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 1.138 by root, Wed Mar 6 07:21:17 2019 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_TFN, // inside true/false/null
		102	INCR_M_NUM, // inside number
		103	INCR_M_STR, // inside string
		104	INCR_M_BS, // inside backslash
		105	INCR_M_C0, // inside comment in initial whitespace sequence
		106	INCR_M_C1, // inside comment in other places
		107	INCR_M_JSON // outside anything, count nesting
		108	};
		109
		110	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		111
		112	typedef struct {
		113	U32 flags;
		114	U32 max_depth;
		115	STRLEN max_size;
		116
		117	SV *cb_object;
		118	HV *cb_sk_object;
		119
		120	// for the incremental parser
		121	SV *incr_text; // the source text so far
		122	STRLEN incr_pos; // the current offset into the text
		123	int incr_nest; // {[]}-nesting level
		124	unsigned char incr_mode;
		125
		126	SV *v_false, *v_true;
		127	} JSON;
		128
		129	INLINE void
		130	json_init (JSON *json)
		131	{
		132	static const JSON init = { F_ALLOW_NONREF, 512 };
		133
		134	*json = init;
		135	}
		136
		137	/////////////////////////////////////////////////////////////////////////////
		138	// utility functions
		139
		140	INLINE SV *
		141	get_bool (const char *name)
		142	{
		143	SV *sv = get_sv (name, 1);
		144
		145	SvREADONLY_on (sv);
		146	SvREADONLY_on (SvRV (sv));
		147
		148	return sv;
		149	}
		150
		151	INLINE void
		152	shrink (SV *sv)
		153	{
		154	sv_utf8_downgrade (sv, 1);
		155
		156	if (SvLEN (sv) > SvCUR (sv) + 1)
		157	{
		158	#ifdef SvPV_shrink_to_cur
		159	SvPV_shrink_to_cur (sv);
		160	#elif defined (SvPV_renew)
		161	SvPV_renew (sv, SvCUR (sv) + 1);
		162	#endif
		163	}
		164	}
		165
		166	/* adds two STRLENs together, slow, and with paranoia */
		167	STRLEN
		168	strlen_sum (STRLEN l1, STRLEN l2)
		169	{
		170	size_t sum = l1 + l2;
		171
		172	if (sum < (size_t)l2 || sum != (size_t)(STRLEN)sum)
		173	croak ("JSON::XS: string size overflow");
		174
		175	return sum;
		176	}
		177
		178	/* similar to SvGROW, but somewhat safer and guarantees exponential realloc strategy */
		179	static char *
		180	json_sv_grow (SV *sv, size_t len1, size_t len2)
		181	{
		182	len1 = strlen_sum (len1, len2);
		183	len1 = strlen_sum (len1, len1 >> 1);
		184
		185	if (len1 > 4096 - 24)
		186	len1 = (len1 | 4095) - 24;
		187
		188	return SvGROW (sv, len1);
		189	}
		190
		191	// decode a utf-8 character and return it, or (UV)-1 in
		192	// case of an error.
		193	// we special-case "safe" characters from U+80 .. U+7FF,
		194	// but use the very good perl function to parse anything else.
		195	// note that we never call this function for a ascii codepoints
		196	INLINE UV
		197	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		198	{
		199	if (expect_true (len >= 2
		200	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
		201	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
		202	{
		203	*clen = 2;
		204	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		205	}
		206	else
		207	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		208	}
		209
		210	// likewise for encoding, also never called for ascii codepoints
		211	// this function takes advantage of this fact, although current gccs
		212	// seem to optimise the check for >= 0x80 away anyways
		213	INLINE unsigned char *
		214	encode_utf8 (unsigned char *s, UV ch)
		215	{
		216	if (expect_false (ch < 0x000080))
		217	*s++ = ch;
		218	else if (expect_true (ch < 0x000800))
		219	*s++ = 0xc0 | ( ch >> 6),
		220	*s++ = 0x80 | ( ch & 0x3f);
		221	else if ( ch < 0x010000)
		222	*s++ = 0xe0 | ( ch >> 12),
		223	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		224	*s++ = 0x80 | ( ch & 0x3f);
		225	else if ( ch < 0x110000)
		226	*s++ = 0xf0 | ( ch >> 18),
		227	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		228	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		229	*s++ = 0x80 | ( ch & 0x3f);
		230
		231	return s;
		232	}
		233
		234	// convert offset pointer to character index, sv must be string
		235	static STRLEN
		236	ptr_to_index (SV *sv, char *offset)
		237	{
		238	return SvUTF8 (sv)
		239	? utf8_distance (offset, SvPVX (sv))
		240	: offset - SvPVX (sv);
		241	}
		242
		243	/////////////////////////////////////////////////////////////////////////////
		244	// fp hell
		245
		246	// scan a group of digits, and a trailing exponent
		247	static void
		248	json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
		249	{
		250	UV uaccum = 0;
		251	int eaccum = 0;
		252
		253	// if we recurse too deep, skip all remaining digits
		254	// to avoid a stack overflow attack
		255	if (expect_false (--maxdepth <= 0))
		256	while (((U8)*s - '0') < 10)
		257	++s;
		258
		259	for (;;)
		260	{
		261	U8 dig = (U8)*s - '0';
		262
		263	if (expect_false (dig >= 10))
		264	{
		265	if (dig == (U8)((U8)'.' - (U8)'0'))
		266	{
		267	++s;
		268	json_atof_scan1 (s, accum, expo, 1, maxdepth);
		269	}
		270	else if ((dig | ' ') == 'e' - '0')
		271	{
		272	int exp2 = 0;
		273	int neg = 0;
		274
		275	++s;
		276
		277	if (*s == '-')
		278	{
		279	++s;
		280	neg = 1;
		281	}
		282	else if (*s == '+')
		283	++s;
		284
		285	while ((dig = (U8)*s - '0') < 10)
		286	exp2 = exp2 * 10 + *s++ - '0';
		287
		288	*expo += neg ? -exp2 : exp2;
		289	}
		290
		291	break;
		292	}
		293
		294	++s;
		295
		296	uaccum = uaccum * 10 + dig;
		297	++eaccum;
		298
		299	// if we have too many digits, then recurse for more
		300	// we actually do this for rather few digits
		301	if (uaccum >= (UV_MAX - 9) / 10)
		302	{
		303	if (postdp) *expo -= eaccum;
		304	json_atof_scan1 (s, accum, expo, postdp, maxdepth);
		305	if (postdp) *expo += eaccum;
		306
		307	break;
		308	}
		309	}
		310
		311	// this relies greatly on the quality of the pow ()
		312	// implementation of the platform, but a good
		313	// implementation is hard to beat.
		314	// (IEEE 754 conformant ones are required to be exact)
		315	if (postdp) *expo -= eaccum;
		316	*accum += uaccum * Perl_pow (10., *expo);
		317	*expo += eaccum;
		318	}
		319
		320	static NV
		321	json_atof (const char *s)
		322	{
		323	NV accum = 0.;
		324	int expo = 0;
		325	int neg = 0;
		326
		327	if (*s == '-')
		328	{
		329	++s;
		330	neg = 1;
		331	}
		332
		333	// a recursion depth of ten gives us >>500 bits
		334	json_atof_scan1 (s, &accum, &expo, 0, 10);
		335
		336	return neg ? -accum : accum;
		337	}
		338
		339	// target of scalar reference is bool? -1 == nope, 0 == false, 1 == true
		340	static int
		341	ref_bool_type (SV *sv)
		342	{
		343	svtype svt = SvTYPE (sv);
		344
		345	if (svt < SVt_PVAV)
		346	{
		347	STRLEN len = 0;
		348	char *pv = svt ? SvPV (sv, len) : 0;
		349
		350	if (len == 1)
		351	if (*pv == '1')
		352	return 1;
		353	else if (*pv == '0')
		354	return 0;
		355	}
		356
		357	return -1;
		358	}
		359
		360	// returns whether scalar is not a reference in the sense of allow_nonref
		361	static int
		362	json_nonref (SV *scalar)
		363	{
		364	if (!SvROK (scalar))
		365	return 1;
		366
		367	scalar = SvRV (scalar);
		368
		369	if (SvTYPE (scalar) >= SVt_PVMG)
		370	{
		371	if (SvSTASH (scalar) == bool_stash)
		372	return 1;
		373
		374	if (!SvOBJECT (scalar) && ref_bool_type (scalar) >= 0)
		375	return 1;
		376	}
		377
		378	return 0;
		379	}
		380
		381	/////////////////////////////////////////////////////////////////////////////
		382	// encoder
27		383
28	// structure used for encoding JSON	384	// structure used for encoding JSON
29	typedef struct	385	typedef struct
30	{	386	{
31	char *cur;	387	char *cur; // SvPVX (sv) + current output position
32	STRLEN len; // SvLEN (sv)
33	char *end; // SvEND (sv)	388	char *end; // SvEND (sv)
34	SV *sv;	389	SV *sv; // result scalar
35	UV flags;	390	JSON json;
36	int max_recurse;	391	U32 indent; // indentation level
37	int indent;	392	UV limit; // escape character values >= this value when encoding
38	} enc_t;	393	} enc_t;
39		394
40	// structure used for decoding JSON	395	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)	396	need (enc_t *enc, STRLEN len)
62	{	397	{
63	if (enc->cur + len >= enc->end)	398	if (expect_false ((uintptr_t)(enc->end - enc->cur) < len))
64	{	399	{
65	STRLEN cur = enc->cur - SvPVX (enc->sv);	400	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
66	SvGROW (enc->sv, cur + len + 1);	401	char *buf = json_sv_grow (enc->sv, cur, len);
67	enc->cur = SvPVX (enc->sv) + cur;	402	enc->cur = buf + cur;
68	enc->end = SvEND (enc->sv);	403	enc->end = buf + SvLEN (enc->sv) - 1;
69	}	404	}
70	}	405	}
71		406
72	static void	407	INLINE void
73	encode_ch (enc_t *enc, char ch)	408	encode_ch (enc_t *enc, char ch)
74	{	409	{
75	need (enc, 1);	410	need (enc, 1);
76	*enc->cur++ = ch;	411	*enc->cur++ = ch;
77	}	412	}
…		…
79	static void	414	static void
80	encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)	415	encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)
81	{	416	{
82	char *end = str + len;	417	char *end = str + len;
83		418
		419	need (enc, len);
		420
84	while (str < end)	421	while (str < end)
85	{	422	{
86	unsigned char ch = *(unsigned char *)str;	423	unsigned char ch = *(unsigned char *)str;
		424
87	if (ch >= 0x20 && ch < 0x80) // most common case	425	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
88	{	426	{
		427	if (expect_false (ch == '"')) // but with slow exceptions
		428	{
		429	need (enc, len + 1);
		430	*enc->cur++ = '\\';
		431	*enc->cur++ = '"';
		432	}
		433	else if (expect_false (ch == '\\'))
		434	{
		435	need (enc, len + 1);
		436	*enc->cur++ = '\\';
		437	*enc->cur++ = '\\';
		438	}
		439	else
89	*enc->cur++ = ch;	440	*enc->cur++ = ch;
		441
90	str++;	442	++str;
91	}	443	}
92	else	444	else
93	{	445	{
94	STRLEN clen;	446	switch (ch)
95	UV uch;
96
97	if (is_utf8)
98	{	447	{
99	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	448	case '\010': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
100	if (clen < 0)	449	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");	450	case '\012': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
102	}	451	case '\014': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
103	else	452	case '\015': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
104	{
105	uch = ch;
106	clen = 1;
107	}
108		453
109	need (enc, len += 6);	454	default:
110
111	if (uch < 0xa0 || enc->flags & F_ASCII)
112	{
113	if (uch > 0xFFFFUL)
114	{	455	{
		456	STRLEN clen;
		457	UV uch;
		458
		459	if (is_utf8)
		460	{
		461	uch = decode_utf8 (str, end - str, &clen);
		462	if (clen == (STRLEN)-1)
		463	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
		464	}
		465	else
		466	{
		467	uch = ch;
115	len += 6;	468	clen = 1;
		469	}
		470
		471	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
		472	{
		473	if (uch >= 0x10000UL)
		474	{
		475	if (uch >= 0x110000UL)
		476	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		477
116	need (enc, len += 6);	478	need (enc, len + 11);
117	sprintf (enc->cur, "\\u%04x\\u%04x",	479	sprintf (enc->cur, "\\u%04x\\u%04x",
118	(uch - 0x10000) / 0x400 + 0xD800,	480	(int)((uch - 0x10000) / 0x400 + 0xD800),
119	(uch - 0x10000) % 0x400 + 0xDC00);	481	(int)((uch - 0x10000) % 0x400 + 0xDC00));
120	enc->cur += 12;	482	enc->cur += 12;
		483	}
		484	else
		485	{
		486	need (enc, len + 5);
		487	*enc->cur++ = '\\';
		488	*enc->cur++ = 'u';
		489	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
		490	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
		491	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
		492	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
		493	}
		494
		495	str += clen;
121	}	496	}
		497	else if (enc->json.flags & F_LATIN1)
		498	{
		499	*enc->cur++ = uch;
		500	str += clen;
		501	}
		502	else if (is_utf8)
		503	{
		504	need (enc, len + clen);
		505	do
		506	{
		507	*enc->cur++ = *str++;
		508	}
		509	while (--clen);
		510	}
122	else	511	else
123	{	512	{
124	sprintf (enc->cur, "\\u%04x", uch);	513	need (enc, len + UTF8_MAXBYTES - 1); // never more than 11 bytes needed
125	enc->cur += 6;	514	enc->cur = encode_utf8 (enc->cur, uch);
		515	++str;
		516	}
126	}	517	}
127	}	518	}
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	}	519	}
138		520
139	--len;	521	--len;
140	}	522	}
141	}	523	}
142		524
143	#define INDENT SB \	525	INLINE void
		526	encode_indent (enc_t *enc)
		527	{
144	if (enc->flags & F_INDENT) \	528	if (enc->json.flags & F_INDENT)
145	{ \	529	{
146	int i_; \	530	int spaces = enc->indent * INDENT_STEP;
147	need (enc, enc->indent); \	531
148	for (i_ = enc->indent * 3; i_--; )\	532	need (enc, spaces);
		533	memset (enc->cur, ' ', spaces);
		534	enc->cur += spaces;
		535	}
		536	}
		537
		538	INLINE void
		539	encode_space (enc_t *enc)
		540	{
		541	need (enc, 1);
		542	encode_ch (enc, ' ');
		543	}
		544
		545	INLINE void
		546	encode_nl (enc_t *enc)
		547	{
		548	if (enc->json.flags & F_INDENT)
		549	{
		550	need (enc, 1);
149	encode_ch (enc, ' '); \	551	encode_ch (enc, '\n');
150	} \	552	}
151	SE	553	}
152		554
153	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	555	INLINE void
154	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	556	encode_comma (enc_t *enc)
155	#define COMMA SB \	557	{
156	encode_ch (enc, ','); \	558	encode_ch (enc, ',');
		559
157	if (enc->flags & F_INDENT) \	560	if (enc->json.flags & F_INDENT)
158	NL; \	561	encode_nl (enc);
159	else if (enc->flags & F_SPACE_AFTER) \	562	else if (enc->json.flags & F_SPACE_AFTER)
160	SPACE; \	563	encode_space (enc);
161	SE	564	}
162		565
163	static void encode_sv (enc_t *enc, SV *sv);	566	static void encode_sv (enc_t *enc, SV *sv);
164		567
165	static void	568	static void
166	encode_av (enc_t *enc, AV *av)	569	encode_av (enc_t *enc, AV *av)
167	{	570	{
168	int i, len = av_len (av);	571	int i, len = av_len (av);
169		572
		573	if (enc->indent >= enc->json.max_depth)
		574	croak (ERR_NESTING_EXCEEDED);
		575
170	encode_ch (enc, '['); NL;	576	encode_ch (enc, '[');
171	++enc->indent;
172		577
		578	if (len >= 0)
		579	{
		580	encode_nl (enc); ++enc->indent;
		581
173	for (i = 0; i <= len; ++i)	582	for (i = 0; i <= len; ++i)
174	{	583	{
175	INDENT;	584	SV **svp = av_fetch (av, i, 0);
176	encode_sv (enc, *av_fetch (av, i, 0));
177		585
		586	encode_indent (enc);
		587
		588	if (svp)
		589	encode_sv (enc, *svp);
		590	else
		591	encode_str (enc, "null", 4, 0);
		592
178	if (i < len)	593	if (i < len)
179	COMMA;	594	encode_comma (enc);
180	}	595	}
181		596
182	NL;	597	encode_nl (enc); --enc->indent; encode_indent (enc);
		598	}
183		599
184	--enc->indent;
185	INDENT; encode_ch (enc, ']');	600	encode_ch (enc, ']');
186	}	601	}
187		602
188	static void	603	static void
189	encode_he (enc_t *enc, HE *he)	604	encode_hk (enc_t *enc, HE *he)
190	{	605	{
191	encode_ch (enc, '"');	606	encode_ch (enc, '"');
192		607
193	if (HeKLEN (he) == HEf_SVKEY)	608	if (HeKLEN (he) == HEf_SVKEY)
194	{	609	{
195	SV *sv = HeSVKEY (he);	610	SV *sv = HeSVKEY (he);
196	STRLEN len;	611	STRLEN len;
		612	char *str;
		613
		614	SvGETMAGIC (sv);
197	char *str = SvPV (sv, len);	615	str = SvPV (sv, len);
198		616
199	encode_str (enc, str, len, SvUTF8 (sv));	617	encode_str (enc, str, len, SvUTF8 (sv));
200	}	618	}
201	else	619	else
202	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	620	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
203		621
204	encode_ch (enc, '"');	622	encode_ch (enc, '"');
205		623
206	if (enc->flags & F_SPACE_BEFORE) SPACE;	624	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
207	encode_ch (enc, ':');	625	encode_ch (enc, ':');
208	if (enc->flags & F_SPACE_AFTER ) SPACE;	626	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
209	encode_sv (enc, HeVAL (he));
210	}	627	}
211		628
212	// compare hash entries, used when all keys are bytestrings	629	// compare hash entries, used when all keys are bytestrings
213	static int	630	static int
214	he_cmp_fast (const void *a_, const void *b_)	631	he_cmp_fast (const void *a_, const void *b_)
…		…
219	HE *b = *(HE **)b_;	636	HE *b = *(HE **)b_;
220		637
221	STRLEN la = HeKLEN (a);	638	STRLEN la = HeKLEN (a);
222	STRLEN lb = HeKLEN (b);	639	STRLEN lb = HeKLEN (b);
223		640
224	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	641	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
225	cmp = la < lb ? -1 : la == lb ? 0 : 1;	642	cmp = lb - la;
226		643
227	return cmp;	644	return cmp;
228	}	645	}
229		646
230	// compare hash entries, used when some keys are sv's or utf-x	647	// compare hash entries, used when some keys are sv's or utf-x
231	static int	648	static int
232	he_cmp_slow (const void *a, const void *b)	649	he_cmp_slow (const void *a, const void *b)
233	{	650	{
234	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	651	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
235	}	652	}
236		653
237	static void	654	static void
238	encode_hv (enc_t *enc, HV *hv)	655	encode_hv (enc_t *enc, HV *hv)
239	{	656	{
240	int count, i;	657	HE *he;
241		658
242	encode_ch (enc, '{'); NL; ++enc->indent;	659	if (enc->indent >= enc->json.max_depth)
		660	croak (ERR_NESTING_EXCEEDED);
243		661
244	if ((count = hv_iterinit (hv)))	662	encode_ch (enc, '{');
245	{	663
246	// for canonical output we have to sort by keys first	664	// for canonical output we have to sort by keys first
247	// actually, this is mostly due to the stupid so-called	665	// actually, this is mostly due to the stupid so-called
248	// security workaround added somewhere in 5.8.x.	666	// security workaround added somewhere in 5.8.x
249	// that randomises hash orderings	667	// that randomises hash orderings
250	if (enc->flags & F_CANONICAL)	668	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
		669	{
		670	int count = hv_iterinit (hv);
		671
		672	if (SvMAGICAL (hv))
251	{	673	{
252	HE *he, *hes [count];	674	// need to count by iterating. could improve by dynamically building the vector below
		675	// but I don't care for the speed of this special case.
		676	// note also that we will run into undefined behaviour when the two iterations
		677	// do not result in the same count, something I might care for in some later release.
		678
		679	count = 0;
		680	while (hv_iternext (hv))
		681	++count;
		682
		683	hv_iterinit (hv);
		684	}
		685
		686	if (count)
		687	{
253	int fast = 1;	688	int i, fast = 1;
		689	HE *hes_stack [STACK_HES];
		690	HE **hes = hes_stack;
		691
		692	// allocate larger arrays on the heap
		693	if (count > STACK_HES)
		694	{
		695	SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
		696	hes = (HE **)SvPVX (sv);
		697	}
254		698
255	i = 0;	699	i = 0;
256	while ((he = hv_iternext (hv)))	700	while ((he = hv_iternext (hv)))
257	{	701	{
258	hes [i++] = he;	702	hes [i++] = he;
…		…
264		708
265	if (fast)	709	if (fast)
266	qsort (hes, count, sizeof (HE *), he_cmp_fast);	710	qsort (hes, count, sizeof (HE *), he_cmp_fast);
267	else	711	else
268	{	712	{
269	// hack to disable "use bytes"	713	// hack to forcefully disable "use bytes"
270	COP *oldcop = PL_curcop, cop;	714	COP cop = *PL_curcop;
271	cop.op_private = 0;	715	cop.op_private = 0;
		716
		717	ENTER;
		718	SAVETMPS;
		719
		720	SAVEVPTR (PL_curcop);
272	PL_curcop = &cop;	721	PL_curcop = &cop;
273		722
274	SAVETMPS;
275	qsort (hes, count, sizeof (HE *), he_cmp_slow);	723	qsort (hes, count, sizeof (HE *), he_cmp_slow);
		724
276	FREETMPS;	725	FREETMPS;
277		726	LEAVE;
278	PL_curcop = oldcop;
279	}	727	}
280		728
281	for (i = 0; i < count; ++i)	729	encode_nl (enc); ++enc->indent;
		730
		731	while (count--)
282	{	732	{
283	INDENT;	733	encode_indent (enc);
		734	he = hes [count];
284	encode_he (enc, hes [i]);	735	encode_hk (enc, he);
		736	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
285		737
286	if (i < count - 1)	738	if (count)
287	COMMA;	739	encode_comma (enc);
288	}	740	}
289		741
290	NL;	742	encode_nl (enc); --enc->indent; encode_indent (enc);
291	}	743	}
		744	}
		745	else
		746	{
		747	if (hv_iterinit (hv) || SvMAGICAL (hv))
		748	if ((he = hv_iternext (hv)))
		749	{
		750	encode_nl (enc); ++enc->indent;
		751
		752	for (;;)
		753	{
		754	encode_indent (enc);
		755	encode_hk (enc, he);
		756	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		757
		758	if (!(he = hv_iternext (hv)))
		759	break;
		760
		761	encode_comma (enc);
		762	}
		763
		764	encode_nl (enc); --enc->indent; encode_indent (enc);
		765	}
		766	}
		767
		768	encode_ch (enc, '}');
		769	}
		770
		771	// encode objects, arrays and special \0=false and \1=true values.
		772	static void
		773	encode_rv (enc_t *enc, SV *sv)
		774	{
		775	svtype svt;
		776	GV *method;
		777
		778	SvGETMAGIC (sv);
		779	svt = SvTYPE (sv);
		780
		781	if (expect_false (SvOBJECT (sv)))
		782	{
		783	HV *stash = SvSTASH (sv);
		784
		785	if (stash == bool_stash)
		786	{
		787	if (SvIV (sv)) encode_str (enc, "true" , 4, 0);
		788	else encode_str (enc, "false", 5, 0);
		789	}
		790	else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
		791	{
		792	int count;
		793	dSP;
		794
		795	ENTER; SAVETMPS;
		796	PUSHMARK (SP);
		797	EXTEND (SP, 2);
		798	// we re-bless the reference to get overload and other niceties right
		799	PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		800	PUSHs (sv_json);
		801
		802	PUTBACK;
		803	count = call_sv ((SV *)GvCV (method), G_ARRAY);
		804	SPAGAIN;
		805
		806	// catch this surprisingly common error
		807	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		808	croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		809
		810	encode_ch (enc, '(');
		811	encode_ch (enc, '"');
		812	encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
		813	encode_ch (enc, '"');
		814	encode_ch (enc, ')');
		815	encode_ch (enc, '[');
		816
		817	if (count)
		818	{
		819	int i;
		820
		821	for (i = 0; i < count - 1; ++i)
		822	{
		823	encode_sv (enc, SP[i + 1 - count]);
		824	encode_ch (enc, ',');
		825	}
		826
		827	encode_sv (enc, TOPs);
		828	SP -= count;
		829	}
		830
		831	encode_ch (enc, ']');
		832
		833	FREETMPS; LEAVE;
		834	}
		835	else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
		836	{
		837	dSP;
		838
		839	ENTER; SAVETMPS;
		840	PUSHMARK (SP);
		841	// we re-bless the reference to get overload and other niceties right
		842	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		843
		844	// calling with G_SCALAR ensures that we always get a 1 return value
		845	PUTBACK;
		846	call_sv ((SV *)GvCV (method), G_SCALAR);
		847	SPAGAIN;
		848
		849	// catch this surprisingly common error
		850	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		851	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		852
		853	sv = POPs;
		854	PUTBACK;
		855
		856	encode_sv (enc, sv);
		857
		858	FREETMPS; LEAVE;
		859	}
		860	else if (enc->json.flags & F_ALLOW_BLESSED)
		861	encode_str (enc, "null", 4, 0);
292	else	862	else
293	{	863	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;	864	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
295	HE *he = hv_iternext (hv);	865	}
296		866	else if (svt == SVt_PVHV)
297	for (;;)	867	encode_hv (enc, (HV *)sv);
298	{	868	else if (svt == SVt_PVAV)
299	INDENT;	869	encode_av (enc, (AV *)sv);
300	encode_he (enc, he);	870	else if (svt < SVt_PVAV)
301
302	if (!(he = hv_iternext (hv)))
303	break;
304
305	COMMA;
306	}
307
308	NL;
309	}
310	}	871	{
		872	int bool_type = ref_bool_type (sv);
311		873
312	--enc->indent; INDENT; encode_ch (enc, '}');	874	if (bool_type == 1)
		875	encode_str (enc, "true", 4, 0);
		876	else if (bool_type == 0)
		877	encode_str (enc, "false", 5, 0);
		878	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		879	encode_str (enc, "null", 4, 0);
		880	else
		881	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		882	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		883	}
		884	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		885	encode_str (enc, "null", 4, 0);
		886	else
		887	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		888	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
313	}	889	}
314		890
315	static void	891	static void
316	encode_sv (enc_t *enc, SV *sv)	892	encode_sv (enc_t *enc, SV *sv)
317	{	893	{
		894	SvGETMAGIC (sv);
		895
318	if (SvPOKp (sv))	896	if (SvPOKp (sv))
319	{	897	{
320	STRLEN len;	898	STRLEN len;
321	char *str = SvPV (sv, len);	899	char *str = SvPV (sv, len);
322	encode_ch (enc, '"');	900	encode_ch (enc, '"');
323	encode_str (enc, str, len, SvUTF8 (sv));	901	encode_str (enc, str, len, SvUTF8 (sv));
324	encode_ch (enc, '"');	902	encode_ch (enc, '"');
325	}	903	}
326	else if (SvNOKp (sv))	904	else if (SvNOKp (sv))
327	{	905	{
		906	// trust that perl will do the right thing w.r.t. JSON syntax.
328	need (enc, NV_DIG + 32);	907	need (enc, NV_DIG + 32);
329	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	908	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
330	enc->cur += strlen (enc->cur);	909	enc->cur += strlen (enc->cur);
331	}	910	}
332	else if (SvIOKp (sv))	911	else if (SvIOKp (sv))
333	{	912	{
		913	// we assume we can always read an IV as a UV and vice versa
		914	// we assume two's complement
		915	// we assume no aliasing issues in the union
		916	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		917	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		918	{
		919	// optimise the "small number case"
		920	// code will likely be branchless and use only a single multiplication
		921	// works for numbers up to 59074
		922	I32 i = SvIVX (sv);
		923	U32 u;
		924	char digit, nz = 0;
		925
334	need (enc, 64);	926	need (enc, 6);
		927
		928	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		929	u = i < 0 ? -i : i;
		930
		931	// convert to 4.28 fixed-point representation
		932	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		933
		934	// now output digit by digit, each time masking out the integer part
		935	// and multiplying by 5 while moving the decimal point one to the right,
		936	// resulting in a net multiplication by 10.
		937	// we always write the digit to memory but conditionally increment
		938	// the pointer, to enable the use of conditional move instructions.
		939	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		940	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		941	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		942	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		943	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		944	}
		945	else
		946	{
		947	// large integer, use the (rather slow) snprintf way.
		948	need (enc, IVUV_MAXCHARS);
335	enc->cur +=	949	enc->cur +=
336	SvIsUV(sv)	950	SvIsUV(sv)
337	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	951	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
338	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	952	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		953	}
339	}	954	}
340	else if (SvROK (sv))	955	else if (SvROK (sv))
341	{	956	encode_rv (enc, SvRV (sv));
342	if (!--enc->max_recurse)	957	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);	958	encode_str (enc, "null", 4, 0);
358	else	959	else
359	croak ("encountered perl type that JSON cannot handle");	960	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
		961	SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
360	}	962	}
361		963
362	static SV *	964	static SV *
363	encode_json (SV *scalar, UV flags)	965	encode_json (SV *scalar, JSON *json)
364	{	966	{
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;	967	enc_t enc;
369	enc.flags = flags;	968
		969	if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar))
		970	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
		971
		972	enc.json = *json;
370	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	973	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
371	enc.cur = SvPVX (enc.sv);	974	enc.cur = SvPVX (enc.sv);
372	enc.end = SvEND (enc.sv);	975	enc.end = SvEND (enc.sv);
373	enc.max_recurse = 0;
374	enc.indent = 0;	976	enc.indent = 0;
		977	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		978	: enc.json.flags & F_LATIN1 ? 0x000100UL
		979	: 0x110000UL;
375		980
376	SvPOK_only (enc.sv);	981	SvPOK_only (enc.sv);
377	encode_sv (&enc, scalar);	982	encode_sv (&enc, scalar);
		983	encode_nl (&enc);
378		984
		985	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		986	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		987
379	if (!(flags & (F_ASCII | F_UTF8)))	988	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
380	SvUTF8_on (enc.sv);	989	SvUTF8_on (enc.sv);
381		990
382	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	991	if (enc.json.flags & F_SHRINK)
		992	shrink (enc.sv);
		993
383	return enc.sv;	994	return enc.sv;
384	}	995	}
385		996
386	/////////////////////////////////////////////////////////////////////////////	997	/////////////////////////////////////////////////////////////////////////////
		998	// decoder
387		999
388	#define WS \	1000	// structure used for decoding JSON
		1001	typedef struct
		1002	{
		1003	char *cur; // current parser pointer
		1004	char *end; // end of input string
		1005	const char *err; // parse error, if != 0
		1006	JSON json;
		1007	U32 depth; // recursion depth
		1008	U32 maxdepth; // recursion depth limit
		1009	} dec_t;
		1010
		1011	INLINE void
		1012	decode_comment (dec_t *dec)
		1013	{
		1014	// only '#'-style comments allowed a.t.m.
		1015
		1016	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		1017	++dec->cur;
		1018	}
		1019
		1020	INLINE void
		1021	decode_ws (dec_t *dec)
		1022	{
389	for (;;) \	1023	for (;;)
390	{ \	1024	{
391	char ch = *dec->cur; \	1025	char ch = *dec->cur;
		1026
392	if (ch > 0x20 \	1027	if (ch > 0x20)
		1028	{
		1029	if (expect_false (ch == '#'))
		1030	{
		1031	if (dec->json.flags & F_RELAXED)
		1032	decode_comment (dec);
		1033	else
		1034	break;
		1035	}
		1036	else
		1037	break;
		1038	}
393	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	1039	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
394	break; \	1040	break; // parse error, but let higher level handle it, gives better error messages
		1041
395	++dec->cur; \	1042	++dec->cur;
396	}	1043	}
		1044	}
397		1045
398	#define ERR(reason) SB dec->err = reason; goto fail; SE	1046	#define ERR(reason) SB dec->err = reason; goto fail; SE
		1047
399	#define EXPECT_CH(ch) SB \	1048	#define EXPECT_CH(ch) SB \
400	if (*dec->cur != ch) \	1049	if (*dec->cur != ch) \
401	ERR (# ch " expected"); \	1050	ERR (# ch " expected"); \
402	++dec->cur; \	1051	++dec->cur; \
403	SE	1052	SE
404		1053
		1054	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
		1055	#define DEC_DEC_DEPTH --dec->depth
		1056
405	static SV *decode_sv (dec_t *dec);	1057	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		1058
412	static signed char decode_hexdigit[256];	1059	static signed char decode_hexdigit[256];
413		1060
414	static UV	1061	static UV
415	decode_4hex (dec_t *dec)	1062	decode_4hex (dec_t *dec)
416	{	1063	{
417	signed char d1, d2, d3, d4;	1064	signed char d1, d2, d3, d4;
		1065	unsigned char *cur = (unsigned char *)dec->cur;
418		1066
419	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	1067	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");	1068	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]];	1069	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");	1070	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		1071
428	dec->cur += 4;	1072	dec->cur += 4;
429		1073
430	return ((UV)d1) << 12	1074	return ((UV)d1) << 12
431	| ((UV)d2) << 8	1075	| ((UV)d2) << 8
…		…
437	}	1081	}
438		1082
439	static SV *	1083	static SV *
440	decode_str (dec_t *dec)	1084	decode_str (dec_t *dec)
441	{	1085	{
442	SV *sv = NEWSV (0,2);	1086	SV *sv = 0;
443	STRLEN cur = 0;
444	int utf8 = 0;	1087	int utf8 = 0;
		1088	char *dec_cur = dec->cur;
445		1089
446	for (;;)	1090	do
447	{	1091	{
448	unsigned char ch = *(unsigned char *)dec->cur;	1092	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		1093	char *cur = buf;
449		1094
450	if (ch == '"')	1095	do
451	break;
452	else if (ch == '\\')
453	{	1096	{
454	switch (*++dec->cur)	1097	unsigned char ch = *(unsigned char *)dec_cur++;
		1098
		1099	if (expect_false (ch == '"'))
455	{	1100	{
456	case '\\':	1101	--dec_cur;
457	case '/':	1102	break;
458	case '"': APPEND_CH (*dec->cur++); break;	1103	}
459		1104	else if (expect_false (ch == '\\'))
460	case 'b': APPEND_CH ('\010'); ++dec->cur; break;	1105	{
461	case 't': APPEND_CH ('\011'); ++dec->cur; break;	1106	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	{	1107	{
468	UV lo, hi;	1108	case '\\':
469	++dec->cur;	1109	case '/':
		1110	case '"': *cur++ = *dec_cur++; break;
470		1111
471	hi = decode_4hex (dec);	1112	case 'b': ++dec_cur; *cur++ = '\010'; break;
472	if (hi == (UV)-1)	1113	case 't': ++dec_cur; *cur++ = '\011'; break;
473	goto fail;	1114	case 'n': ++dec_cur; *cur++ = '\012'; break;
		1115	case 'f': ++dec_cur; *cur++ = '\014'; break;
		1116	case 'r': ++dec_cur; *cur++ = '\015'; break;
474		1117
475	// possibly a surrogate pair	1118	case 'u':
476	if (hi >= 0xd800 && hi < 0xdc00)
477	{	1119	{
478	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	1120	UV lo, hi;
479	ERR ("illegal surrogate character");	1121	++dec_cur;
480		1122
481	dec->cur += 2;	1123	dec->cur = dec_cur;
482
483	lo = decode_4hex (dec);	1124	hi = decode_4hex (dec);
		1125	dec_cur = dec->cur;
484	if (lo == (UV)-1)	1126	if (hi == (UV)-1)
485	goto fail;	1127	goto fail;
486		1128
		1129	// possibly a surrogate pair
		1130	if (hi >= 0xd800)
		1131	if (hi < 0xdc00)
		1132	{
		1133	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		1134	ERR ("missing low surrogate character in surrogate pair");
		1135
		1136	dec_cur += 2;
		1137
		1138	dec->cur = dec_cur;
		1139	lo = decode_4hex (dec);
		1140	dec_cur = dec->cur;
		1141	if (lo == (UV)-1)
		1142	goto fail;
		1143
487	if (lo < 0xdc00 || lo >= 0xe000)	1144	if (lo < 0xdc00 || lo >= 0xe000)
488	ERR ("surrogate pair expected");	1145	ERR ("surrogate pair expected");
489		1146
490	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	1147	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		1148	}
		1149	else if (hi < 0xe000)
		1150	ERR ("missing high surrogate character in surrogate pair");
		1151
		1152	if (hi >= 0x80)
		1153	{
		1154	utf8 = 1;
		1155
		1156	cur = encode_utf8 (cur, hi);
		1157	}
		1158	else
		1159	*cur++ = hi;
491	}	1160	}
492	else if (lo >= 0xdc00 && lo < 0xe000)
493	ERR ("illegal surrogate character");
494
495	if (hi >= 0x80)
496	{	1161	break;
497	utf8 = 1;
498		1162
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	1163	default:
503	APPEND_CH (hi);	1164	--dec_cur;
		1165	ERR ("illegal backslash escape sequence in string");
504	}	1166	}
		1167	}
		1168	else if (expect_true (ch >= 0x20 && ch < 0x80))
		1169	*cur++ = ch;
		1170	else if (ch >= 0x80)
		1171	{
		1172	STRLEN clen;
		1173
		1174	--dec_cur;
		1175
		1176	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		1177	if (clen == (STRLEN)-1)
		1178	ERR ("malformed UTF-8 character in JSON string");
		1179
		1180	do
		1181	*cur++ = *dec_cur++;
		1182	while (--clen);
		1183
		1184	utf8 = 1;
		1185	}
		1186	else if (ch == '\t' && dec->json.flags & F_RELAXED)
		1187	*cur++ = ch;
		1188	else
		1189	{
		1190	--dec_cur;
		1191
		1192	if (!ch)
		1193	ERR ("unexpected end of string while parsing JSON string");
505	break;	1194	else
		1195	ERR ("invalid character encountered while parsing JSON string");
506	}	1196	}
507	}	1197	}
508	else if (ch >= 0x20 && ch <= 0x7f)	1198	while (cur < buf + SHORT_STRING_LEN);
509	APPEND_CH (*dec->cur++);	1199
510	else if (ch >= 0x80)	1200	{
		1201	STRLEN len = cur - buf;
		1202
		1203	if (sv)
511	{	1204	{
512	STRLEN clen;	1205	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		1206
517	SvGROW (sv, cur + clen + 1); // at most 4 bytes for 21 bits	1207	if (SvLEN (sv) - cur <= len)
518	memcpy (SvPVX (sv) + cur, dec->cur, clen);	1208	json_sv_grow (sv, cur, len);
519	dec->cur += clen;	1209
		1210	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		1211	SvCUR_set (sv, SvCUR (sv) + len);
520	}	1212	}
521	else	1213	else
522	ERR ("invalid character encountered");	1214	sv = newSVpvn (buf, len);
523	}	1215	}
		1216	}
		1217	while (*dec_cur != '"');
524		1218
525	++dec->cur;	1219	++dec_cur;
526		1220
		1221	if (sv)
		1222	{
527	SvPOK_only (sv);	1223	SvPOK_only (sv);
528
529	SvCUR_set (sv, cur);
530	*SvEND (sv) = 0;	1224	*SvEND (sv) = 0;
531		1225
532	if (utf8)	1226	if (utf8)
533	SvUTF8_on (sv);	1227	SvUTF8_on (sv);
		1228	}
		1229	else
		1230	sv = newSVpvn ("", 0);
534		1231
		1232	dec->cur = dec_cur;
535	return sv;	1233	return sv;
536		1234
537	fail:	1235	fail:
538	SvREFCNT_dec (sv);	1236	dec->cur = dec_cur;
539	return 0;	1237	return 0;
540	}	1238	}
541		1239
542	static SV *	1240	static SV *
543	decode_num (dec_t *dec)	1241	decode_num (dec_t *dec)
…		…
553	{	1251	{
554	++dec->cur;	1252	++dec->cur;
555	if (*dec->cur >= '0' && *dec->cur <= '9')	1253	if (*dec->cur >= '0' && *dec->cur <= '9')
556	ERR ("malformed number (leading zero must not be followed by another digit)");	1254	ERR ("malformed number (leading zero must not be followed by another digit)");
557	}	1255	}
558		1256	else if (*dec->cur < '0' || *dec->cur > '9')
559	// int	1257	ERR ("malformed number (no digits after initial minus)");
		1258	else
		1259	do
		1260	{
		1261	++dec->cur;
		1262	}
560	while (*dec->cur >= '0' && *dec->cur <= '9')	1263	while (*dec->cur >= '0' && *dec->cur <= '9');
561	++dec->cur;
562		1264
563	// [frac]	1265	// [frac]
564	if (*dec->cur == '.')	1266	if (*dec->cur == '.')
565	{	1267	{
566	is_nv = 1;	1268	++dec->cur;
		1269
		1270	if (*dec->cur < '0' || *dec->cur > '9')
		1271	ERR ("malformed number (no digits after decimal point)");
567		1272
568	do	1273	do
569	{	1274	{
570	++dec->cur;	1275	++dec->cur;
571	}	1276	}
572	while (*dec->cur >= '0' && *dec->cur <= '9');	1277	while (*dec->cur >= '0' && *dec->cur <= '9');
		1278
		1279	is_nv = 1;
573	}	1280	}
574		1281
575	// [exp]	1282	// [exp]
576	if (*dec->cur == 'e' || *dec->cur == 'E')	1283	if (*dec->cur == 'e' || *dec->cur == 'E')
577	{	1284	{
578	is_nv = 1;
579
580	++dec->cur;	1285	++dec->cur;
		1286
581	if (*dec->cur == '-' || *dec->cur == '+')	1287	if (*dec->cur == '-' || *dec->cur == '+')
582	++dec->cur;	1288	++dec->cur;
583		1289
		1290	if (*dec->cur < '0' || *dec->cur > '9')
		1291	ERR ("malformed number (no digits after exp sign)");
		1292
		1293	do
		1294	{
		1295	++dec->cur;
		1296	}
584	while (*dec->cur >= '0' && *dec->cur <= '9')	1297	while (*dec->cur >= '0' && *dec->cur <= '9');
585	++dec->cur;	1298
		1299	is_nv = 1;
586	}	1300	}
587		1301
588	if (!is_nv)	1302	if (!is_nv)
589	{	1303	{
590	UV uv;	1304	int len = dec->cur - start;
591	int numtype = grok_number (start, dec->cur - start, &uv);	1305
592	if (numtype & IS_NUMBER_IN_UV)	1306	// special case the rather common 1..5-digit-int case
593	if (numtype & IS_NUMBER_NEG)	1307	if (*start == '-')
		1308	switch (len)
594	{	1309	{
595	if (uv < (UV)IV_MIN)	1310	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
596	return newSViv (-(IV)uv);	1311	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
		1312	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1313	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1314	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
597	}	1315	}
		1316	else
		1317	switch (len)
		1318	{
		1319	case 1: return newSViv ( start [0] - '0' * 1);
		1320	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1321	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1322	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1323	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1324	}
		1325
		1326	{
		1327	UV uv;
		1328	int numtype = grok_number (start, len, &uv);
		1329	if (numtype & IS_NUMBER_IN_UV)
		1330	if (numtype & IS_NUMBER_NEG)
		1331	{
		1332	if (uv < (UV)IV_MIN)
		1333	return newSViv (-(IV)uv);
		1334	}
598	else	1335	else
599	return newSVuv (uv);	1336	return newSVuv (uv);
600	}	1337	}
601		1338
		1339	len -= *start == '-' ? 1 : 0;
		1340
		1341	// does not fit into IV or UV, try NV
		1342	if (len <= NV_DIG)
		1343	// fits into NV without loss of precision
		1344	return newSVnv (json_atof (start));
		1345
		1346	// everything else fails, convert it to a string
		1347	return newSVpvn (start, dec->cur - start);
		1348	}
		1349
		1350	// loss of precision here
602	return newSVnv (Atof (start));	1351	return newSVnv (json_atof (start));
603		1352
604	fail:	1353	fail:
605	return 0;	1354	return 0;
606	}	1355	}
607		1356
608	static SV *	1357	static SV *
609	decode_av (dec_t *dec)	1358	decode_av (dec_t *dec)
610	{	1359	{
611	AV *av = newAV ();	1360	AV *av = newAV ();
612		1361
		1362	DEC_INC_DEPTH;
		1363	decode_ws (dec);
		1364
		1365	if (*dec->cur == ']')
		1366	++dec->cur;
		1367	else
613	for (;;)	1368	for (;;)
614	{	1369	{
615	SV *value;	1370	SV *value;
616		1371
617	value = decode_sv (dec);	1372	value = decode_sv (dec);
618	if (!value)	1373	if (!value)
619	goto fail;	1374	goto fail;
620		1375
621	av_push (av, value);	1376	av_push (av, value);
622		1377
623	WS;	1378	decode_ws (dec);
624		1379
625	if (*dec->cur == ']')	1380	if (*dec->cur == ']')
626	{	1381	{
627	++dec->cur;	1382	++dec->cur;
628	break;	1383	break;
629	}	1384	}
630		1385
631	if (*dec->cur != ',')	1386	if (*dec->cur != ',')
632	ERR (", or ] expected while parsing array");	1387	ERR (", or ] expected while parsing array");
633		1388
634	++dec->cur;	1389	++dec->cur;
		1390
		1391	decode_ws (dec);
		1392
		1393	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1394	{
		1395	++dec->cur;
		1396	break;
		1397	}
635	}	1398	}
636		1399
		1400	DEC_DEC_DEPTH;
637	return newRV_noinc ((SV *)av);	1401	return newRV_noinc ((SV *)av);
638		1402
639	fail:	1403	fail:
640	SvREFCNT_dec (av);	1404	SvREFCNT_dec (av);
		1405	DEC_DEC_DEPTH;
641	return 0;	1406	return 0;
642	}	1407	}
643		1408
644	static SV *	1409	static SV *
645	decode_hv (dec_t *dec)	1410	decode_hv (dec_t *dec)
646	{	1411	{
		1412	SV *sv;
647	HV *hv = newHV ();	1413	HV *hv = newHV ();
648		1414
		1415	DEC_INC_DEPTH;
		1416	decode_ws (dec);
		1417
		1418	if (*dec->cur == '}')
		1419	++dec->cur;
		1420	else
649	for (;;)	1421	for (;;)
650	{	1422	{
651	SV *key, *value;
652
653	WS; EXPECT_CH ('"');	1423	EXPECT_CH ('"');
654		1424
655	key = decode_str (dec);	1425	// heuristic: assume that
656	if (!key)	1426	// a) decode_str + hv_store_ent are abysmally slow.
657	goto fail;	1427	// b) most hash keys are short, simple ascii text.
658		1428	// => try to "fast-match" such strings to avoid
659	WS; EXPECT_CH (':');	1429	// the overhead of decode_str + hv_store_ent.
660
661	value = decode_sv (dec);
662	if (!value)
663	{	1430	{
		1431	SV *value;
		1432	char *p = dec->cur;
		1433	char *e = p + 24; // only try up to 24 bytes
		1434
		1435	for (;;)
		1436	{
		1437	// the >= 0x80 is false on most architectures
		1438	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1439	{
		1440	// slow path, back up and use decode_str
		1441	SV *key = decode_str (dec);
		1442	if (!key)
		1443	goto fail;
		1444
		1445	decode_ws (dec); EXPECT_CH (':');
		1446
		1447	decode_ws (dec);
		1448	value = decode_sv (dec);
		1449	if (!value)
		1450	{
		1451	SvREFCNT_dec (key);
		1452	goto fail;
		1453	}
		1454
		1455	hv_store_ent (hv, key, value, 0);
664	SvREFCNT_dec (key);	1456	SvREFCNT_dec (key);
		1457
		1458	break;
		1459	}
		1460	else if (*p == '"')
		1461	{
		1462	// fast path, got a simple key
		1463	char *key = dec->cur;
		1464	int len = p - key;
		1465	dec->cur = p + 1;
		1466
		1467	decode_ws (dec); EXPECT_CH (':');
		1468
		1469	decode_ws (dec);
		1470	value = decode_sv (dec);
		1471	if (!value)
665	goto fail;	1472	goto fail;
		1473
		1474	hv_store (hv, key, len, value, 0);
		1475
		1476	break;
		1477	}
		1478
		1479	++p;
		1480	}
666	}	1481	}
667		1482
668	//TODO: optimise	1483	decode_ws (dec);
669	hv_store_ent (hv, key, value, 0);
670		1484
671	WS;
672
673	if (*dec->cur == '}')	1485	if (*dec->cur == '}')
		1486	{
		1487	++dec->cur;
		1488	break;
		1489	}
		1490
		1491	if (*dec->cur != ',')
		1492	ERR (", or } expected while parsing object/hash");
		1493
		1494	++dec->cur;
		1495
		1496	decode_ws (dec);
		1497
		1498	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1499	{
		1500	++dec->cur;
		1501	break;
		1502	}
		1503	}
		1504
		1505	DEC_DEC_DEPTH;
		1506	sv = newRV_noinc ((SV *)hv);
		1507
		1508	// check filter callbacks
		1509	if (expect_false (dec->json.flags & F_HOOK))
		1510	{
		1511	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
674	{	1512	{
675	++dec->cur;	1513	HE *cb, *he;
676	break;	1514
		1515	hv_iterinit (hv);
		1516	he = hv_iternext (hv);
		1517	hv_iterinit (hv);
		1518
		1519	// the next line creates a mortal sv each time it's called.
		1520	// might want to optimise this for common cases.
		1521	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1522
		1523	if (cb)
		1524	{
		1525	dSP;
		1526	int count;
		1527
		1528	ENTER; SAVETMPS;
		1529	PUSHMARK (SP);
		1530	XPUSHs (HeVAL (he));
		1531	sv_2mortal (sv);
		1532
		1533	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1534
		1535	if (count == 1)
		1536	{
		1537	sv = newSVsv (POPs);
		1538	FREETMPS; LEAVE;
		1539	return sv;
		1540	}
		1541	else if (count)
		1542	croak ("filter_json_single_key_object callbacks must not return more than one scalar");
		1543
		1544	SvREFCNT_inc (sv);
		1545	FREETMPS; LEAVE;
		1546	}
677	}	1547	}
678		1548
679	if (*dec->cur != ',')	1549	if (dec->json.cb_object)
680	ERR (", or } expected while parsing object/hash");	1550	{
		1551	dSP;
		1552	int count;
681		1553
682	++dec->cur;	1554	ENTER; SAVETMPS;
683	}	1555	PUSHMARK (SP);
		1556	XPUSHs (sv_2mortal (sv));
684		1557
685	return newRV_noinc ((SV *)hv);	1558	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1559
		1560	if (count == 1)
		1561	sv = newSVsv (POPs);
		1562	else if (count == 0)
		1563	SvREFCNT_inc (sv);
		1564	else
		1565	croak ("filter_json_object callbacks must not return more than one scalar");
		1566
		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 != dec.end)
		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	switch (self->incr_mode)
		1848	{
		1849	// reached end of a scalar, see if we are inside a nested structure or not
		1850	end_of_scalar:
		1851	self->incr_mode = INCR_M_JSON;
		1852
		1853	if (self->incr_nest) // end of a scalar inside array, object or tag
		1854	goto incr_m_json;
		1855	else // end of scalar outside structure, json text ends here
		1856	goto interrupt;
		1857
		1858	// only used for initial whitespace skipping
		1859	case INCR_M_WS:
		1860	for (;;)
		1861	{
		1862	if (*p > 0x20)
		1863	{
		1864	if (*p == '#')
		1865	{
		1866	self->incr_mode = INCR_M_C0;
		1867	goto incr_m_c;
		1868	}
		1869	else
		1870	{
		1871	self->incr_mode = INCR_M_JSON;
		1872	goto incr_m_json;
		1873	}
		1874	}
		1875	else if (!*p)
		1876	goto interrupt;
		1877
		1878	++p;
		1879	}
		1880
		1881	// skip a single char inside a string (for \\-processing)
		1882	case INCR_M_BS:
		1883	if (!*p)
		1884	goto interrupt;
		1885
		1886	++p;
		1887	self->incr_mode = INCR_M_STR;
		1888	goto incr_m_str;
		1889
		1890	// inside #-style comments
		1891	case INCR_M_C0:
		1892	case INCR_M_C1:
		1893	incr_m_c:
		1894	for (;;)
		1895	{
		1896	if (*p == '\n')
		1897	{
		1898	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1899	break;
		1900	}
		1901	else if (!*p)
		1902	goto interrupt;
		1903
		1904	++p;
		1905	}
		1906
		1907	break;
		1908
		1909	// inside true/false/null
		1910	case INCR_M_TFN:
		1911	incr_m_tfn:
		1912	for (;;)
		1913	switch (*p++)
		1914	{
		1915	case 'r': case 'u': case 'e': // tRUE, falsE, nUll
		1916	case 'a': case 'l': case 's': // fALSe, nuLL
		1917	// allowed
		1918	break;
		1919
		1920	default:
		1921	--p;
		1922	goto end_of_scalar;
		1923	}
		1924
		1925	// inside a number
		1926	case INCR_M_NUM:
		1927	incr_m_num:
		1928	for (;;)
		1929	switch (*p++)
		1930	{
		1931	case 'e': case 'E': case '.': case '+':
		1932	case '-':
		1933	case '0': case '1': case '2': case '3': case '4':
		1934	case '5': case '6': case '7': case '8': case '9':
		1935	// allowed
		1936	break;
		1937
		1938	default:
		1939	--p;
		1940	goto end_of_scalar;
		1941	}
		1942
		1943	// inside a string
		1944	case INCR_M_STR:
		1945	incr_m_str:
		1946	for (;;)
		1947	{
		1948	if (*p == '"')
		1949	{
		1950	++p;
		1951	goto end_of_scalar;
		1952	}
		1953	else if (*p == '\\')
		1954	{
		1955	++p; // "virtually" consumes character after \
		1956
		1957	if (!*p) // if at end of string we have to switch modes
		1958	{
		1959	self->incr_mode = INCR_M_BS;
		1960	goto interrupt;
		1961	}
		1962	}
		1963	else if (!*p)
		1964	goto interrupt;
		1965
		1966	++p;
		1967	}
		1968
		1969	// after initial ws, outside string
		1970	case INCR_M_JSON:
		1971	incr_m_json:
		1972	for (;;)
		1973	{
		1974	switch (*p++)
		1975	{
		1976	case 0:
		1977	--p;
		1978	goto interrupt;
		1979
		1980	case 0x09:
		1981	case 0x0a:
		1982	case 0x0d:
		1983	case 0x20:
		1984	if (!self->incr_nest)
		1985	{
		1986	--p; // do not eat the whitespace, let the next round do it
		1987	goto interrupt;
		1988	}
		1989	break;
		1990
		1991	// the following three blocks handle scalars. this makes the parser
		1992	// more strict than required inside arrays or objects, and could
		1993	// be moved to a special case on the toplevel (except strings)
		1994	case 't':
		1995	case 'f':
		1996	case 'n':
		1997	self->incr_mode = INCR_M_TFN;
		1998	goto incr_m_tfn;
		1999
		2000	case '-':
		2001	case '0': case '1': case '2': case '3': case '4':
		2002	case '5': case '6': case '7': case '8': case '9':
		2003	self->incr_mode = INCR_M_NUM;
		2004	goto incr_m_num;
		2005
		2006	case '"':
		2007	self->incr_mode = INCR_M_STR;
		2008	goto incr_m_str;
		2009
		2010	case '[':
		2011	case '{':
		2012	case '(':
		2013	if (++self->incr_nest > self->max_depth)
		2014	croak (ERR_NESTING_EXCEEDED);
		2015	break;
		2016
		2017	case ']':
		2018	case '}':
		2019	if (--self->incr_nest <= 0)
		2020	goto interrupt;
		2021	break;
		2022
		2023	case ')':
		2024	--self->incr_nest;
		2025	break;
		2026
		2027	case '#':
		2028	self->incr_mode = INCR_M_C1;
		2029	goto incr_m_c;
		2030	}
		2031	}
		2032	}
		2033
		2034	modechange:
		2035	;
		2036	}
		2037
		2038	interrupt:
		2039	self->incr_pos = p - SvPVX (self->incr_text);
		2040	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		2041	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
		2042	}
		2043
		2044	/////////////////////////////////////////////////////////////////////////////
		2045	// XS interface functions
		2046
789	MODULE = JSON::XS PACKAGE = JSON::XS	2047	MODULE = JSON::XS PACKAGE = JSON::XS
790		2048
791	BOOT:	2049	BOOT:
792	{	2050	{
793	int i;	2051	int i;
794		2052
795	memset (decode_hexdigit, 0xff, 256);
796	for (i = 10; i--; )	2053	for (i = 0; i < 256; ++i)
797	decode_hexdigit ['0' + i] = i;	2054	decode_hexdigit [i] =
		2055	i >= '0' && i <= '9' ? i - '0'
		2056	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		2057	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		2058	: -1;
798		2059
799	for (i = 6; --i; )	2060	json_stash = gv_stashpv ("JSON::XS" , 1);
800	{	2061	bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
801	decode_hexdigit ['a' + i] = 10 + i;	2062	bool_false = get_bool ("Types::Serialiser::false");
802	decode_hexdigit ['A' + i] = 10 + i;	2063	bool_true = get_bool ("Types::Serialiser::true");
		2064
		2065	sv_json = newSVpv ("JSON", 0);
		2066	SvREADONLY_on (sv_json);
		2067
		2068	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
		2069	}
		2070
		2071	PROTOTYPES: DISABLE
		2072
		2073	void CLONE (...)
		2074	CODE:
		2075	// as long as these writes are atomic, the race should not matter
		2076	// as existing threads either already use 0, or use the old value,
		2077	// which is sitll correct for the initial thread.
		2078	json_stash = 0;
		2079	bool_stash = 0;
		2080	bool_false = 0;
		2081	bool_true = 0;
		2082
		2083	void new (char *klass)
		2084	PPCODE:
		2085	{
		2086	SV *pv = NEWSV (0, sizeof (JSON));
		2087	SvPOK_only (pv);
		2088	json_init ((JSON *)SvPVX (pv));
		2089	XPUSHs (sv_2mortal (sv_bless (
		2090	newRV_noinc (pv),
		2091	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		2092	)));
		2093	}
		2094
		2095	void boolean_values (JSON *self, SV *v_false = 0, SV *v_true = 0)
		2096	PPCODE:
		2097	self->v_false = newSVsv (v_false);
		2098	self->v_true = newSVsv (v_true);
		2099	XPUSHs (ST (0));
		2100
		2101	void get_boolean_values (JSON *self)
		2102	PPCODE:
		2103	if (self->v_false && self->v_true)
		2104	{
		2105	EXTEND (SP, 2);
		2106	PUSHs (self->v_false);
		2107	PUSHs (self->v_true);
803	}	2108	}
804		2109
805	json_stash = gv_stashpv ("JSON::XS", 1);	2110	void ascii (JSON *self, int enable = 1)
806	}	2111	ALIAS:
		2112	ascii = F_ASCII
		2113	latin1 = F_LATIN1
		2114	utf8 = F_UTF8
		2115	indent = F_INDENT
		2116	canonical = F_CANONICAL
		2117	space_before = F_SPACE_BEFORE
		2118	space_after = F_SPACE_AFTER
		2119	pretty = F_PRETTY
		2120	allow_nonref = F_ALLOW_NONREF
		2121	shrink = F_SHRINK
		2122	allow_blessed = F_ALLOW_BLESSED
		2123	convert_blessed = F_CONV_BLESSED
		2124	relaxed = F_RELAXED
		2125	allow_unknown = F_ALLOW_UNKNOWN
		2126	allow_tags = F_ALLOW_TAGS
		2127	PPCODE:
		2128	{
		2129	if (enable)
		2130	self->flags |= ix;
		2131	else
		2132	self->flags &= ~ix;
807		2133
808	SV *new (char *dummy)	2134	XPUSHs (ST (0));
		2135	}
		2136
		2137	void get_ascii (JSON *self)
		2138	ALIAS:
		2139	get_ascii = F_ASCII
		2140	get_latin1 = F_LATIN1
		2141	get_utf8 = F_UTF8
		2142	get_indent = F_INDENT
		2143	get_canonical = F_CANONICAL
		2144	get_space_before = F_SPACE_BEFORE
		2145	get_space_after = F_SPACE_AFTER
		2146	get_allow_nonref = F_ALLOW_NONREF
		2147	get_shrink = F_SHRINK
		2148	get_allow_blessed = F_ALLOW_BLESSED
		2149	get_convert_blessed = F_CONV_BLESSED
		2150	get_relaxed = F_RELAXED
		2151	get_allow_unknown = F_ALLOW_UNKNOWN
		2152	get_allow_tags = F_ALLOW_TAGS
		2153	PPCODE:
		2154	XPUSHs (boolSV (self->flags & ix));
		2155
		2156	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		2157	PPCODE:
		2158	self->max_depth = max_depth;
		2159	XPUSHs (ST (0));
		2160
		2161	U32 get_max_depth (JSON *self)
809	CODE:	2162	CODE:
810	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	2163	RETVAL = self->max_depth;
811	OUTPUT:	2164	OUTPUT:
812	RETVAL	2165	RETVAL
813		2166
814	SV *ascii (SV *self, int enable)	2167	void max_size (JSON *self, U32 max_size = 0)
815	ALIAS:	2168	PPCODE:
816	ascii = F_ASCII	2169	self->max_size = max_size;
817	utf8 = F_UTF8	2170	XPUSHs (ST (0));
818	indent = F_INDENT	2171
819	canonical = F_CANONICAL	2172	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:	2173	CODE:
826	{	2174	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:	2175	OUTPUT:
836	RETVAL	2176	RETVAL
837		2177
838	void encode (SV *self, SV *scalar)	2178	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
839	PPCODE:	2179	PPCODE:
840	XPUSHs (encode_json (scalar, *SvJSON (self)));	2180	{
		2181	SvREFCNT_dec (self->cb_object);
		2182	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
841		2183
842	void decode (SV *self, SV *jsonstr)	2184	XPUSHs (ST (0));
		2185	}
		2186
		2187	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
843	PPCODE:	2188	PPCODE:
844	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	2189	{
		2190	if (!self->cb_sk_object)
		2191	self->cb_sk_object = newHV ();
845		2192
846	void to_json (SV *scalar)	2193	if (SvOK (cb))
		2194	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		2195	else
		2196	{
		2197	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		2198
		2199	if (!HvKEYS (self->cb_sk_object))
		2200	{
		2201	SvREFCNT_dec (self->cb_sk_object);
		2202	self->cb_sk_object = 0;
		2203	}
		2204	}
		2205
		2206	XPUSHs (ST (0));
		2207	}
		2208
		2209	void encode (JSON *self, SV *scalar)
847	PPCODE:	2210	PPCODE:
848	XPUSHs (encode_json (scalar, F_UTF8));	2211	PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
		2212	XPUSHs (scalar);
849		2213
850	void from_json (SV *jsonstr)	2214	void decode (JSON *self, SV *jsonstr)
851	PPCODE:	2215	PPCODE:
852	XPUSHs (decode_json (jsonstr, F_UTF8));	2216	PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
		2217	XPUSHs (jsonstr);
853		2218
		2219	void decode_prefix (JSON *self, SV *jsonstr)
		2220	PPCODE:
		2221	{
		2222	SV *sv;
		2223	STRLEN offset;
		2224	PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
		2225	EXTEND (SP, 2);
		2226	PUSHs (sv);
		2227	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset))));
		2228	}
		2229
		2230	void incr_parse (JSON *self, SV *jsonstr = 0)
		2231	PPCODE:
		2232	{
		2233	if (!self->incr_text)
		2234	self->incr_text = newSVpvn ("", 0);
		2235
		2236	/* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
		2237	if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
		2238	if (DECODE_WANTS_OCTETS (self))
		2239	{
		2240	if (self->incr_pos)
		2241	self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
		2242	(U8 *)SvPVX (self->incr_text) + self->incr_pos);
		2243
		2244	sv_utf8_downgrade (self->incr_text, 0);
		2245	}
		2246	else
		2247	{
		2248	sv_utf8_upgrade (self->incr_text);
		2249
		2250	if (self->incr_pos)
		2251	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		2252	- (U8 *)SvPVX (self->incr_text);
		2253	}
		2254
		2255	// append data, if any
		2256	if (jsonstr)
		2257	{
		2258	/* make sure both strings have same encoding */
		2259	if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
		2260	if (SvUTF8 (jsonstr))
		2261	sv_utf8_downgrade (jsonstr, 0);
		2262	else
		2263	sv_utf8_upgrade (jsonstr);
		2264
		2265	/* and then just blindly append */
		2266	{
		2267	STRLEN len;
		2268	const char *str = SvPV (jsonstr, len);
		2269	STRLEN cur = SvCUR (self->incr_text);
		2270
		2271	if (SvLEN (self->incr_text) - cur <= len)
		2272	json_sv_grow (self->incr_text, cur, len);
		2273
		2274	Move (str, SvEND (self->incr_text), len, char);
		2275	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		2276	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		2277	}
		2278	}
		2279
		2280	if (GIMME_V != G_VOID)
		2281	do
		2282	{
		2283	SV *sv;
		2284	STRLEN offset;
		2285
		2286	if (!INCR_DONE (self))
		2287	{
		2288	incr_parse (self);
		2289
		2290	if (self->incr_pos > self->max_size && self->max_size)
		2291	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		2292	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		2293
		2294	if (!INCR_DONE (self))
		2295	{
		2296	// as an optimisation, do not accumulate white space in the incr buffer
		2297	if (self->incr_mode == INCR_M_WS && self->incr_pos)
		2298	{
		2299	self->incr_pos = 0;
		2300	SvCUR_set (self->incr_text, 0);
		2301	}
		2302
		2303	break;
		2304	}
		2305	}
		2306
		2307	PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
		2308	XPUSHs (sv);
		2309
		2310	self->incr_pos -= offset;
		2311	self->incr_nest = 0;
		2312	self->incr_mode = 0;
		2313
		2314	sv_chop (self->incr_text, SvPVX (self->incr_text) + offset);
		2315	}
		2316	while (GIMME_V == G_ARRAY);
		2317	}
		2318
		2319	SV *incr_text (JSON *self)
		2320	ATTRS: lvalue
		2321	CODE:
		2322	{
		2323	if (self->incr_pos)
		2324	croak ("incr_text can not be called when the incremental parser already started parsing");
		2325
		2326	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		2327	}
		2328	OUTPUT:
		2329	RETVAL
		2330
		2331	void incr_skip (JSON *self)
		2332	CODE:
		2333	{
		2334	if (self->incr_pos)
		2335	{
		2336	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2337	self->incr_pos = 0;
		2338	self->incr_nest = 0;
		2339	self->incr_mode = 0;
		2340	}
		2341	}
		2342
		2343	void incr_reset (JSON *self)
		2344	CODE:
		2345	{
		2346	SvREFCNT_dec (self->incr_text);
		2347	self->incr_text = 0;
		2348	self->incr_pos = 0;
		2349	self->incr_nest = 0;
		2350	self->incr_mode = 0;
		2351	}
		2352
		2353	void DESTROY (JSON *self)
		2354	CODE:
		2355	SvREFCNT_dec (self->v_false);
		2356	SvREFCNT_dec (self->v_true);
		2357	SvREFCNT_dec (self->cb_sk_object);
		2358	SvREFCNT_dec (self->cb_object);
		2359	SvREFCNT_dec (self->incr_text);
		2360
		2361	PROTOTYPES: ENABLE
		2362
		2363	void encode_json (SV *scalar)
		2364	PPCODE:
		2365	{
		2366	JSON json;
		2367	json_init (&json);
		2368	json.flags |= F_UTF8;
		2369	PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
		2370	XPUSHs (scalar);
		2371	}
		2372
		2373	void decode_json (SV *jsonstr)
		2374	PPCODE:
		2375	{
		2376	JSON json;
		2377	json_init (&json);
		2378	json.flags |= F_UTF8;
		2379	PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
		2380	XPUSHs (jsonstr);
		2381	}
		2382

Diff Legend

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