ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.55 by root, Mon Jul 23 22:57:40 2007 UTC vs.
Revision 1.143 by root, Mon Jun 21 01:04:30 2021 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"	8	#include <stdio.h>
		9	#include <limits.h>
		10	#include <float.h>
		11	#include <inttypes.h>
9		12
10	#if defined(__BORLANDC__) || defined(_MSC_VER)	13	#if defined(__BORLANDC__) || defined(_MSC_VER)
11	# define snprintf _snprintf // C compilers have this in stdio.h	14	# define snprintf _snprintf // C compilers have this in stdio.h
12	#endif	15	#endif
13		16
14	// some old perls do not have this, try to make it work, no	17	// some old perls do not have this, try to make it work, no
15	// guarentees, though. if it breaks, you get to keep the pieces.	18	// guarantees, though. if it breaks, you get to keep the pieces.
16	#ifndef UTF8_MAXBYTES	19	#ifndef UTF8_MAXBYTES
17	# define UTF8_MAXBYTES 13	20	# define UTF8_MAXBYTES 13
18	#endif	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)
19		36
20	#define F_ASCII 0x00000001UL	37	#define F_ASCII 0x00000001UL
21	#define F_LATIN1 0x00000002UL	38	#define F_LATIN1 0x00000002UL
22	#define F_UTF8 0x00000004UL	39	#define F_UTF8 0x00000004UL
23	#define F_INDENT 0x00000008UL	40	#define F_INDENT 0x00000008UL
…		…
26	#define F_SPACE_AFTER 0x00000040UL	43	#define F_SPACE_AFTER 0x00000040UL
27	#define F_ALLOW_NONREF 0x00000100UL	44	#define F_ALLOW_NONREF 0x00000100UL
28	#define F_SHRINK 0x00000200UL	45	#define F_SHRINK 0x00000200UL
29	#define F_ALLOW_BLESSED 0x00000400UL	46	#define F_ALLOW_BLESSED 0x00000400UL
30	#define F_CONV_BLESSED 0x00000800UL	47	#define F_CONV_BLESSED 0x00000800UL
31	#define F_MAXDEPTH 0xf8000000UL
32	#define S_MAXDEPTH 27
33	#define F_MAXSIZE 0x01f00000UL	48	#define F_RELAXED 0x00001000UL
34	#define S_MAXSIZE 20	49	#define F_ALLOW_UNKNOWN 0x00002000UL
		50	#define F_ALLOW_TAGS 0x00004000UL
35	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing	51	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
36		52
37	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
38	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
39
40	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	53	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
41	#define F_DEFAULT (9UL << S_MAXDEPTH)
42		54
43	#define INIT_SIZE 32 // initial scalar size to be allocated	55	#define INIT_SIZE 64 // initial scalar size to be allocated
44	#define INDENT_STEP 3 // spaces per indentation level	56	#define INDENT_STEP 3 // spaces per indentation level
45		57
46	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size	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)
47		61
48	#define SB do {	62	#define SB do {
49	#define SE } while (0)	63	#define SE } while (0)
50		64
51	#if __GNUC__ >= 3	65	#if __GNUC__ >= 3
52	# define expect(expr,value) __builtin_expect ((expr),(value))	66	# define expect(expr,value) __builtin_expect ((expr), (value))
53	# define inline inline	67	# define INLINE static inline
54	#else	68	#else
55	# define expect(expr,value) (expr)	69	# define expect(expr,value) (expr)
56	# define inline static	70	# define INLINE static
57	#endif	71	#endif
58		72
59	#define expect_false(expr) expect ((expr) != 0, 0)	73	#define expect_false(expr) expect ((expr) != 0, 0)
60	#define expect_true(expr) expect ((expr) != 0, 1)	74	#define expect_true(expr) expect ((expr) != 0, 1)
61		75
62	static HV *json_stash, *json_boolean_stash; // JSON::XS::	76	#define IN_RANGE_INC(type,val,beg,end) \
63	static SV *json_true, *json_false;	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)
64		111
65	typedef struct {	112	typedef struct {
66	U32 flags;	113	U32 flags;
		114	U32 max_depth;
		115	STRLEN max_size;
		116
67	SV *cb_object;	117	SV *cb_object;
68	HV *cb_sk_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;
69	} JSON;	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	}
70		136
71	/////////////////////////////////////////////////////////////////////////////	137	/////////////////////////////////////////////////////////////////////////////
72	// utility functions	138	// utility functions
73		139
74	inline void	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
75	shrink (SV *sv)	152	shrink (SV *sv)
76	{	153	{
77	sv_utf8_downgrade (sv, 1);	154	sv_utf8_downgrade (sv, 1);
		155
78	if (SvLEN (sv) > SvCUR (sv) + 1)	156	if (SvLEN (sv) > SvCUR (sv) + 1)
79	{	157	{
80	#ifdef SvPV_shrink_to_cur	158	#ifdef SvPV_shrink_to_cur
81	SvPV_shrink_to_cur (sv);	159	SvPV_shrink_to_cur (sv);
82	#elif defined (SvPV_renew)	160	#elif defined (SvPV_renew)
83	SvPV_renew (sv, SvCUR (sv) + 1);	161	SvPV_renew (sv, SvCUR (sv) + 1);
84	#endif	162	#endif
85	}	163	}
86	}	164	}
87		165
		166	/* adds two STRLENs together, slow, and with paranoia */
		167	static 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
88	// decode an utf-8 character and return it, or (UV)-1 in	191	// decode a utf-8 character and return it, or (UV)-1 in
89	// case of an error.	192	// case of an error.
90	// we special-case "safe" characters from U+80 .. U+7FF,	193	// we special-case "safe" characters from U+80 .. U+7FF,
91	// but use the very good perl function to parse anything else.	194	// but use the very good perl function to parse anything else.
92	// note that we never call this function for a ascii codepoints	195	// note that we never call this function for a ascii codepoints
93	inline UV	196	INLINE UV
94	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)	197	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
95	{	198	{
96	if (expect_false (s[0] > 0xdf || s[0] < 0xc2))	199	if (expect_true (len >= 2
97	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	200	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
98	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	201	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
99	{	202	{
100	*clen = 2;	203	*clen = 2;
101	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);	204	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
102	}	205	}
103	else	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	}
104	{	309	}
105	*clen = (STRLEN)-1;	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)
106	return (UV)-1;	372	return 1;
		373
		374	if (!SvOBJECT (scalar) && ref_bool_type (scalar) >= 0)
		375	return 1;
107	}	376	}
		377
		378	return 0;
108	}	379	}
109		380
110	/////////////////////////////////////////////////////////////////////////////	381	/////////////////////////////////////////////////////////////////////////////
111	// encoder	382	// encoder
112		383
…		…
116	char *cur; // SvPVX (sv) + current output position	387	char *cur; // SvPVX (sv) + current output position
117	char *end; // SvEND (sv)	388	char *end; // SvEND (sv)
118	SV *sv; // result scalar	389	SV *sv; // result scalar
119	JSON json;	390	JSON json;
120	U32 indent; // indentation level	391	U32 indent; // indentation level
121	U32 maxdepth; // max. indentation/recursion level	392	UV limit; // escape character values >= this value when encoding
122	} enc_t;	393	} enc_t;
123		394
124	inline void	395	INLINE void
125	need (enc_t *enc, STRLEN len)	396	need (enc_t *enc, STRLEN len)
126	{	397	{
127	if (expect_false (enc->cur + len >= enc->end))	398	if (expect_false ((uintptr_t)(enc->end - enc->cur) < len))
128	{	399	{
129	STRLEN cur = enc->cur - SvPVX (enc->sv);	400	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
130	SvGROW (enc->sv, cur + len + 1);	401	char *buf = json_sv_grow (enc->sv, cur, len);
131	enc->cur = SvPVX (enc->sv) + cur;	402	enc->cur = buf + cur;
132	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	403	enc->end = buf + SvLEN (enc->sv) - 1;
133	}	404	}
134	}	405	}
135		406
136	inline void	407	INLINE void
137	encode_ch (enc_t *enc, char ch)	408	encode_ch (enc_t *enc, char ch)
138	{	409	{
139	need (enc, 1);	410	need (enc, 1);
140	*enc->cur++ = ch;	411	*enc->cur++ = ch;
141	}	412	}
…		…
153		424
154	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case	425	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
155	{	426	{
156	if (expect_false (ch == '"')) // but with slow exceptions	427	if (expect_false (ch == '"')) // but with slow exceptions
157	{	428	{
158	need (enc, len += 1);	429	need (enc, len + 1);
159	*enc->cur++ = '\\';	430	*enc->cur++ = '\\';
160	*enc->cur++ = '"';	431	*enc->cur++ = '"';
161	}	432	}
162	else if (expect_false (ch == '\\'))	433	else if (expect_false (ch == '\\'))
163	{	434	{
164	need (enc, len += 1);	435	need (enc, len + 1);
165	*enc->cur++ = '\\';	436	*enc->cur++ = '\\';
166	*enc->cur++ = '\\';	437	*enc->cur++ = '\\';
167	}	438	}
168	else	439	else
169	*enc->cur++ = ch;	440	*enc->cur++ = ch;
…		…
172	}	443	}
173	else	444	else
174	{	445	{
175	switch (ch)	446	switch (ch)
176	{	447	{
177	case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;	448	case '\010': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
178	case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;	449	case '\011': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
179	case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;	450	case '\012': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
180	case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;	451	case '\014': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
181	case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;	452	case '\015': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
182		453
183	default:	454	default:
184	{	455	{
185	STRLEN clen;	456	STRLEN clen;
186	UV uch;	457	UV uch;
…		…
195	{	466	{
196	uch = ch;	467	uch = ch;
197	clen = 1;	468	clen = 1;
198	}	469	}
199		470
200	if (uch > 0x10FFFFUL)	471	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
201	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
202
203	if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
204	{	472	{
205	if (uch > 0xFFFFUL)	473	if (uch >= 0x10000UL)
206	{	474	{
		475	if (uch >= 0x110000UL)
		476	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		477
207	need (enc, len += 11);	478	need (enc, len + 11);
208	sprintf (enc->cur, "\\u%04x\\u%04x",	479	sprintf (enc->cur, "\\u%04x\\u%04x",
209	(int)((uch - 0x10000) / 0x400 + 0xD800),	480	(int)((uch - 0x10000) / 0x400 + 0xD800),
210	(int)((uch - 0x10000) % 0x400 + 0xDC00));	481	(int)((uch - 0x10000) % 0x400 + 0xDC00));
211	enc->cur += 12;	482	enc->cur += 12;
212	}	483	}
213	else	484	else
214	{	485	{
215	static char hexdigit [16] = "0123456789abcdef";
216	need (enc, len += 5);	486	need (enc, len + 5);
217	*enc->cur++ = '\\';	487	*enc->cur++ = '\\';
218	*enc->cur++ = 'u';	488	*enc->cur++ = 'u';
219	*enc->cur++ = hexdigit [ uch >> 12 ];	489	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
220	*enc->cur++ = hexdigit [(uch >> 8) & 15];	490	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
221	*enc->cur++ = hexdigit [(uch >> 4) & 15];	491	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
222	*enc->cur++ = hexdigit [(uch >> 0) & 15];	492	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
223	}	493	}
224		494
225	str += clen;	495	str += clen;
226	}	496	}
227	else if (enc->json.flags & F_LATIN1)	497	else if (enc->json.flags & F_LATIN1)
…		…
229	*enc->cur++ = uch;	499	*enc->cur++ = uch;
230	str += clen;	500	str += clen;
231	}	501	}
232	else if (is_utf8)	502	else if (is_utf8)
233	{	503	{
234	need (enc, len += clen);	504	need (enc, len + clen);
235	do	505	do
236	{	506	{
237	*enc->cur++ = *str++;	507	*enc->cur++ = *str++;
238	}	508	}
239	while (--clen);	509	while (--clen);
240	}	510	}
241	else	511	else
242	{	512	{
243	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	513	need (enc, len + UTF8_MAXBYTES - 1); // never more than 11 bytes needed
244	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	514	enc->cur = encode_utf8 (enc->cur, uch);
245	++str;	515	++str;
246	}	516	}
247	}	517	}
248	}	518	}
249	}	519	}
250		520
251	--len;	521	--len;
252	}	522	}
253	}	523	}
254		524
255	inline void	525	INLINE void
256	encode_indent (enc_t *enc)	526	encode_indent (enc_t *enc)
257	{	527	{
258	if (enc->json.flags & F_INDENT)	528	if (enc->json.flags & F_INDENT)
259	{	529	{
260	int spaces = enc->indent * INDENT_STEP;	530	int spaces = enc->indent * INDENT_STEP;
…		…
263	memset (enc->cur, ' ', spaces);	533	memset (enc->cur, ' ', spaces);
264	enc->cur += spaces;	534	enc->cur += spaces;
265	}	535	}
266	}	536	}
267		537
268	inline void	538	INLINE void
269	encode_space (enc_t *enc)	539	encode_space (enc_t *enc)
270	{	540	{
271	need (enc, 1);	541	need (enc, 1);
272	encode_ch (enc, ' ');	542	encode_ch (enc, ' ');
273	}	543	}
274		544
275	inline void	545	INLINE void
276	encode_nl (enc_t *enc)	546	encode_nl (enc_t *enc)
277	{	547	{
278	if (enc->json.flags & F_INDENT)	548	if (enc->json.flags & F_INDENT)
279	{	549	{
280	need (enc, 1);	550	need (enc, 1);
281	encode_ch (enc, '\n');	551	encode_ch (enc, '\n');
282	}	552	}
283	}	553	}
284		554
285	inline void	555	INLINE void
286	encode_comma (enc_t *enc)	556	encode_comma (enc_t *enc)
287	{	557	{
288	encode_ch (enc, ',');	558	encode_ch (enc, ',');
289		559
290	if (enc->json.flags & F_INDENT)	560	if (enc->json.flags & F_INDENT)
…		…
298	static void	568	static void
299	encode_av (enc_t *enc, AV *av)	569	encode_av (enc_t *enc, AV *av)
300	{	570	{
301	int i, len = av_len (av);	571	int i, len = av_len (av);
302		572
303	if (enc->indent >= enc->maxdepth)	573	if (enc->indent >= enc->json.max_depth)
304	croak ("data structure too deep (hit recursion limit)");	574	croak (ERR_NESTING_EXCEEDED);
305		575
306	encode_ch (enc, '['); encode_nl (enc);	576	encode_ch (enc, '[');
307	++enc->indent;
308		577
		578	if (len >= 0)
		579	{
		580	encode_nl (enc); ++enc->indent;
		581
309	for (i = 0; i <= len; ++i)	582	for (i = 0; i <= len; ++i)
310	{	583	{
311	SV **svp = av_fetch (av, i, 0);	584	SV **svp = av_fetch (av, i, 0);
312		585
313	encode_indent (enc);	586	encode_indent (enc);
314		587
315	if (svp)	588	if (svp)
316	encode_sv (enc, *svp);	589	encode_sv (enc, *svp);
317	else	590	else
318	encode_str (enc, "null", 4, 0);	591	encode_str (enc, "null", 4, 0);
319		592
320	if (i < len)	593	if (i < len)
321	encode_comma (enc);	594	encode_comma (enc);
322	}	595	}
323		596
		597	encode_nl (enc); --enc->indent; encode_indent (enc);
		598	}
		599
324	encode_nl (enc);	600	encode_ch (enc, ']');
325
326	--enc->indent;
327	encode_indent (enc); encode_ch (enc, ']');
328	}	601	}
329		602
330	static void	603	static void
331	encode_he (enc_t *enc, HE *he)	604	encode_hk (enc_t *enc, HE *he)
332	{	605	{
333	encode_ch (enc, '"');	606	encode_ch (enc, '"');
334		607
335	if (HeKLEN (he) == HEf_SVKEY)	608	if (HeKLEN (he) == HEf_SVKEY)
336	{	609	{
337	SV *sv = HeSVKEY (he);	610	SV *sv = HeSVKEY (he);
338	STRLEN len;	611	STRLEN len;
339	char *str;	612	char *str;
340		613
341	SvGETMAGIC (sv);	614	SvGETMAGIC (sv);
342	str = SvPV (sv, len);	615	str = SvPV (sv, len);
343		616
344	encode_str (enc, str, len, SvUTF8 (sv));	617	encode_str (enc, str, len, SvUTF8 (sv));
345	}	618	}
…		…
349	encode_ch (enc, '"');	622	encode_ch (enc, '"');
350		623
351	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);	624	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
352	encode_ch (enc, ':');	625	encode_ch (enc, ':');
353	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);	626	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
354	encode_sv (enc, HeVAL (he));
355	}	627	}
356		628
357	// compare hash entries, used when all keys are bytestrings	629	// compare hash entries, used when all keys are bytestrings
358	static int	630	static int
359	he_cmp_fast (const void *a_, const void *b_)	631	he_cmp_fast (const void *a_, const void *b_)
…		…
364	HE *b = *(HE **)b_;	636	HE *b = *(HE **)b_;
365		637
366	STRLEN la = HeKLEN (a);	638	STRLEN la = HeKLEN (a);
367	STRLEN lb = HeKLEN (b);	639	STRLEN lb = HeKLEN (b);
368		640
369	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	641	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
370	cmp = la - lb;	642	cmp = lb - la;
371		643
372	return cmp;	644	return cmp;
373	}	645	}
374		646
375	// 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
376	static int	648	static int
377	he_cmp_slow (const void *a, const void *b)	649	he_cmp_slow (const void *a, const void *b)
378	{	650	{
379	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	651	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
380	}	652	}
381		653
382	static void	654	static void
383	encode_hv (enc_t *enc, HV *hv)	655	encode_hv (enc_t *enc, HV *hv)
384	{	656	{
385	int count, i;	657	HE *he;
386		658
387	if (enc->indent >= enc->maxdepth)	659	if (enc->indent >= enc->json.max_depth)
388	croak ("data structure too deep (hit recursion limit)");	660	croak (ERR_NESTING_EXCEEDED);
389		661
390	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	662	encode_ch (enc, '{');
391		663
392	if ((count = hv_iterinit (hv)))
393	{
394	// for canonical output we have to sort by keys first	664	// for canonical output we have to sort by keys first
395	// actually, this is mostly due to the stupid so-called	665	// actually, this is mostly due to the stupid so-called
396	// security workaround added somewhere in 5.8.x.	666	// security workaround added somewhere in 5.8.x
397	// that randomises hash orderings	667	// that randomises hash orderings
398	if (enc->json.flags & F_CANONICAL)	668	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
		669	{
		670	int count = hv_iterinit (hv);
		671
		672	if (SvMAGICAL (hv))
399	{	673	{
		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	{
400	int fast = 1;	688	int i, fast = 1;
401	HE *he;	689	HE *hes_stack [STACK_HES];
402	#if defined(__BORLANDC__) || defined(_MSC_VER)	690	HE **hes = hes_stack;
403	HE **hes = _alloca (count * sizeof (HE));	691
404	#else	692	// allocate larger arrays on the heap
405	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode	693	if (count > STACK_HES)
406	#endif	694	{
		695	SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
		696	hes = (HE **)SvPVX (sv);
		697	}
407		698
408	i = 0;	699	i = 0;
409	while ((he = hv_iternext (hv)))	700	while ((he = hv_iternext (hv)))
410	{	701	{
411	hes [i++] = he;	702	hes [i++] = he;
…		…
433		724
434	FREETMPS;	725	FREETMPS;
435	LEAVE;	726	LEAVE;
436	}	727	}
437		728
438	for (i = 0; i < count; ++i)	729	encode_nl (enc); ++enc->indent;
		730
		731	while (count--)
439	{	732	{
440	encode_indent (enc);	733	encode_indent (enc);
		734	he = hes [count];
441	encode_he (enc, hes [i]);	735	encode_hk (enc, he);
		736	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
442		737
443	if (i < count - 1)	738	if (count)
444	encode_comma (enc);	739	encode_comma (enc);
445	}	740	}
446		741
447	encode_nl (enc);	742	encode_nl (enc); --enc->indent; encode_indent (enc);
448	}	743	}
		744	}
449	else	745	else
		746	{
		747	if (hv_iterinit (hv) || SvMAGICAL (hv))
		748	if ((he = hv_iternext (hv)))
450	{	749	{
451	HE *he = hv_iternext (hv);	750	encode_nl (enc); ++enc->indent;
452		751
453	for (;;)	752	for (;;)
454	{	753	{
455	encode_indent (enc);	754	encode_indent (enc);
456	encode_he (enc, he);	755	encode_hk (enc, he);
		756	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
457		757
458	if (!(he = hv_iternext (hv)))	758	if (!(he = hv_iternext (hv)))
459	break;	759	break;
460		760
461	encode_comma (enc);	761	encode_comma (enc);
462	}	762	}
463		763
464	encode_nl (enc);	764	encode_nl (enc); --enc->indent; encode_indent (enc);
465	}	765	}
466	}	766	}
467		767
468	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	768	encode_ch (enc, '}');
469	}	769	}
470		770
471	// encode objects, arrays and special \0=false and \1=true values.	771	// encode objects, arrays and special \0=false and \1=true values.
472	static void	772	static void
473	encode_rv (enc_t *enc, SV *sv)	773	encode_rv (enc_t *enc, SV *sv)
474	{	774	{
475	svtype svt;	775	svtype svt;
		776	GV *method;
476		777
477	SvGETMAGIC (sv);	778	SvGETMAGIC (sv);
478	svt = SvTYPE (sv);	779	svt = SvTYPE (sv);
479		780
480	if (expect_false (SvOBJECT (sv)))	781	if (expect_false (SvOBJECT (sv)))
481	{	782	{
482	if (SvSTASH (sv) == json_boolean_stash)	783	HV *stash = SvSTASH (sv);
		784
		785	if (stash == bool_stash)
483	{	786	{
484	if (SvIV (sv))
485	encode_str (enc, "true", 4, 0);	787	if (SvIV (sv)) encode_str (enc, "true" , 4, 0);
486	else
487	encode_str (enc, "false", 5, 0);	788	else encode_str (enc, "false", 5, 0);
488	}	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	PUTBACK;
		832
		833	encode_ch (enc, ']');
		834
		835	FREETMPS; LEAVE;
		836	}
		837	else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
		838	{
		839	dSP;
		840
		841	ENTER; SAVETMPS;
		842	PUSHMARK (SP);
		843	// we re-bless the reference to get overload and other niceties right
		844	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		845
		846	// calling with G_SCALAR ensures that we always get a 1 return value
		847	PUTBACK;
		848	call_sv ((SV *)GvCV (method), G_SCALAR);
		849	SPAGAIN;
		850
		851	// catch this surprisingly common error
		852	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		853	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		854
		855	sv = POPs;
		856	PUTBACK;
		857
		858	encode_sv (enc, sv);
		859
		860	FREETMPS; LEAVE;
		861	}
		862	else if (enc->json.flags & F_ALLOW_BLESSED)
		863	encode_str (enc, "null", 4, 0);
489	else	864	else
490	{
491	#if 0
492	if (0 && sv_derived_from (rv, "JSON::Literal"))
493	{
494	// not yet
495	}
496	#endif
497	if (enc->json.flags & F_CONV_BLESSED)
498	{
499	// we re-bless the reference to get overload and other niceties right
500	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
501
502	if (to_json)
503	{
504	int count;
505	dSP;
506
507	ENTER; SAVETMPS; PUSHMARK (SP);
508	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
509
510	// calling with G_SCALAR ensures that we always get a 1 return value
511	PUTBACK;
512	call_sv ((SV *)GvCV (to_json), G_SCALAR);
513	SPAGAIN;
514
515	// catch this surprisingly common error
516	if (SvROK (TOPs) && SvRV (TOPs) == sv)
517	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
518
519	sv = POPs;
520	PUTBACK;
521
522	encode_sv (enc, sv);
523
524	FREETMPS; LEAVE;
525	}
526	else if (enc->json.flags & F_ALLOW_BLESSED)
527	encode_str (enc, "null", 4, 0);
528	else
529	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
530	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
531	}
532	else if (enc->json.flags & F_ALLOW_BLESSED)
533	encode_str (enc, "null", 4, 0);
534	else
535	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",	865	croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
536	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	866	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
537	}
538	}	867	}
539	else if (svt == SVt_PVHV)	868	else if (svt == SVt_PVHV)
540	encode_hv (enc, (HV *)sv);	869	encode_hv (enc, (HV *)sv);
541	else if (svt == SVt_PVAV)	870	else if (svt == SVt_PVAV)
542	encode_av (enc, (AV *)sv);	871	encode_av (enc, (AV *)sv);
543	else if (svt < SVt_PVAV)	872	else if (svt < SVt_PVAV)
544	{	873	{
545	STRLEN len = 0;	874	int bool_type = ref_bool_type (sv);
546	char *pv = svt ? SvPV (sv, len) : 0;
547		875
548	if (len == 1 && *pv == '1')	876	if (bool_type == 1)
549	encode_str (enc, "true", 4, 0);	877	encode_str (enc, "true", 4, 0);
550	else if (len == 1 && *pv == '0')	878	else if (bool_type == 0)
551	encode_str (enc, "false", 5, 0);	879	encode_str (enc, "false", 5, 0);
		880	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		881	encode_str (enc, "null", 4, 0);
552	else	882	else
553	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	883	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
554	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	884	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
555	}	885	}
		886	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		887	encode_str (enc, "null", 4, 0);
556	else	888	else
557	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	889	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
558	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	890	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
559	}	891	}
560		892
…		…
578	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	910	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
579	enc->cur += strlen (enc->cur);	911	enc->cur += strlen (enc->cur);
580	}	912	}
581	else if (SvIOKp (sv))	913	else if (SvIOKp (sv))
582	{	914	{
583	// we assume we can always read an IV as a UV	915	// we assume we can always read an IV as a UV and vice versa
584	if (SvUV (sv) & ~(UV)0x7fff)	916	// we assume two's complement
585	{	917	// we assume no aliasing issues in the union
586	// large integer, use the (rather slow) snprintf way.	918	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
587	need (enc, sizeof (UV) * 3);	919	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
588	enc->cur +=
589	SvIsUV(sv)
590	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
591	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
592	}
593	else
594	{	920	{
595	// optimise the "small number case"	921	// optimise the "small number case"
596	// code will likely be branchless and use only a single multiplication	922	// code will likely be branchless and use only a single multiplication
		923	// 4.28 works for numbers up to 59074
597	I32 i = SvIV (sv);	924	I32 i = SvIVX (sv);
598	U32 u;	925	U32 u;
599	char digit, nz = 0;	926	char digit, nz = 0;
600		927
601	need (enc, 6);	928	need (enc, 6);
602		929
…		…
608		935
609	// now output digit by digit, each time masking out the integer part	936	// now output digit by digit, each time masking out the integer part
610	// and multiplying by 5 while moving the decimal point one to the right,	937	// and multiplying by 5 while moving the decimal point one to the right,
611	// resulting in a net multiplication by 10.	938	// resulting in a net multiplication by 10.
612	// we always write the digit to memory but conditionally increment	939	// we always write the digit to memory but conditionally increment
613	// the pointer, to ease the usage of conditional move instructions.	940	// the pointer, to enable the use of conditional move instructions.
614	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;	941	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
615	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;	942	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
616	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;	943	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
617	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;	944	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
618	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'	945	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
619	}	946	}
		947	else
		948	{
		949	// large integer, use the (rather slow) snprintf way.
		950	need (enc, IVUV_MAXCHARS);
		951	enc->cur +=
		952	SvIsUV(sv)
		953	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
		954	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		955	}
620	}	956	}
621	else if (SvROK (sv))	957	else if (SvROK (sv))
622	encode_rv (enc, SvRV (sv));	958	encode_rv (enc, SvRV (sv));
623	else if (!SvOK (sv))	959	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
624	encode_str (enc, "null", 4, 0);	960	encode_str (enc, "null", 4, 0);
625	else	961	else
626	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	962	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
627	SvPV_nolen (sv), SvFLAGS (sv));	963	SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
628	}	964	}
629		965
630	static SV *	966	static SV *
631	encode_json (SV *scalar, JSON *json)	967	encode_json (SV *scalar, JSON *json)
632	{	968	{
633	enc_t enc;	969	enc_t enc;
634		970
635	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))	971	if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar))
636	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	972	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
637		973
638	enc.json = *json;	974	enc.json = *json;
639	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	975	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
640	enc.cur = SvPVX (enc.sv);	976	enc.cur = SvPVX (enc.sv);
641	enc.end = SvEND (enc.sv);	977	enc.end = SvEND (enc.sv);
642	enc.indent = 0;	978	enc.indent = 0;
643	enc.maxdepth = DEC_DEPTH (enc.json.flags);	979	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		980	: enc.json.flags & F_LATIN1 ? 0x000100UL
		981	: 0x110000UL;
644		982
645	SvPOK_only (enc.sv);	983	SvPOK_only (enc.sv);
646	encode_sv (&enc, scalar);	984	encode_sv (&enc, scalar);
		985	encode_nl (&enc);
647		986
648	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	987	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
649	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	988	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
650		989
651	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))	990	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
…		…
669	JSON json;	1008	JSON json;
670	U32 depth; // recursion depth	1009	U32 depth; // recursion depth
671	U32 maxdepth; // recursion depth limit	1010	U32 maxdepth; // recursion depth limit
672	} dec_t;	1011	} dec_t;
673		1012
674	inline void	1013	INLINE void
		1014	decode_comment (dec_t *dec)
		1015	{
		1016	// only '#'-style comments allowed a.t.m.
		1017
		1018	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		1019	++dec->cur;
		1020	}
		1021
		1022	INLINE void
675	decode_ws (dec_t *dec)	1023	decode_ws (dec_t *dec)
676	{	1024	{
677	for (;;)	1025	for (;;)
678	{	1026	{
679	char ch = *dec->cur;	1027	char ch = *dec->cur;
680		1028
681	if (ch > 0x20	1029	if (ch > 0x20)
		1030	{
		1031	if (expect_false (ch == '#'))
		1032	{
		1033	if (dec->json.flags & F_RELAXED)
		1034	decode_comment (dec);
		1035	else
		1036	break;
		1037	}
		1038	else
		1039	break;
		1040	}
682	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	1041	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
683	break;	1042	break; // parse error, but let higher level handle it, gives better error messages
684		1043	else
685	++dec->cur;	1044	++dec->cur;
686	}	1045	}
687	}	1046	}
688		1047
689	#define ERR(reason) SB dec->err = reason; goto fail; SE	1048	#define ERR(reason) SB dec->err = reason; goto fail; SE
690		1049
…		…
692	if (*dec->cur != ch) \	1051	if (*dec->cur != ch) \
693	ERR (# ch " expected"); \	1052	ERR (# ch " expected"); \
694	++dec->cur; \	1053	++dec->cur; \
695	SE	1054	SE
696		1055
697	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	1056	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
698	#define DEC_DEC_DEPTH --dec->depth	1057	#define DEC_DEC_DEPTH --dec->depth
699		1058
700	static SV *decode_sv (dec_t *dec);	1059	static SV *decode_sv (dec_t *dec);
701		1060
702	static signed char decode_hexdigit[256];	1061	static signed char decode_hexdigit[256];
…		…
794		1153
795	if (hi >= 0x80)	1154	if (hi >= 0x80)
796	{	1155	{
797	utf8 = 1;	1156	utf8 = 1;
798		1157
799	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	1158	cur = encode_utf8 (cur, hi);
800	}	1159	}
801	else	1160	else
802	*cur++ = hi;	1161	*cur++ = hi;
803	}	1162	}
804	break;	1163	break;
…		…
806	default:	1165	default:
807	--dec_cur;	1166	--dec_cur;
808	ERR ("illegal backslash escape sequence in string");	1167	ERR ("illegal backslash escape sequence in string");
809	}	1168	}
810	}	1169	}
811	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	1170	else if (expect_true (ch >= 0x20 && ch < 0x80))
812	*cur++ = ch;	1171	*cur++ = ch;
813	else if (ch >= 0x80)	1172	else if (ch >= 0x80)
814	{	1173	{
815	STRLEN clen;	1174	STRLEN clen;
816	UV uch;
817		1175
818	--dec_cur;	1176	--dec_cur;
819		1177
820	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);	1178	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
821	if (clen == (STRLEN)-1)	1179	if (clen == (STRLEN)-1)
822	ERR ("malformed UTF-8 character in JSON string");	1180	ERR ("malformed UTF-8 character in JSON string");
823		1181
824	do	1182	do
825	*cur++ = *dec_cur++;	1183	*cur++ = *dec_cur++;
826	while (--clen);	1184	while (--clen);
827		1185
828	utf8 = 1;	1186	utf8 = 1;
829	}	1187	}
		1188	else if (ch == '\t' && dec->json.flags & F_RELAXED)
		1189	*cur++ = ch;
830	else	1190	else
831	{	1191	{
832	--dec_cur;	1192	--dec_cur;
833		1193
834	if (!ch)	1194	if (!ch)
…		…
842	{	1202	{
843	STRLEN len = cur - buf;	1203	STRLEN len = cur - buf;
844		1204
845	if (sv)	1205	if (sv)
846	{	1206	{
847	SvGROW (sv, SvCUR (sv) + len + 1);	1207	STRLEN cur = SvCUR (sv);
		1208
		1209	if (SvLEN (sv) - cur <= len)
		1210	json_sv_grow (sv, cur, len);
		1211
848	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	1212	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
849	SvCUR_set (sv, SvCUR (sv) + len);	1213	SvCUR_set (sv, SvCUR (sv) + len);
850	}	1214	}
851	else	1215	else
852	sv = newSVpvn (buf, len);	1216	sv = newSVpvn (buf, len);
…		…
937	is_nv = 1;	1301	is_nv = 1;
938	}	1302	}
939		1303
940	if (!is_nv)	1304	if (!is_nv)
941	{	1305	{
		1306	int len = dec->cur - start;
		1307
942	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1308	// special case the rather common 1..5-digit-int case
943	if (*start == '-')	1309	if (*start == '-')
944	switch (dec->cur - start)	1310	switch (len)
945	{	1311	{
946	case 2: return newSViv (-( start [1] - '0' * 1));	1312	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
947	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1313	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
948	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));	1314	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
949	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));	1315	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1316	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
950	}	1317	}
951	else	1318	else
952	switch (dec->cur - start)	1319	switch (len)
953	{	1320	{
954	case 1: return newSViv ( start [0] - '0' * 1);	1321	case 1: return newSViv ( start [0] - '0' * 1);
955	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1322	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
956	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);	1323	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
957	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);	1324	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1325	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
958	}	1326	}
959		1327
960	{	1328	{
961	UV uv;	1329	UV uv;
962	int numtype = grok_number (start, dec->cur - start, &uv);	1330	int numtype = grok_number (start, len, &uv);
963	if (numtype & IS_NUMBER_IN_UV)	1331	if (numtype & IS_NUMBER_IN_UV)
964	if (numtype & IS_NUMBER_NEG)	1332	if (numtype & IS_NUMBER_NEG)
965	{	1333	{
966	if (uv < (UV)IV_MIN)	1334	if (uv < (UV)IV_MIN)
967	return newSViv (-(IV)uv);	1335	return newSViv (-(IV)uv);
968	}	1336	}
969	else	1337	else
970	return newSVuv (uv);	1338	return newSVuv (uv);
971
972	// here would likely be the place for bigint support
973	}	1339	}
974	}
975		1340
976	// if we ever support bigint or bigfloat, this is the place for bigfloat	1341	len -= *start == '-' ? 1 : 0;
		1342
		1343	// does not fit into IV or UV, try NV
		1344	if (len <= NV_DIG)
		1345	// fits into NV without loss of precision
		1346	return newSVnv (json_atof (start));
		1347
		1348	// everything else fails, convert it to a string
		1349	return newSVpvn (start, dec->cur - start);
		1350	}
		1351
		1352	// loss of precision here
977	return newSVnv (Atof (start));	1353	return newSVnv (json_atof (start));
978		1354
979	fail:	1355	fail:
980	return 0;	1356	return 0;
981	}	1357	}
982		1358
…		…
1006	if (*dec->cur == ']')	1382	if (*dec->cur == ']')
1007	{	1383	{
1008	++dec->cur;	1384	++dec->cur;
1009	break;	1385	break;
1010	}	1386	}
1011		1387
1012	if (*dec->cur != ',')	1388	if (*dec->cur != ',')
1013	ERR (", or ] expected while parsing array");	1389	ERR (", or ] expected while parsing array");
1014		1390
1015	++dec->cur;	1391	++dec->cur;
		1392
		1393	decode_ws (dec);
		1394
		1395	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1396	{
		1397	++dec->cur;
		1398	break;
		1399	}
1016	}	1400	}
1017		1401
1018	DEC_DEC_DEPTH;	1402	DEC_DEC_DEPTH;
1019	return newRV_noinc ((SV *)av);	1403	return newRV_noinc ((SV *)av);
1020		1404
…		…
1036	if (*dec->cur == '}')	1420	if (*dec->cur == '}')
1037	++dec->cur;	1421	++dec->cur;
1038	else	1422	else
1039	for (;;)	1423	for (;;)
1040	{	1424	{
1041	decode_ws (dec); EXPECT_CH ('"');	1425	EXPECT_CH ('"');
1042		1426
1043	// heuristic: assume that	1427	// heuristic: assume that
1044	// a) decode_str + hv_store_ent are abysmally slow.	1428	// a) decode_str + hv_store_ent are abysmally slow.
1045	// b) most hash keys are short, simple ascii text.	1429	// b) most hash keys are short, simple ascii text.
1046	// => try to "fast-match" such strings to avoid	1430	// => try to "fast-match" such strings to avoid
…		…
1050	char *p = dec->cur;	1434	char *p = dec->cur;
1051	char *e = p + 24; // only try up to 24 bytes	1435	char *e = p + 24; // only try up to 24 bytes
1052		1436
1053	for (;;)	1437	for (;;)
1054	{	1438	{
1055	// the >= 0x80 is true on most architectures	1439	// the >= 0x80 is false on most architectures
1056	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')	1440	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
1057	{	1441	{
1058	// slow path, back up and use decode_str	1442	// slow path, back up and use decode_str
1059	SV *key = decode_str (dec);	1443	SV *key = decode_str (dec);
1060	if (!key)	1444	if (!key)
1061	goto fail;	1445	goto fail;
1062		1446
1063	decode_ws (dec); EXPECT_CH (':');	1447	decode_ws (dec); EXPECT_CH (':');
1064		1448
		1449	decode_ws (dec);
1065	value = decode_sv (dec);	1450	value = decode_sv (dec);
1066	if (!value)	1451	if (!value)
1067	{	1452	{
1068	SvREFCNT_dec (key);	1453	SvREFCNT_dec (key);
1069	goto fail;	1454	goto fail;
…		…
1081	int len = p - key;	1466	int len = p - key;
1082	dec->cur = p + 1;	1467	dec->cur = p + 1;
1083		1468
1084	decode_ws (dec); EXPECT_CH (':');	1469	decode_ws (dec); EXPECT_CH (':');
1085		1470
		1471	decode_ws (dec);
1086	value = decode_sv (dec);	1472	value = decode_sv (dec);
1087	if (!value)	1473	if (!value)
1088	goto fail;	1474	goto fail;
1089		1475
1090	hv_store (hv, key, len, value, 0);	1476	hv_store (hv, key, len, value, 0);
…		…
1106		1492
1107	if (*dec->cur != ',')	1493	if (*dec->cur != ',')
1108	ERR (", or } expected while parsing object/hash");	1494	ERR (", or } expected while parsing object/hash");
1109		1495
1110	++dec->cur;	1496	++dec->cur;
		1497
		1498	decode_ws (dec);
		1499
		1500	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1501	{
		1502	++dec->cur;
		1503	break;
		1504	}
1111	}	1505	}
1112		1506
1113	DEC_DEC_DEPTH;	1507	DEC_DEC_DEPTH;
1114	sv = newRV_noinc ((SV *)hv);	1508	sv = newRV_noinc ((SV *)hv);
1115		1509
1116	// check filter callbacks	1510	// check filter callbacks
1117	if (dec->json.flags & F_HOOK)	1511	if (expect_false (dec->json.flags & F_HOOK))
1118	{	1512	{
1119	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)	1513	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
1120	{	1514	{
1121	HE *cb, *he;	1515	HE *cb, *he;
1122		1516
1123	hv_iterinit (hv);	1517	hv_iterinit (hv);
1124	he = hv_iternext (hv);	1518	he = hv_iternext (hv);
1125	hv_iterinit (hv);	1519	hv_iterinit (hv);
1126		1520
1127	// the next line creates a mortal sv each time its called.	1521	// the next line creates a mortal sv each time it's called.
1128	// might want to optimise this for common cases.	1522	// might want to optimise this for common cases.
1129	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);	1523	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1130		1524
1131	if (cb)	1525	if (cb)
1132	{	1526	{
1133	dSP;	1527	dSP;
1134	int count;	1528	int count;
1135		1529
1136	ENTER; SAVETMPS; PUSHMARK (SP);	1530	ENTER; SAVETMPS;
		1531	PUSHMARK (SP);
1137	XPUSHs (HeVAL (he));	1532	XPUSHs (HeVAL (he));
		1533	sv_2mortal (sv);
1138		1534
1139	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;	1535	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1140		1536
1141	if (count == 1)	1537	if (count == 1)
1142	{	1538	{
1143	sv = newSVsv (POPs);	1539	sv = newSVsv (POPs);
		1540	PUTBACK;
1144	FREETMPS; LEAVE;	1541	FREETMPS; LEAVE;
1145	return sv;	1542	return sv;
1146	}	1543	}
		1544	else if (count)
		1545	croak ("filter_json_single_key_object callbacks must not return more than one scalar");
		1546
		1547	PUTBACK;
		1548
		1549	SvREFCNT_inc (sv);
1147		1550
1148	FREETMPS; LEAVE;	1551	FREETMPS; LEAVE;
1149	}	1552	}
1150	}	1553	}
1151		1554
1152	if (dec->json.cb_object)	1555	if (dec->json.cb_object)
1153	{	1556	{
1154	dSP;	1557	dSP;
1155	int count;	1558	int count;
1156		1559
1157	ENTER; SAVETMPS; PUSHMARK (SP);	1560	ENTER; SAVETMPS;
		1561	PUSHMARK (SP);
1158	XPUSHs (sv_2mortal (sv));	1562	XPUSHs (sv_2mortal (sv));
1159		1563
1160	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;	1564	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1161		1565
1162	if (count == 1)	1566	if (count == 1)
1163	{
1164	sv = newSVsv (POPs);	1567	sv = newSVsv (POPs);
1165	FREETMPS; LEAVE;	1568	else if (count == 0)
1166	return sv;
1167	}
1168
1169	SvREFCNT_inc (sv);	1569	SvREFCNT_inc (sv);
		1570	else
		1571	croak ("filter_json_object callbacks must not return more than one scalar");
		1572
		1573	PUTBACK;
		1574
1170	FREETMPS; LEAVE;	1575	FREETMPS; LEAVE;
1171	}	1576	}
1172	}	1577	}
1173		1578
1174	return sv;	1579	return sv;
…		…
1178	DEC_DEC_DEPTH;	1583	DEC_DEC_DEPTH;
1179	return 0;	1584	return 0;
1180	}	1585	}
1181		1586
1182	static SV *	1587	static SV *
		1588	decode_tag (dec_t *dec)
		1589	{
		1590	SV *tag = 0;
		1591	SV *val = 0;
		1592
		1593	if (!(dec->json.flags & F_ALLOW_TAGS))
		1594	ERR ("malformed JSON string, neither array, object, number, string or atom");
		1595
		1596	++dec->cur;
		1597
		1598	decode_ws (dec);
		1599
		1600	tag = decode_sv (dec);
		1601	if (!tag)
		1602	goto fail;
		1603
		1604	if (!SvPOK (tag))
		1605	ERR ("malformed JSON string, (tag) must be a string");
		1606
		1607	decode_ws (dec);
		1608
		1609	if (*dec->cur != ')')
		1610	ERR (") expected after tag");
		1611
		1612	++dec->cur;
		1613
		1614	decode_ws (dec);
		1615
		1616	val = decode_sv (dec);
		1617	if (!val)
		1618	goto fail;
		1619
		1620	if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
		1621	ERR ("malformed JSON string, tag value must be an array");
		1622
		1623	{
		1624	AV *av = (AV *)SvRV (val);
		1625	int i, len = av_len (av) + 1;
		1626	HV *stash = gv_stashsv (tag, 0);
		1627	SV *sv;
		1628
		1629	if (!stash)
		1630	ERR ("cannot decode perl-object (package does not exist)");
		1631
		1632	GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
		1633
		1634	if (!method)
		1635	ERR ("cannot decode perl-object (package does not have a THAW method)");
		1636
		1637	dSP;
		1638
		1639	ENTER; SAVETMPS;
		1640	PUSHMARK (SP);
		1641	EXTEND (SP, len + 2);
		1642	// we re-bless the reference to get overload and other niceties right
		1643	PUSHs (tag);
		1644	PUSHs (sv_json);
		1645
		1646	for (i = 0; i < len; ++i)
		1647	PUSHs (*av_fetch (av, i, 1));
		1648
		1649	PUTBACK;
		1650	call_sv ((SV *)GvCV (method), G_SCALAR);
		1651	SPAGAIN;
		1652
		1653	SvREFCNT_dec (tag);
		1654	SvREFCNT_dec (val);
		1655	sv = SvREFCNT_inc (POPs);
		1656
		1657	PUTBACK;
		1658
		1659	FREETMPS; LEAVE;
		1660
		1661	return sv;
		1662	}
		1663
		1664	fail:
		1665	SvREFCNT_dec (tag);
		1666	SvREFCNT_dec (val);
		1667	return 0;
		1668	}
		1669
		1670	static SV *
1183	decode_sv (dec_t *dec)	1671	decode_sv (dec_t *dec)
1184	{	1672	{
1185	decode_ws (dec);
1186
1187	// the beauty of JSON: you need exactly one character lookahead	1673	// the beauty of JSON: you need exactly one character lookahead
1188	// to parse anything.	1674	// to parse everything.
1189	switch (*dec->cur)	1675	switch (*dec->cur)
1190	{	1676	{
1191	case '"': ++dec->cur; return decode_str (dec);	1677	case '"': ++dec->cur; return decode_str (dec);
1192	case '[': ++dec->cur; return decode_av (dec);	1678	case '[': ++dec->cur; return decode_av (dec);
1193	case '{': ++dec->cur; return decode_hv (dec);	1679	case '{': ++dec->cur; return decode_hv (dec);
		1680	case '(': return decode_tag (dec);
1194		1681
1195	case '-':	1682	case '-':
1196	case '0': case '1': case '2': case '3': case '4':	1683	case '0': case '1': case '2': case '3': case '4':
1197	case '5': case '6': case '7': case '8': case '9':	1684	case '5': case '6': case '7': case '8': case '9':
1198	return decode_num (dec);	1685	return decode_num (dec);
1199		1686
		1687	case 'f':
		1688	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
		1689	{
		1690	dec->cur += 5;
		1691
		1692	if (expect_false (!dec->json.v_false))
		1693	dec->json.v_false = GET_BOOL (false);
		1694
		1695	return newSVsv (dec->json.v_false);
		1696	}
		1697	else
		1698	ERR ("'false' expected");
		1699
		1700	break;
		1701
1200	case 't':	1702	case 't':
1201	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1703	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1202	{	1704	{
1203	dec->cur += 4;	1705	dec->cur += 4;
1204	return SvREFCNT_inc (json_true);	1706
		1707	if (expect_false (!dec->json.v_true))
		1708	dec->json.v_true = GET_BOOL (true);
		1709
		1710	return newSVsv (dec->json.v_true);
1205	}	1711	}
1206	else	1712	else
1207	ERR ("'true' expected");	1713	ERR ("'true' expected");
1208
1209	break;
1210
1211	case 'f':
1212	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1213	{
1214	dec->cur += 5;
1215	return SvREFCNT_inc (json_false);
1216	}
1217	else
1218	ERR ("'false' expected");
1219		1714
1220	break;	1715	break;
1221		1716
1222	case 'n':	1717	case 'n':
1223	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))	1718	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
…		…
1229	ERR ("'null' expected");	1724	ERR ("'null' expected");
1230		1725
1231	break;	1726	break;
1232		1727
1233	default:	1728	default:
1234	ERR ("malformed JSON string, neither array, object, number, string or atom");	1729	ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1235	break;	1730	break;
1236	}	1731	}
1237		1732
1238	fail:	1733	fail:
1239	return 0;	1734	return 0;
1240	}	1735	}
1241		1736
1242	static SV *	1737	static SV *
1243	decode_json (SV *string, JSON *json, UV *offset_return)	1738	decode_json (SV *string, JSON *json, STRLEN *offset_return)
1244	{	1739	{
1245	dec_t dec;	1740	dec_t dec;
1246	UV offset;
1247	SV *sv;	1741	SV *sv;
1248		1742
		1743	/* work around bugs in 5.10 where manipulating magic values
		1744	* makes perl ignore the magic in subsequent accesses.
		1745	* also make a copy of non-PV values, to get them into a clean
		1746	* state (SvPV should do that, but it's buggy, see below).
		1747	*
		1748	* SvIsCOW_shared_hash works around a bug in perl (possibly 5.16),
		1749	* as reported by Reini Urban.
		1750	*/
1249	SvGETMAGIC (string);	1751	/*SvGETMAGIC (string);*/
		1752	if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
		1753	string = sv_2mortal (newSVsv (string));
		1754
1250	SvUPGRADE (string, SVt_PV);	1755	SvUPGRADE (string, SVt_PV);
1251		1756
1252	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))	1757	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1758	* assertion with -DDEBUGGING, although SvCUR is documented to
		1759	* return the xpv_cur field which certainly exists after upgrading.
		1760	* according to nicholas clark, calling SvPOK fixes this.
		1761	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1762	* and hope for the best.
		1763	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1764	* assertion business is seriously broken, try yet another workaround
		1765	* for the broken -DDEBUGGING.
		1766	*/
		1767	{
		1768	#ifdef DEBUGGING
		1769	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1770	#else
		1771	STRLEN offset = SvCUR (string);
		1772	#endif
		1773
		1774	if (offset > json->max_size && json->max_size)
1253	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",	1775	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1254	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));	1776	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1777	}
1255		1778
1256	if (json->flags & F_UTF8)	1779	if (DECODE_WANTS_OCTETS (json))
1257	sv_utf8_downgrade (string, 0);	1780	sv_utf8_downgrade (string, 0);
1258	else	1781	else
1259	sv_utf8_upgrade (string);	1782	sv_utf8_upgrade (string);
1260		1783
1261	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1784	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1262		1785
1263	dec.json = *json;	1786	dec.json = *json;
1264	dec.cur = SvPVX (string);	1787	dec.cur = SvPVX (string);
1265	dec.end = SvEND (string);	1788	dec.end = SvEND (string);
1266	dec.err = 0;	1789	dec.err = 0;
1267	dec.depth = 0;	1790	dec.depth = 0;
1268	dec.maxdepth = DEC_DEPTH (dec.json.flags);
1269		1791
1270	if (dec.json.cb_object || dec.json.cb_sk_object)	1792	if (dec.json.cb_object || dec.json.cb_sk_object)
1271	dec.json.flags |= F_HOOK;	1793	dec.json.flags |= F_HOOK;
1272		1794
1273	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1795	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1796
		1797	decode_ws (&dec);
1274	sv = decode_sv (&dec);	1798	sv = decode_sv (&dec);
1275		1799
1276	if (!(offset_return || !sv))	1800	if (offset_return)
		1801	*offset_return = dec.cur - SvPVX (string);
		1802	else if (sv)
1277	{	1803	{
1278	// check for trailing garbage	1804	// check for trailing garbage
1279	decode_ws (&dec);	1805	decode_ws (&dec);
1280		1806
1281	if (*dec.cur)	1807	if (dec.cur != dec.end)
1282	{	1808	{
1283	dec.err = "garbage after JSON object";	1809	dec.err = "garbage after JSON object";
1284	SvREFCNT_dec (sv);	1810	SvREFCNT_dec (sv);
1285	sv = 0;	1811	sv = 0;
1286	}	1812	}
1287	}
1288
1289	if (offset_return || !sv)
1290	{
1291	offset = dec.json.flags & F_UTF8
1292	? dec.cur - SvPVX (string)
1293	: utf8_distance (dec.cur, SvPVX (string));
1294
1295	if (offset_return)
1296	*offset_return = offset;
1297	}	1813	}
1298		1814
1299	if (!sv)	1815	if (!sv)
1300	{	1816	{
1301	SV *uni = sv_newmortal ();	1817	SV *uni = sv_newmortal ();
…		…
1307	SAVEVPTR (PL_curcop);	1823	SAVEVPTR (PL_curcop);
1308	PL_curcop = &cop;	1824	PL_curcop = &cop;
1309	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1825	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1310	LEAVE;	1826	LEAVE;
1311		1827
1312	croak ("%s, at character offset %d [\"%s\"]",	1828	croak ("%s, at character offset %d (before \"%s\")",
1313	dec.err,	1829	dec.err,
1314	(int)offset,	1830	(int)ptr_to_index (string, dec.cur),
1315	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1831	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1316	}	1832	}
1317		1833
1318	sv = sv_2mortal (sv);	1834	sv = sv_2mortal (sv);
1319		1835
1320	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))	1836	if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv))
1321	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1837	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1322		1838
1323	return sv;	1839	return sv;
		1840	}
		1841
		1842	/////////////////////////////////////////////////////////////////////////////
		1843	// incremental parser
		1844
		1845	static void
		1846	incr_parse (JSON *self)
		1847	{
		1848	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1849
		1850	// the state machine here is a bit convoluted and could be simplified a lot
		1851	// but this would make it slower, so...
		1852
		1853	for (;;)
		1854	{
		1855	switch (self->incr_mode)
		1856	{
		1857	// reached end of a scalar, see if we are inside a nested structure or not
		1858	end_of_scalar:
		1859	self->incr_mode = INCR_M_JSON;
		1860
		1861	if (self->incr_nest) // end of a scalar inside array, object or tag
		1862	goto incr_m_json;
		1863	else // end of scalar outside structure, json text ends here
		1864	goto interrupt;
		1865
		1866	// only used for initial whitespace skipping
		1867	case INCR_M_WS:
		1868	for (;;)
		1869	{
		1870	if (*p > 0x20)
		1871	{
		1872	if (*p == '#')
		1873	{
		1874	self->incr_mode = INCR_M_C0;
		1875	goto incr_m_c;
		1876	}
		1877	else
		1878	{
		1879	self->incr_mode = INCR_M_JSON;
		1880	goto incr_m_json;
		1881	}
		1882	}
		1883	else if (!*p)
		1884	goto interrupt;
		1885
		1886	++p;
		1887	}
		1888
		1889	// skip a single char inside a string (for \\-processing)
		1890	case INCR_M_BS:
		1891	if (!*p)
		1892	goto interrupt;
		1893
		1894	++p;
		1895	self->incr_mode = INCR_M_STR;
		1896	goto incr_m_str;
		1897
		1898	// inside #-style comments
		1899	case INCR_M_C0:
		1900	case INCR_M_C1:
		1901	incr_m_c:
		1902	for (;;)
		1903	{
		1904	if (*p == '\n')
		1905	{
		1906	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1907	break;
		1908	}
		1909	else if (!*p)
		1910	goto interrupt;
		1911
		1912	++p;
		1913	}
		1914
		1915	break;
		1916
		1917	// inside true/false/null
		1918	case INCR_M_TFN:
		1919	incr_m_tfn:
		1920	for (;;)
		1921	switch (*p++)
		1922	{
		1923	case 'r': case 'u': case 'e': // tRUE, falsE, nUll
		1924	case 'a': case 'l': case 's': // fALSe, nuLL
		1925	// allowed
		1926	break;
		1927
		1928	default:
		1929	--p;
		1930	goto end_of_scalar;
		1931	}
		1932
		1933	// inside a number
		1934	case INCR_M_NUM:
		1935	incr_m_num:
		1936	for (;;)
		1937	switch (*p++)
		1938	{
		1939	case 'e': case 'E': case '.': case '+':
		1940	case '-':
		1941	case '0': case '1': case '2': case '3': case '4':
		1942	case '5': case '6': case '7': case '8': case '9':
		1943	// allowed
		1944	break;
		1945
		1946	default:
		1947	--p;
		1948	goto end_of_scalar;
		1949	}
		1950
		1951	// inside a string
		1952	case INCR_M_STR:
		1953	incr_m_str:
		1954	for (;;)
		1955	{
		1956	if (*p == '"')
		1957	{
		1958	++p;
		1959	goto end_of_scalar;
		1960	}
		1961	else if (*p == '\\')
		1962	{
		1963	++p; // "virtually" consumes character after \
		1964
		1965	if (!*p) // if at end of string we have to switch modes
		1966	{
		1967	self->incr_mode = INCR_M_BS;
		1968	goto interrupt;
		1969	}
		1970	}
		1971	else if (!*p)
		1972	goto interrupt;
		1973
		1974	++p;
		1975	}
		1976
		1977	// after initial ws, outside string
		1978	case INCR_M_JSON:
		1979	incr_m_json:
		1980	for (;;)
		1981	{
		1982	switch (*p++)
		1983	{
		1984	case 0:
		1985	--p;
		1986	goto interrupt;
		1987
		1988	case 0x09:
		1989	case 0x0a:
		1990	case 0x0d:
		1991	case 0x20:
		1992	if (!self->incr_nest)
		1993	{
		1994	--p; // do not eat the whitespace, let the next round do it
		1995	goto interrupt;
		1996	}
		1997	break;
		1998
		1999	// the following three blocks handle scalars. this makes the parser
		2000	// more strict than required inside arrays or objects, and could
		2001	// be moved to a special case on the toplevel (except strings)
		2002	case 't':
		2003	case 'f':
		2004	case 'n':
		2005	self->incr_mode = INCR_M_TFN;
		2006	goto incr_m_tfn;
		2007
		2008	case '-':
		2009	case '0': case '1': case '2': case '3': case '4':
		2010	case '5': case '6': case '7': case '8': case '9':
		2011	self->incr_mode = INCR_M_NUM;
		2012	goto incr_m_num;
		2013
		2014	case '"':
		2015	self->incr_mode = INCR_M_STR;
		2016	goto incr_m_str;
		2017
		2018	case '[':
		2019	case '{':
		2020	case '(':
		2021	if (++self->incr_nest > self->max_depth)
		2022	croak (ERR_NESTING_EXCEEDED);
		2023	break;
		2024
		2025	case ']':
		2026	case '}':
		2027	if (--self->incr_nest <= 0)
		2028	goto interrupt;
		2029	break;
		2030
		2031	case ')':
		2032	--self->incr_nest;
		2033	break;
		2034
		2035	case '#':
		2036	self->incr_mode = INCR_M_C1;
		2037	goto incr_m_c;
		2038	}
		2039	}
		2040	}
		2041
		2042	modechange:
		2043	;
		2044	}
		2045
		2046	interrupt:
		2047	self->incr_pos = p - SvPVX (self->incr_text);
		2048	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		2049	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1324	}	2050	}
1325		2051
1326	/////////////////////////////////////////////////////////////////////////////	2052	/////////////////////////////////////////////////////////////////////////////
1327	// XS interface functions	2053	// XS interface functions
1328		2054
…		…
1337	i >= '0' && i <= '9' ? i - '0'	2063	i >= '0' && i <= '9' ? i - '0'
1338	: i >= 'a' && i <= 'f' ? i - 'a' + 10	2064	: i >= 'a' && i <= 'f' ? i - 'a' + 10
1339	: i >= 'A' && i <= 'F' ? i - 'A' + 10	2065	: i >= 'A' && i <= 'F' ? i - 'A' + 10
1340	: -1;	2066	: -1;
1341		2067
1342	json_stash = gv_stashpv ("JSON::XS" , 1);	2068	json_stash = gv_stashpv ("JSON::XS" , 1);
1343	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);	2069	bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
		2070	bool_false = get_bool ("Types::Serialiser::false");
		2071	bool_true = get_bool ("Types::Serialiser::true");
1344		2072
1345	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );	2073	sv_json = newSVpv ("JSON", 0);
1346	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	2074	SvREADONLY_on (sv_json);
		2075
		2076	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1347	}	2077	}
1348		2078
1349	PROTOTYPES: DISABLE	2079	PROTOTYPES: DISABLE
		2080
		2081	void CLONE (...)
		2082	CODE:
		2083	// as long as these writes are atomic, the race should not matter
		2084	// as existing threads either already use 0, or use the old value,
		2085	// which is sitll correct for the initial thread.
		2086	json_stash = 0;
		2087	bool_stash = 0;
		2088	bool_false = 0;
		2089	bool_true = 0;
1350		2090
1351	void new (char *klass)	2091	void new (char *klass)
1352	PPCODE:	2092	PPCODE:
1353	{	2093	{
1354	SV *pv = NEWSV (0, sizeof (JSON));	2094	SV *pv = NEWSV (0, sizeof (JSON));
1355	SvPOK_only (pv);	2095	SvPOK_only (pv);
1356	Zero (SvPVX (pv), 1, JSON);	2096	json_init ((JSON *)SvPVX (pv));
1357	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	2097	XPUSHs (sv_2mortal (sv_bless (
1358	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash)));	2098	newRV_noinc (pv),
		2099	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		2100	)));
1359	}	2101	}
		2102
		2103	void boolean_values (JSON *self, SV *v_false = 0, SV *v_true = 0)
		2104	PPCODE:
		2105	self->v_false = newSVsv (v_false);
		2106	self->v_true = newSVsv (v_true);
		2107	XPUSHs (ST (0));
		2108
		2109	void get_boolean_values (JSON *self)
		2110	PPCODE:
		2111	if (self->v_false && self->v_true)
		2112	{
		2113	EXTEND (SP, 2);
		2114	PUSHs (self->v_false);
		2115	PUSHs (self->v_true);
		2116	}
1360		2117
1361	void ascii (JSON *self, int enable = 1)	2118	void ascii (JSON *self, int enable = 1)
1362	ALIAS:	2119	ALIAS:
1363	ascii = F_ASCII	2120	ascii = F_ASCII
1364	latin1 = F_LATIN1	2121	latin1 = F_LATIN1
…		…
1370	pretty = F_PRETTY	2127	pretty = F_PRETTY
1371	allow_nonref = F_ALLOW_NONREF	2128	allow_nonref = F_ALLOW_NONREF
1372	shrink = F_SHRINK	2129	shrink = F_SHRINK
1373	allow_blessed = F_ALLOW_BLESSED	2130	allow_blessed = F_ALLOW_BLESSED
1374	convert_blessed = F_CONV_BLESSED	2131	convert_blessed = F_CONV_BLESSED
		2132	relaxed = F_RELAXED
		2133	allow_unknown = F_ALLOW_UNKNOWN
		2134	allow_tags = F_ALLOW_TAGS
1375	PPCODE:	2135	PPCODE:
1376	{	2136	{
1377	if (enable)	2137	if (enable)
1378	self->flags |= ix;	2138	self->flags |= ix;
1379	else	2139	else
1380	self->flags &= ~ix;	2140	self->flags &= ~ix;
1381		2141
1382	XPUSHs (ST (0));	2142	XPUSHs (ST (0));
1383	}	2143	}
1384		2144
1385	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	2145	void get_ascii (JSON *self)
		2146	ALIAS:
		2147	get_ascii = F_ASCII
		2148	get_latin1 = F_LATIN1
		2149	get_utf8 = F_UTF8
		2150	get_indent = F_INDENT
		2151	get_canonical = F_CANONICAL
		2152	get_space_before = F_SPACE_BEFORE
		2153	get_space_after = F_SPACE_AFTER
		2154	get_allow_nonref = F_ALLOW_NONREF
		2155	get_shrink = F_SHRINK
		2156	get_allow_blessed = F_ALLOW_BLESSED
		2157	get_convert_blessed = F_CONV_BLESSED
		2158	get_relaxed = F_RELAXED
		2159	get_allow_unknown = F_ALLOW_UNKNOWN
		2160	get_allow_tags = F_ALLOW_TAGS
1386	PPCODE:	2161	PPCODE:
1387	{	2162	XPUSHs (boolSV (self->flags & ix));
1388	UV log2 = 0;
1389		2163
1390	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	2164	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1391		2165	PPCODE:
1392	while ((1UL << log2) < max_depth)	2166	self->max_depth = max_depth;
1393	++log2;
1394
1395	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1396
1397	XPUSHs (ST (0));	2167	XPUSHs (ST (0));
1398	}
1399		2168
		2169	U32 get_max_depth (JSON *self)
		2170	CODE:
		2171	RETVAL = self->max_depth;
		2172	OUTPUT:
		2173	RETVAL
		2174
1400	void max_size (JSON *self, UV max_size = 0)	2175	void max_size (JSON *self, U32 max_size = 0)
1401	PPCODE:	2176	PPCODE:
1402	{	2177	self->max_size = max_size;
1403	UV log2 = 0;
1404
1405	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1406	if (max_size == 1) max_size = 2;
1407
1408	while ((1UL << log2) < max_size)
1409	++log2;
1410
1411	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1412
1413	XPUSHs (ST (0));	2178	XPUSHs (ST (0));
1414	}	2179
		2180	int get_max_size (JSON *self)
		2181	CODE:
		2182	RETVAL = self->max_size;
		2183	OUTPUT:
		2184	RETVAL
1415		2185
1416	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)	2186	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1417	PPCODE:	2187	PPCODE:
1418	{	2188	{
1419	SvREFCNT_dec (self->cb_object);	2189	SvREFCNT_dec (self->cb_object);
…		…
1423	}	2193	}
1424		2194
1425	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)	2195	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1426	PPCODE:	2196	PPCODE:
1427	{	2197	{
1428	if (!self->cb_sk_object)	2198	if (!self->cb_sk_object)
1429	self->cb_sk_object = newHV ();	2199	self->cb_sk_object = newHV ();
1430		2200
1431	if (SvOK (cb))	2201	if (SvOK (cb))
1432	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);	2202	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1433	else	2203	else
…		…
1444	XPUSHs (ST (0));	2214	XPUSHs (ST (0));
1445	}	2215	}
1446		2216
1447	void encode (JSON *self, SV *scalar)	2217	void encode (JSON *self, SV *scalar)
1448	PPCODE:	2218	PPCODE:
1449	XPUSHs (encode_json (scalar, self));	2219	PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
		2220	XPUSHs (scalar);
1450		2221
1451	void decode (JSON *self, SV *jsonstr)	2222	void decode (JSON *self, SV *jsonstr)
1452	PPCODE:	2223	PPCODE:
1453	XPUSHs (decode_json (jsonstr, self, 0));	2224	PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
		2225	XPUSHs (jsonstr);
1454		2226
1455	void decode_prefix (JSON *self, SV *jsonstr)	2227	void decode_prefix (JSON *self, SV *jsonstr)
1456	PPCODE:	2228	PPCODE:
1457	{	2229	{
1458	UV offset;	2230	SV *sv;
		2231	STRLEN offset;
		2232	PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1459	EXTEND (SP, 2);	2233	EXTEND (SP, 2);
1460	PUSHs (decode_json (jsonstr, self, &offset));	2234	PUSHs (sv);
1461	PUSHs (sv_2mortal (newSVuv (offset)));	2235	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset))));
		2236	}
		2237
		2238	void incr_parse (JSON *self, SV *jsonstr = 0)
		2239	PPCODE:
		2240	{
		2241	if (!self->incr_text)
		2242	self->incr_text = newSVpvn ("", 0);
		2243
		2244	/* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
		2245	if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
		2246	if (DECODE_WANTS_OCTETS (self))
		2247	{
		2248	if (self->incr_pos)
		2249	self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
		2250	(U8 *)SvPVX (self->incr_text) + self->incr_pos);
		2251
		2252	sv_utf8_downgrade (self->incr_text, 0);
		2253	}
		2254	else
		2255	{
		2256	sv_utf8_upgrade (self->incr_text);
		2257
		2258	if (self->incr_pos)
		2259	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		2260	- (U8 *)SvPVX (self->incr_text);
		2261	}
		2262
		2263	// append data, if any
		2264	if (jsonstr)
		2265	{
		2266	/* make sure both strings have same encoding */
		2267	if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
		2268	if (SvUTF8 (jsonstr))
		2269	sv_utf8_downgrade (jsonstr, 0);
		2270	else
		2271	sv_utf8_upgrade (jsonstr);
		2272
		2273	/* and then just blindly append */
		2274	{
		2275	STRLEN len;
		2276	const char *str = SvPV (jsonstr, len);
		2277	STRLEN cur = SvCUR (self->incr_text);
		2278
		2279	if (SvLEN (self->incr_text) - cur <= len)
		2280	json_sv_grow (self->incr_text, cur, len);
		2281
		2282	Move (str, SvEND (self->incr_text), len, char);
		2283	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		2284	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		2285	}
		2286	}
		2287
		2288	if (GIMME_V != G_VOID)
		2289	do
		2290	{
		2291	SV *sv;
		2292	STRLEN offset;
		2293
		2294	if (!INCR_DONE (self))
		2295	{
		2296	incr_parse (self);
		2297
		2298	if (self->incr_pos > self->max_size && self->max_size)
		2299	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		2300	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		2301
		2302	if (!INCR_DONE (self))
		2303	{
		2304	// as an optimisation, do not accumulate white space in the incr buffer
		2305	if (self->incr_mode == INCR_M_WS && self->incr_pos)
		2306	{
		2307	self->incr_pos = 0;
		2308	SvCUR_set (self->incr_text, 0);
		2309	}
		2310
		2311	break;
		2312	}
		2313	}
		2314
		2315	PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
		2316	XPUSHs (sv);
		2317
		2318	self->incr_pos -= offset;
		2319	self->incr_nest = 0;
		2320	self->incr_mode = 0;
		2321
		2322	sv_chop (self->incr_text, SvPVX (self->incr_text) + offset);
		2323	}
		2324	while (GIMME_V == G_ARRAY);
		2325	}
		2326
		2327	SV *incr_text (JSON *self)
		2328	ATTRS: lvalue
		2329	CODE:
		2330	{
		2331	if (self->incr_pos)
		2332	croak ("incr_text can not be called when the incremental parser already started parsing");
		2333
		2334	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		2335	}
		2336	OUTPUT:
		2337	RETVAL
		2338
		2339	void incr_skip (JSON *self)
		2340	CODE:
		2341	{
		2342	if (self->incr_pos)
		2343	{
		2344	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2345	self->incr_pos = 0;
		2346	self->incr_nest = 0;
		2347	self->incr_mode = 0;
		2348	}
		2349	}
		2350
		2351	void incr_reset (JSON *self)
		2352	CODE:
		2353	{
		2354	SvREFCNT_dec (self->incr_text);
		2355	self->incr_text = 0;
		2356	self->incr_pos = 0;
		2357	self->incr_nest = 0;
		2358	self->incr_mode = 0;
1462	}	2359	}
1463		2360
1464	void DESTROY (JSON *self)	2361	void DESTROY (JSON *self)
1465	CODE:	2362	CODE:
		2363	SvREFCNT_dec (self->v_false);
		2364	SvREFCNT_dec (self->v_true);
1466	SvREFCNT_dec (self->cb_sk_object);	2365	SvREFCNT_dec (self->cb_sk_object);
1467	SvREFCNT_dec (self->cb_object);	2366	SvREFCNT_dec (self->cb_object);
		2367	SvREFCNT_dec (self->incr_text);
1468		2368
1469	PROTOTYPES: ENABLE	2369	PROTOTYPES: ENABLE
1470		2370
1471	void to_json (SV *scalar)	2371	void encode_json (SV *scalar)
1472	PPCODE:	2372	PPCODE:
1473	{	2373	{
1474	JSON json = { F_DEFAULT | F_UTF8 };	2374	JSON json;
1475	XPUSHs (encode_json (scalar, &json));	2375	json_init (&json);
		2376	json.flags |= F_UTF8;
		2377	PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
		2378	XPUSHs (scalar);
1476	}	2379	}
1477		2380
1478	void from_json (SV *jsonstr)	2381	void decode_json (SV *jsonstr)
1479	PPCODE:	2382	PPCODE:
1480	{	2383	{
1481	JSON json = { F_DEFAULT | F_UTF8 };	2384	JSON json;
1482	XPUSHs (decode_json (jsonstr, &json, 0));	2385	json_init (&json);
		2386	json.flags |= F_UTF8;
		2387	PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
		2388	XPUSHs (jsonstr);
1483	}	2389	}
1484		2390

Diff Legend

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