630
		1631	ENTER; SAVETMPS;
		1632	PUSHMARK (SP);
		1633	EXTEND (SP, len + 2);
		1634	// we re-bless the reference to get overload and other niceties right
		1635	PUSHs (tag);
		1636	PUSHs (sv_json);
		1637
		1638	for (i = 0; i < len; ++i)
		1639	PUSHs (*av_fetch (av, i, 1));
		1640
		1641	PUTBACK;
		1642	call_sv ((SV *)GvCV (method), G_SCALAR);
		1643	SPAGAIN;
		1644
		1645	SvREFCNT_dec (tag);
		1646	SvREFCNT_dec (val);
		1647	sv = SvREFCNT_inc (POPs);
		1648
		1649	PUTBACK;
		1650
		1651	FREETMPS; LEAVE;
		1652
		1653	return sv;
		1654	}
		1655
		1656	fail:
		1657	SvREFCNT_dec (tag);
		1658	SvREFCNT_dec (val);
689	return 0;	1659	return 0;
690	}	1660	}
691		1661
692	static SV *	1662	static SV *
693	decode_sv (dec_t *dec)	1663	decode_sv (dec_t *dec)
694	{	1664	{
695	WS;	1665	// the beauty of JSON: you need exactly one character lookahead
		1666	// to parse everything.
696	switch (*dec->cur)	1667	switch (*dec->cur)
697	{	1668	{
698	case '"': ++dec->cur; return decode_str (dec);	1669	case '"': ++dec->cur; return decode_str (dec);
699	case '[': ++dec->cur; return decode_av (dec);	1670	case '[': ++dec->cur; return decode_av (dec);
700	case '{': ++dec->cur; return decode_hv (dec);	1671	case '{': ++dec->cur; return decode_hv (dec);
		1672	case '(': return decode_tag (dec);
701		1673
702	case '-':	1674	case '-':
703	case '0': case '1': case '2': case '3': case '4':	1675	case '0': case '1': case '2': case '3': case '4':
704	case '5': case '6': case '7': case '8': case '9':	1676	case '5': case '6': case '7': case '8': case '9':
705	return decode_num (dec);	1677	return decode_num (dec);
706		1678
		1679	case 'f':
		1680	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
		1681	{
		1682	dec->cur += 5;
		1683
		1684	if (expect_false (!dec->json.v_false))
		1685	dec->json.v_false = GET_BOOL (false);
		1686
		1687	return newSVsv (dec->json.v_false);
		1688	}
		1689	else
		1690	ERR ("'false' expected");
		1691
		1692	break;
		1693
707	case 't':	1694	case 't':
708	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1695	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
709	{	1696	{
710	dec->cur += 4;	1697	dec->cur += 4;
		1698
		1699	if (expect_false (!dec->json.v_true))
		1700	dec->json.v_true = GET_BOOL (true);
		1701
711	return newSViv (1);	1702	return newSVsv (dec->json.v_true);
712	}	1703	}
713	else	1704	else
714	ERR ("'true' expected");	1705	ERR ("'true' expected");
715
716	break;
717
718	case 'f':
719	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
720	{
721	dec->cur += 5;
722	return newSViv (0);
723	}
724	else
725	ERR ("'false' expected");
726		1706
727	break;	1707	break;
728		1708
729	case 'n':	1709	case 'n':
730	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))	1710	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
731	{	1711	{
732	dec->cur += 4;	1712	dec->cur += 4;
733	return newSViv (1);	1713	return newSVsv (&PL_sv_undef);
734	}	1714	}
735	else	1715	else
736	ERR ("'null' expected");	1716	ERR ("'null' expected");
737		1717
738	break;	1718	break;
739		1719
740	default:	1720	default:
741	ERR ("malformed json string");	1721	ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
742	break;	1722	break;
743	}	1723	}
744		1724
745	fail:	1725	fail:
746	return 0;	1726	return 0;
747	}	1727	}
748		1728
749	static SV *	1729	static SV *
750	decode_json (SV *string, UV flags)	1730	decode_json (SV *string, JSON *json, STRLEN *offset_return)
751	{	1731	{
		1732	dec_t dec;
752	SV *sv;	1733	SV *sv;
753		1734
754	if (!(flags & F_UTF8))	1735	/* work around bugs in 5.10 where manipulating magic values
		1736	* makes perl ignore the magic in subsequent accesses.
		1737	* also make a copy of non-PV values, to get them into a clean
		1738	* state (SvPV should do that, but it's buggy, see below).
		1739	*
		1740	* SvIsCOW_shared_hash works around a bug in perl (possibly 5.16),
		1741	* as reported by Reini Urban.
		1742	*/
		1743	/*SvGETMAGIC (string);*/
		1744	if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
		1745	string = sv_2mortal (newSVsv (string));
		1746
		1747	SvUPGRADE (string, SVt_PV);
		1748
		1749	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1750	* assertion with -DDEBUGGING, although SvCUR is documented to
		1751	* return the xpv_cur field which certainly exists after upgrading.
		1752	* according to nicholas clark, calling SvPOK fixes this.
		1753	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1754	* and hope for the best.
		1755	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1756	* assertion business is seriously broken, try yet another workaround
		1757	* for the broken -DDEBUGGING.
		1758	*/
		1759	{
		1760	#ifdef DEBUGGING
		1761	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1762	#else
		1763	STRLEN offset = SvCUR (string);
		1764	#endif
		1765
		1766	if (offset > json->max_size && json->max_size)
		1767	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1768	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1769	}
		1770
		1771	if (DECODE_WANTS_OCTETS (json))
		1772	sv_utf8_downgrade (string, 0);
		1773	else
755	sv_utf8_upgrade (string);	1774	sv_utf8_upgrade (string);
756		1775
757	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1776	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
758		1777
759	dec_t dec;	1778	dec.json = *json;
760	dec.flags = flags;
761	dec.cur = SvPVX (string);	1779	dec.cur = SvPVX (string);
762	dec.end = SvEND (string);	1780	dec.end = SvEND (string);
763	dec.err = 0;	1781	dec.err = 0;
		1782	dec.depth = 0;
764		1783
765	*dec.end = 1; // invalid anywhere	1784	if (dec.json.cb_object || dec.json.cb_sk_object)
		1785	dec.json.flags |= F_HOOK;
		1786
		1787	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1788
		1789	decode_ws (&dec);
766	sv = decode_sv (&dec);	1790	sv = decode_sv (&dec);
767	*dec.end = 0;	1791
		1792	if (offset_return)
		1793	*offset_return = dec.cur - SvPVX (string);
		1794	else if (sv)
		1795	{
		1796	// check for trailing garbage
		1797	decode_ws (&dec);
		1798
		1799	if (*dec.cur)
		1800	{
		1801	dec.err = "garbage after JSON object";
		1802	SvREFCNT_dec (sv);
		1803	sv = 0;
		1804	}
		1805	}
768		1806
769	if (!sv)	1807	if (!sv)
770	{	1808	{
771	IV offset = utf8_distance (dec.cur, SvPVX (string));
772	SV *uni = sv_newmortal ();	1809	SV *uni = sv_newmortal ();
773		1810
		1811	// horrible hack to silence warning inside pv_uni_display
		1812	COP cop = *PL_curcop;
		1813	cop.cop_warnings = pWARN_NONE;
		1814	ENTER;
		1815	SAVEVPTR (PL_curcop);
		1816	PL_curcop = &cop;
774	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1817	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
		1818	LEAVE;
		1819
775	croak ("%s, at character %d (%s)",	1820	croak ("%s, at character offset %d (before \"%s\")",
776	dec.err,	1821	dec.err,
777	(int)offset,	1822	(int)ptr_to_index (string, dec.cur),
778	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1823	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
779	}	1824	}
780		1825
781	sv = sv_2mortal (sv);	1826	sv = sv_2mortal (sv);
782		1827
783	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1828	if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv))
784	croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)");	1829	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
785		1830
786	return sv;	1831	return sv;
787	}	1832	}
788		1833
		1834	/////////////////////////////////////////////////////////////////////////////
		1835	// incremental parser
		1836
		1837	static void
		1838	incr_parse (JSON *self)
		1839	{
		1840	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1841
		1842	// the state machine here is a bit convoluted and could be simplified a lot
		1843	// but this would make it slower, so...
		1844
		1845	for (;;)
		1846	{
		1847	//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
		1848	switch (self->incr_mode)
		1849	{
		1850	// only used for initial whitespace skipping
		1851	case INCR_M_WS:
		1852	for (;;)
		1853	{
		1854	if (*p > 0x20)
		1855	{
		1856	if (*p == '#')
		1857	{
		1858	self->incr_mode = INCR_M_C0;
		1859	goto incr_m_c;
		1860	}
		1861	else
		1862	{
		1863	self->incr_mode = INCR_M_JSON;
		1864	goto incr_m_json;
		1865	}
		1866	}
		1867	else if (!*p)
		1868	goto interrupt;
		1869
		1870	++p;
		1871	}
		1872
		1873	// skip a single char inside a string (for \\-processing)
		1874	case INCR_M_BS:
		1875	if (!*p)
		1876	goto interrupt;
		1877
		1878	++p;
		1879	self->incr_mode = INCR_M_STR;
		1880	goto incr_m_str;
		1881
		1882	// inside #-style comments
		1883	case INCR_M_C0:
		1884	case INCR_M_C1:
		1885	incr_m_c:
		1886	for (;;)
		1887	{
		1888	if (*p == '\n')
		1889	{
		1890	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1891	break;
		1892	}
		1893	else if (!*p)
		1894	goto interrupt;
		1895
		1896	++p;
		1897	}
		1898
		1899	break;
		1900
		1901	// inside a string
		1902	case INCR_M_STR:
		1903	incr_m_str:
		1904	for (;;)
		1905	{
		1906	if (*p == '"')
		1907	{
		1908	++p;
		1909	self->incr_mode = INCR_M_JSON;
		1910
		1911	if (!self->incr_nest)
		1912	goto interrupt;
		1913
		1914	goto incr_m_json;
		1915	}
		1916	else if (*p == '\\')
		1917	{
		1918	++p; // "virtually" consumes character after \
		1919
		1920	if (!*p) // if at end of string we have to switch modes
		1921	{
		1922	self->incr_mode = INCR_M_BS;
		1923	goto interrupt;
		1924	}
		1925	}
		1926	else if (!*p)
		1927	goto interrupt;
		1928
		1929	++p;
		1930	}
		1931
		1932	// after initial ws, outside string
		1933	case INCR_M_JSON:
		1934	incr_m_json:
		1935	for (;;)
		1936	{
		1937	switch (*p++)
		1938	{
		1939	case 0:
		1940	--p;
		1941	goto interrupt;
		1942
		1943	case 0x09:
		1944	case 0x0a:
		1945	case 0x0d:
		1946	case 0x20:
		1947	if (!self->incr_nest)
		1948	{
		1949	--p; // do not eat the whitespace, let the next round do it
		1950	goto interrupt;
		1951	}
		1952	break;
		1953
		1954	case '"':
		1955	self->incr_mode = INCR_M_STR;
		1956	goto incr_m_str;
		1957
		1958	case '[':
		1959	case '{':
		1960	case '(':
		1961	if (++self->incr_nest > self->max_depth)
		1962	croak (ERR_NESTING_EXCEEDED);
		1963	break;
		1964
		1965	case ']':
		1966	case '}':
		1967	if (--self->incr_nest <= 0)
		1968	goto interrupt;
		1969	break;
		1970
		1971	case ')':
		1972	--self->incr_nest;
		1973	break;
		1974
		1975	case '#':
		1976	self->incr_mode = INCR_M_C1;
		1977	goto incr_m_c;
		1978	}
		1979	}
		1980	}
		1981
		1982	modechange:
		1983	;
		1984	}
		1985
		1986	interrupt:
		1987	self->incr_pos = p - SvPVX (self->incr_text);
		1988	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1989	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
		1990	}
		1991
		1992	/////////////////////////////////////////////////////////////////////////////
		1993	// XS interface functions
		1994
789	MODULE = JSON::XS PACKAGE = JSON::XS	1995	MODULE = JSON::XS PACKAGE = JSON::XS
790		1996
791	BOOT:	1997	BOOT:
792	{	1998	{
793	int i;	1999	int i;
794		2000
795	memset (decode_hexdigit, 0xff, 256);
796	for (i = 10; i--; )	2001	for (i = 0; i < 256; ++i)
797	decode_hexdigit ['0' + i] = i;	2002	decode_hexdigit [i] =
		2003	i >= '0' && i <= '9' ? i - '0'
		2004	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		2005	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		2006	: -1;
798		2007
799	for (i = 6; --i; )	2008	json_stash = gv_stashpv ("JSON::XS" , 1);
800	{	2009	bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
801	decode_hexdigit ['a' + i] = 10 + i;	2010	bool_false = get_bool ("Types::Serialiser::false");
802	decode_hexdigit ['A' + i] = 10 + i;	2011	bool_true = get_bool ("Types::Serialiser::true");
		2012
		2013	sv_json = newSVpv ("JSON", 0);
		2014	SvREADONLY_on (sv_json);
		2015
		2016	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
		2017	}
		2018
		2019	PROTOTYPES: DISABLE
		2020
		2021	void CLONE (...)
		2022	CODE:
		2023	// as long as these writes are atomic, the race should not matter
		2024	// as existing threads either already use 0, or use the old value,
		2025	// which is sitll correct for the initial thread.
		2026	json_stash = 0;
		2027	bool_stash = 0;
		2028	bool_false = 0;
		2029	bool_true = 0;
		2030
		2031	void new (char *klass)
		2032	PPCODE:
		2033	{
		2034	SV *pv = NEWSV (0, sizeof (JSON));
		2035	SvPOK_only (pv);
		2036	json_init ((JSON *)SvPVX (pv));
		2037	XPUSHs (sv_2mortal (sv_bless (
		2038	newRV_noinc (pv),
		2039	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		2040	)));
		2041	}
		2042
		2043	void boolean_values (JSON *self, SV *v_false = 0, SV *v_true = 0)
		2044	PPCODE:
		2045	self->v_false = newSVsv (v_false);
		2046	self->v_true = newSVsv (v_true);
		2047	XPUSHs (ST (0));
		2048
		2049	void get_boolean_values (JSON *self)
		2050	PPCODE:
		2051	if (self->v_false && self->v_true)
		2052	{
		2053	EXTEND (SP, 2);
		2054	PUSHs (self->v_false);
		2055	PUSHs (self->v_true);
803	}	2056	}
804		2057
805	json_stash = gv_stashpv ("JSON::XS", 1);	2058	void ascii (JSON *self, int enable = 1)
806	}	2059	ALIAS:
		2060	ascii = F_ASCII
		2061	latin1 = F_LATIN1
		2062	utf8 = F_UTF8
		2063	indent = F_INDENT
		2064	canonical = F_CANONICAL
		2065	space_before = F_SPACE_BEFORE
		2066	space_after = F_SPACE_AFTER
		2067	pretty = F_PRETTY
		2068	allow_nonref = F_ALLOW_NONREF
		2069	shrink = F_SHRINK
		2070	allow_blessed = F_ALLOW_BLESSED
		2071	convert_blessed = F_CONV_BLESSED
		2072	relaxed = F_RELAXED
		2073	allow_unknown = F_ALLOW_UNKNOWN
		2074	allow_tags = F_ALLOW_TAGS
		2075	PPCODE:
		2076	{
		2077	if (enable)
		2078	self->flags |= ix;
		2079	else
		2080	self->flags &= ~ix;
807		2081
808	SV *new (char *dummy)	2082	XPUSHs (ST (0));
		2083	}
		2084
		2085	void get_ascii (JSON *self)
		2086	ALIAS:
		2087	get_ascii = F_ASCII
		2088	get_latin1 = F_LATIN1
		2089	get_utf8 = F_UTF8
		2090	get_indent = F_INDENT
		2091	get_canonical = F_CANONICAL
		2092	get_space_before = F_SPACE_BEFORE
		2093	get_space_after = F_SPACE_AFTER
		2094	get_allow_nonref = F_ALLOW_NONREF
		2095	get_shrink = F_SHRINK
		2096	get_allow_blessed = F_ALLOW_BLESSED
		2097	get_convert_blessed = F_CONV_BLESSED
		2098	get_relaxed = F_RELAXED
		2099	get_allow_unknown = F_ALLOW_UNKNOWN
		2100	get_allow_tags = F_ALLOW_TAGS
		2101	PPCODE:
		2102	XPUSHs (boolSV (self->flags & ix));
		2103
		2104	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		2105	PPCODE:
		2106	self->max_depth = max_depth;
		2107	XPUSHs (ST (0));
		2108
		2109	U32 get_max_depth (JSON *self)
809	CODE:	2110	CODE:
810	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	2111	RETVAL = self->max_depth;
811	OUTPUT:	2112	OUTPUT:
812	RETVAL	2113	RETVAL
813		2114
814	SV *ascii (SV *self, int enable)	2115	void max_size (JSON *self, U32 max_size = 0)
815	ALIAS:	2116	PPCODE:
816	ascii = F_ASCII	2117	self->max_size = max_size;
817	utf8 = F_UTF8	2118	XPUSHs (ST (0));
818	indent = F_INDENT	2119
819	canonical = F_CANONICAL	2120	int get_max_size (JSON *self)
820	space_before = F_SPACE_BEFORE
821	space_after = F_SPACE_AFTER
822	json_rpc = F_JSON_RPC
823	pretty = F_PRETTY
824	allow_nonref = F_ALLOW_NONREF
825	CODE:	2121	CODE:
826	{	2122	RETVAL = self->max_size;
827	UV *uv = SvJSON (self);
828	if (enable)
829	*uv |= ix;
830	else
831	*uv &= ~ix;
832
833	RETVAL = newSVsv (self);
834	}
835	OUTPUT:	2123	OUTPUT:
836	RETVAL	2124	RETVAL
837		2125
838	void encode (SV *self, SV *scalar)	2126	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
839	PPCODE:	2127	PPCODE:
840	XPUSHs (encode_json (scalar, *SvJSON (self)));	2128	{
		2129	SvREFCNT_dec (self->cb_object);
		2130	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
841		2131
842	void decode (SV *self, SV *jsonstr)	2132	XPUSHs (ST (0));
		2133	}
		2134
		2135	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
843	PPCODE:	2136	PPCODE:
844	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	2137	{
		2138	if (!self->cb_sk_object)
		2139	self->cb_sk_object = newHV ();
845		2140
846	void to_json (SV *scalar)	2141	if (SvOK (cb))
		2142	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		2143	else
		2144	{
		2145	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		2146
		2147	if (!HvKEYS (self->cb_sk_object))
		2148	{
		2149	SvREFCNT_dec (self->cb_sk_object);
		2150	self->cb_sk_object = 0;
		2151	}
		2152	}
		2153
		2154	XPUSHs (ST (0));
		2155	}
		2156
		2157	void encode (JSON *self, SV *scalar)
847	PPCODE:	2158	PPCODE:
848	XPUSHs (encode_json (scalar, F_UTF8));	2159	PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
		2160	XPUSHs (scalar);
849		2161
850	void from_json (SV *jsonstr)	2162	void decode (JSON *self, SV *jsonstr)
851	PPCODE:	2163	PPCODE:
852	XPUSHs (decode_json (jsonstr, F_UTF8));	2164	PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
		2165	XPUSHs (jsonstr);
853		2166
		2167	void decode_prefix (JSON *self, SV *jsonstr)
		2168	PPCODE:
		2169	{
		2170	SV *sv;
		2171	STRLEN offset;
		2172	PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
		2173	EXTEND (SP, 2);
		2174	PUSHs (sv);
		2175	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset))));
		2176	}
		2177
		2178	void incr_parse (JSON *self, SV *jsonstr = 0)
		2179	PPCODE:
		2180	{
		2181	if (!self->incr_text)
		2182	self->incr_text = newSVpvn ("", 0);
		2183
		2184	/* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
		2185	if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
		2186	if (DECODE_WANTS_OCTETS (self))
		2187	{
		2188	if (self->incr_pos)
		2189	self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
		2190	(U8 *)SvPVX (self->incr_text) + self->incr_pos);
		2191
		2192	sv_utf8_downgrade (self->incr_text, 0);
		2193	}
		2194	else
		2195	{
		2196	sv_utf8_upgrade (self->incr_text);
		2197
		2198	if (self->incr_pos)
		2199	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		2200	- (U8 *)SvPVX (self->incr_text);
		2201	}
		2202
		2203	// append data, if any
		2204	if (jsonstr)
		2205	{
		2206	/* make sure both strings have same encoding */
		2207	if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
		2208	if (SvUTF8 (jsonstr))
		2209	sv_utf8_downgrade (jsonstr, 0);
		2210	else
		2211	sv_utf8_upgrade (jsonstr);
		2212
		2213	/* and then just blindly append */
		2214	{
		2215	STRLEN len;
		2216	const char *str = SvPV (jsonstr, len);
		2217	STRLEN cur = SvCUR (self->incr_text);
		2218
		2219	if (SvLEN (self->incr_text) - cur <= len)
		2220	json_sv_grow (self->incr_text, cur, len);
		2221
		2222	Move (str, SvEND (self->incr_text), len, char);
		2223	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		2224	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		2225	}
		2226	}
		2227
		2228	if (GIMME_V != G_VOID)
		2229	do
		2230	{
		2231	SV *sv;
		2232	STRLEN offset;
		2233
		2234	if (!INCR_DONE (self))
		2235	{
		2236	incr_parse (self);
		2237
		2238	if (self->incr_pos > self->max_size && self->max_size)
		2239	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		2240	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		2241
		2242	if (!INCR_DONE (self))
		2243	{
		2244	// as an optimisation, do not accumulate white space in the incr buffer
		2245	if (self->incr_mode == INCR_M_WS && self->incr_pos)
		2246	{
		2247	self->incr_pos = 0;
		2248	SvCUR_set (self->incr_text, 0);
		2249	}
		2250
		2251	break;
		2252	}
		2253	}
		2254
		2255	PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
		2256	XPUSHs (sv);
		2257
		2258	self->incr_pos -= offset;
		2259	self->incr_nest = 0;
		2260	self->incr_mode = 0;
		2261
		2262	sv_chop (self->incr_text, SvPVX (self->incr_text) + offset);
		2263	}
		2264	while (GIMME_V == G_ARRAY);
		2265	}
		2266
		2267	SV *incr_text (JSON *self)
		2268	ATTRS: lvalue
		2269	CODE:
		2270	{
		2271	if (self->incr_pos)
		2272	croak ("incr_text can not be called when the incremental parser already started parsing");
		2273
		2274	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		2275	}
		2276	OUTPUT:
		2277	RETVAL
		2278
		2279	void incr_skip (JSON *self)
		2280	CODE:
		2281	{
		2282	if (self->incr_pos)
		2283	{
		2284	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2285	self->incr_pos = 0;
		2286	self->incr_nest = 0;
		2287	self->incr_mode = 0;
		2288	}
		2289	}
		2290
		2291	void incr_reset (JSON *self)
		2292	CODE:
		2293	{
		2294	SvREFCNT_dec (self->incr_text);
		2295	self->incr_text = 0;
		2296	self->incr_pos = 0;
		2297	self->incr_nest = 0;
		2298	self->incr_mode = 0;
		2299	}
		2300
		2301	void DESTROY (JSON *self)
		2302	CODE:
		2303	SvREFCNT_dec (self->v_false);
		2304	SvREFCNT_dec (self->v_true);
		2305	SvREFCNT_dec (self->cb_sk_object);
		2306	SvREFCNT_dec (self->cb_object);
		2307	SvREFCNT_dec (self->incr_text);
		2308
		2309	PROTOTYPES: ENABLE
		2310
		2311	void encode_json (SV *scalar)
		2312	PPCODE:
		2313	{
		2314	JSON json;
		2315	json_init (&json);
		2316	json.flags |= F_UTF8;
		2317	PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
		2318	XPUSHs (scalar);
		2319	}
		2320
		2321	void decode_json (SV *jsonstr)
		2322	PPCODE:
		2323	{
		2324	JSON json;
		2325	json_init (&json);
		2326	json.flags |= F_UTF8;
		2327	PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
		2328	XPUSHs (jsonstr);
		2329	}
		2330

Diff Legend

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