ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.32 by root, Wed May 9 16:41:12 2007 UTC vs.
Revision 1.135 by root, Thu Nov 15 20:49:12 2018 UTC

1	#include "EXTERN.h"	1	#include "EXTERN.h"
2	#include "perl.h"	2	#include "perl.h"
3	#include "XSUB.h"	3	#include "XSUB.h"
4		4
5	#include "assert.h"	5	#include <assert.h>
6	#include "string.h"	6	#include <string.h>
7	#include "stdlib.h"	7	#include <stdlib.h>
8	#include "stdio.h"	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
		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
14	#define F_ASCII 0x00000001UL	37	#define F_ASCII 0x00000001UL
15	#define F_LATIN1 0x00000002UL	38	#define F_LATIN1 0x00000002UL
16	#define F_UTF8 0x00000004UL	39	#define F_UTF8 0x00000004UL
17	#define F_INDENT 0x00000008UL	40	#define F_INDENT 0x00000008UL
18	#define F_CANONICAL 0x00000010UL	41	#define F_CANONICAL 0x00000010UL
19	#define F_SPACE_BEFORE 0x00000020UL	42	#define F_SPACE_BEFORE 0x00000020UL
20	#define F_SPACE_AFTER 0x00000040UL	43	#define F_SPACE_AFTER 0x00000040UL
21	#define F_ALLOW_NONREF 0x00000100UL	44	#define F_ALLOW_NONREF 0x00000100UL
22	#define F_SHRINK 0x00000200UL	45	#define F_SHRINK 0x00000200UL
		46	#define F_ALLOW_BLESSED 0x00000400UL
		47	#define F_CONV_BLESSED 0x00000800UL
23	#define F_MAXDEPTH 0xf8000000UL	48	#define F_RELAXED 0x00001000UL
24	#define S_MAXDEPTH 27	49	#define F_ALLOW_UNKNOWN 0x00002000UL
25		50	#define F_ALLOW_TAGS 0x00004000UL
26	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))	51	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
27
28	// F_SELFCONVERT? <=> to_json/toJson
29	// F_BLESSED? <=> { $__class__$ => }
30		52
31	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	53	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
32	#define F_DEFAULT (9UL << S_MAXDEPTH)
33		54
34	#define INIT_SIZE 32 // initial scalar size to be allocated	55	#define INIT_SIZE 64 // initial scalar size to be allocated
35	#define INDENT_STEP 3 // spaces per indentation level	56	#define INDENT_STEP 3 // spaces per indentation level
36		57
37	#define SHORT_STRING_LEN 512 // 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)
38		61
39	#define SB do {	62	#define SB do {
40	#define SE } while (0)	63	#define SE } while (0)
41		64
42	static HV *json_stash; // JSON::XS::	65	#if __GNUC__ >= 3
		66	# define expect(expr,value) __builtin_expect ((expr), (value))
		67	# define INLINE static inline
		68	#else
		69	# define expect(expr,value) (expr)
		70	# define INLINE static
		71	#endif
		72
		73	#define expect_false(expr) expect ((expr) != 0, 0)
		74	#define expect_true(expr) expect ((expr) != 0, 1)
		75
		76	#define IN_RANGE_INC(type,val,beg,end) \
		77	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		78	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		79
		80	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
		81
		82	#ifdef USE_ITHREADS
		83	# define JSON_STASH (expect_true (json_stash) ? json_stash : gv_stashpv ("JSON::XS", 1))
		84	# define BOOL_STASH (expect_true (bool_stash) ? bool_stash : gv_stashpv ("Types::Serialiser::Boolean", 1))
		85	# define GET_BOOL(value) (expect_true (bool_ ## value) ? bool_ ## value : get_bool ("Types::Serialiser::" # value))
		86	#else
		87	# define JSON_STASH json_stash
		88	# define BOOL_STASH bool_stash
		89	# define GET_BOOL(value) bool_ ## value
		90	#endif
		91
		92	// the amount of HEs to allocate on the stack, when sorting keys
		93	#define STACK_HES 64
		94
		95	static HV *json_stash, *bool_stash; // JSON::XS::, Types::Serialiser::Boolean::
		96	static SV *bool_false, *bool_true;
		97	static SV *sv_json;
		98
		99	enum {
		100	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		101	INCR_M_STR, // inside string
		102	INCR_M_BS, // inside backslash
		103	INCR_M_C0, // inside comment in initial whitespace sequence
		104	INCR_M_C1, // inside comment in other places
		105	INCR_M_JSON // outside anything, count nesting
		106	};
		107
		108	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		109
		110	typedef struct {
		111	U32 flags;
		112	U32 max_depth;
		113	STRLEN max_size;
		114
		115	SV *cb_object;
		116	HV *cb_sk_object;
		117
		118	// for the incremental parser
		119	SV *incr_text; // the source text so far
		120	STRLEN incr_pos; // the current offset into the text
		121	int incr_nest; // {[]}-nesting level
		122	unsigned char incr_mode;
		123
		124	SV *v_false, *v_true;
		125	} JSON;
		126
		127	INLINE void
		128	json_init (JSON *json)
		129	{
		130	static const JSON init = { F_ALLOW_NONREF, 512 };
		131
		132	*json = init;
		133	}
43		134
44	/////////////////////////////////////////////////////////////////////////////	135	/////////////////////////////////////////////////////////////////////////////
45	// utility functions	136	// utility functions
46		137
47	static UV *	138	INLINE SV *
48	SvJSON (SV *sv)	139	get_bool (const char *name)
49	{	140	{
50	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))	141	SV *sv = get_sv (name, 1);
51	croak ("object is not of type JSON::XS");
52		142
53	return &SvUVX (SvRV (sv));	143	SvREADONLY_on (sv);
54	}	144	SvREADONLY_on (SvRV (sv));
55		145
56	static void	146	return sv;
		147	}
		148
		149	INLINE void
57	shrink (SV *sv)	150	shrink (SV *sv)
58	{	151	{
59	sv_utf8_downgrade (sv, 1);	152	sv_utf8_downgrade (sv, 1);
		153
60	if (SvLEN (sv) > SvCUR (sv) + 1)	154	if (SvLEN (sv) > SvCUR (sv) + 1)
61	{	155	{
62	#ifdef SvPV_shrink_to_cur	156	#ifdef SvPV_shrink_to_cur
63	SvPV_shrink_to_cur (sv);	157	SvPV_shrink_to_cur (sv);
64	#elif defined (SvPV_renew)	158	#elif defined (SvPV_renew)
65	SvPV_renew (sv, SvCUR (sv) + 1);	159	SvPV_renew (sv, SvCUR (sv) + 1);
66	#endif	160	#endif
67	}	161	}
68	}	162	}
69		163
		164	/* adds two STRLENs together, slow, and with paranoia */
		165	STRLEN
		166	strlen_sum (STRLEN l1, STRLEN l2)
		167	{
		168	size_t sum = l1 + l2;
		169
		170	if (sum < (size_t)l2 || sum != (size_t)(STRLEN)sum)
		171	croak ("JSON::XS: string size overflow");
		172
		173	return sum;
		174	}
		175
		176	/* similar to SvGROW, but somewhat safer and guarantees exponential realloc strategy */
		177	static char *
		178	json_sv_grow (SV *sv, size_t len1, size_t len2)
		179	{
		180	len1 = strlen_sum (len1, len2);
		181	len1 = strlen_sum (len1, len1 >> 1);
		182
		183	if (len1 > 4096 - 24)
		184	len1 = (len1 | 4095) - 24;
		185
		186	return SvGROW (sv, len1);
		187	}
		188
70	// decode an utf-8 character and return it, or (UV)-1 in	189	// decode a utf-8 character and return it, or (UV)-1 in
71	// case of an error.	190	// case of an error.
72	// we special-case "safe" characters from U+80 .. U+7FF,	191	// we special-case "safe" characters from U+80 .. U+7FF,
73	// but use the very good perl function to parse anything else.	192	// but use the very good perl function to parse anything else.
74	// note that we never call this function for a ascii codepoints	193	// note that we never call this function for a ascii codepoints
75	static UV	194	INLINE UV
76	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)	195	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
77	{	196	{
78	if (s[0] > 0xdf || s[0] < 0xc2)	197	if (expect_true (len >= 2
79	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	198	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
80	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	199	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
81	{	200	{
82	*clen = 2;	201	*clen = 2;
83	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);	202	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
84	}	203	}
85	else	204	else
		205	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		206	}
		207
		208	// likewise for encoding, also never called for ascii codepoints
		209	// this function takes advantage of this fact, although current gccs
		210	// seem to optimise the check for >= 0x80 away anyways
		211	INLINE unsigned char *
		212	encode_utf8 (unsigned char *s, UV ch)
		213	{
		214	if (expect_false (ch < 0x000080))
		215	*s++ = ch;
		216	else if (expect_true (ch < 0x000800))
		217	*s++ = 0xc0 | ( ch >> 6),
		218	*s++ = 0x80 | ( ch & 0x3f);
		219	else if ( ch < 0x010000)
		220	*s++ = 0xe0 | ( ch >> 12),
		221	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		222	*s++ = 0x80 | ( ch & 0x3f);
		223	else if ( ch < 0x110000)
		224	*s++ = 0xf0 | ( ch >> 18),
		225	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		226	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		227	*s++ = 0x80 | ( ch & 0x3f);
		228
		229	return s;
		230	}
		231
		232	// convert offset pointer to character index, sv must be string
		233	static STRLEN
		234	ptr_to_index (SV *sv, char *offset)
		235	{
		236	return SvUTF8 (sv)
		237	? utf8_distance (offset, SvPVX (sv))
		238	: offset - SvPVX (sv);
		239	}
		240
		241	/////////////////////////////////////////////////////////////////////////////
		242	// fp hell
		243
		244	// scan a group of digits, and a trailing exponent
		245	static void
		246	json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
		247	{
		248	UV uaccum = 0;
		249	int eaccum = 0;
		250
		251	// if we recurse too deep, skip all remaining digits
		252	// to avoid a stack overflow attack
		253	if (expect_false (--maxdepth <= 0))
		254	while (((U8)*s - '0') < 10)
		255	++s;
		256
		257	for (;;)
		258	{
		259	U8 dig = (U8)*s - '0';
		260
		261	if (expect_false (dig >= 10))
		262	{
		263	if (dig == (U8)((U8)'.' - (U8)'0'))
		264	{
		265	++s;
		266	json_atof_scan1 (s, accum, expo, 1, maxdepth);
		267	}
		268	else if ((dig | ' ') == 'e' - '0')
		269	{
		270	int exp2 = 0;
		271	int neg = 0;
		272
		273	++s;
		274
		275	if (*s == '-')
		276	{
		277	++s;
		278	neg = 1;
		279	}
		280	else if (*s == '+')
		281	++s;
		282
		283	while ((dig = (U8)*s - '0') < 10)
		284	exp2 = exp2 * 10 + *s++ - '0';
		285
		286	*expo += neg ? -exp2 : exp2;
		287	}
		288
		289	break;
		290	}
		291
		292	++s;
		293
		294	uaccum = uaccum * 10 + dig;
		295	++eaccum;
		296
		297	// if we have too many digits, then recurse for more
		298	// we actually do this for rather few digits
		299	if (uaccum >= (UV_MAX - 9) / 10)
		300	{
		301	if (postdp) *expo -= eaccum;
		302	json_atof_scan1 (s, accum, expo, postdp, maxdepth);
		303	if (postdp) *expo += eaccum;
		304
		305	break;
		306	}
86	{	307	}
87	*clen = (STRLEN)-1;	308
		309	// this relies greatly on the quality of the pow ()
		310	// implementation of the platform, but a good
		311	// implementation is hard to beat.
		312	// (IEEE 754 conformant ones are required to be exact)
		313	if (postdp) *expo -= eaccum;
		314	*accum += uaccum * Perl_pow (10., *expo);
		315	*expo += eaccum;
		316	}
		317
		318	static NV
		319	json_atof (const char *s)
		320	{
		321	NV accum = 0.;
		322	int expo = 0;
		323	int neg = 0;
		324
		325	if (*s == '-')
		326	{
		327	++s;
		328	neg = 1;
		329	}
		330
		331	// a recursion depth of ten gives us >>500 bits
		332	json_atof_scan1 (s, &accum, &expo, 0, 10);
		333
		334	return neg ? -accum : accum;
		335	}
		336
		337	// target of scalar reference is bool? -1 == nope, 0 == false, 1 == true
		338	static int
		339	ref_bool_type (SV *sv)
		340	{
		341	svtype svt = SvTYPE (sv);
		342
		343	if (svt < SVt_PVAV)
		344	{
		345	STRLEN len = 0;
		346	char *pv = svt ? SvPV (sv, len) : 0;
		347
		348	if (len == 1)
		349	if (*pv == '1')
		350	return 1;
		351	else if (*pv == '0')
		352	return 0;
		353	}
		354
		355	return -1;
		356	}
		357
		358	// returns whether scalar is not a reference in the sense of allow_nonref
		359	static int
		360	json_nonref (SV *scalar)
		361	{
		362	if (!SvROK (scalar))
		363	return 1;
		364
		365	scalar = SvRV (scalar);
		366
		367	if (SvTYPE (scalar) >= SVt_PVMG)
		368	{
		369	if (SvSTASH (scalar) == bool_stash)
88	return (UV)-1;	370	return 1;
		371
		372	if (!SvOBJECT (scalar) && ref_bool_type (scalar) >= 0)
		373	return 1;
89	}	374	}
		375
		376	return 0;
90	}	377	}
91		378
92	/////////////////////////////////////////////////////////////////////////////	379	/////////////////////////////////////////////////////////////////////////////
93	// encoder	380	// encoder
94		381
…		…
96	typedef struct	383	typedef struct
97	{	384	{
98	char *cur; // SvPVX (sv) + current output position	385	char *cur; // SvPVX (sv) + current output position
99	char *end; // SvEND (sv)	386	char *end; // SvEND (sv)
100	SV *sv; // result scalar	387	SV *sv; // result scalar
101	U32 flags; // F_*	388	JSON json;
102	U32 indent; // indentation level	389	U32 indent; // indentation level
103	U32 maxdepth; // max. indentation/recursion level	390	UV limit; // escape character values >= this value when encoding
104	} enc_t;	391	} enc_t;
105		392
106	static void	393	INLINE void
107	need (enc_t *enc, STRLEN len)	394	need (enc_t *enc, STRLEN len)
108	{	395	{
109	if (enc->cur + len >= enc->end)	396	if (expect_false ((uintptr_t)(enc->end - enc->cur) < len))
110	{	397	{
111	STRLEN cur = enc->cur - SvPVX (enc->sv);	398	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
112	SvGROW (enc->sv, cur + len + 1);	399	char *buf = json_sv_grow (enc->sv, cur, len);
113	enc->cur = SvPVX (enc->sv) + cur;	400	enc->cur = buf + cur;
114	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	401	enc->end = buf + SvLEN (enc->sv) - 1;
115	}	402	}
116	}	403	}
117		404
118	static void	405	INLINE void
119	encode_ch (enc_t *enc, char ch)	406	encode_ch (enc_t *enc, char ch)
120	{	407	{
121	need (enc, 1);	408	need (enc, 1);
122	*enc->cur++ = ch;	409	*enc->cur++ = ch;
123	}	410	}
…		…
131		418
132	while (str < end)	419	while (str < end)
133	{	420	{
134	unsigned char ch = *(unsigned char *)str;	421	unsigned char ch = *(unsigned char *)str;
135		422
136	if (ch >= 0x20 && ch < 0x80) // most common case	423	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
137	{	424	{
138	if (ch == '"') // but with slow exceptions	425	if (expect_false (ch == '"')) // but with slow exceptions
139	{	426	{
140	need (enc, len += 1);	427	need (enc, len + 1);
141	*enc->cur++ = '\\';	428	*enc->cur++ = '\\';
142	*enc->cur++ = '"';	429	*enc->cur++ = '"';
143	}	430	}
144	else if (ch == '\\')	431	else if (expect_false (ch == '\\'))
145	{	432	{
146	need (enc, len += 1);	433	need (enc, len + 1);
147	*enc->cur++ = '\\';	434	*enc->cur++ = '\\';
148	*enc->cur++ = '\\';	435	*enc->cur++ = '\\';
149	}	436	}
150	else	437	else
151	*enc->cur++ = ch;	438	*enc->cur++ = ch;
…		…
154	}	441	}
155	else	442	else
156	{	443	{
157	switch (ch)	444	switch (ch)
158	{	445	{
159	case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;	446	case '\010': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
160	case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;	447	case '\011': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
161	case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;	448	case '\012': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
162	case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;	449	case '\014': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
163	case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;	450	case '\015': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
164		451
165	default:	452	default:
166	{	453	{
167	STRLEN clen;	454	STRLEN clen;
168	UV uch;	455	UV uch;
169		456
170	if (is_utf8)	457	if (is_utf8)
171	{	458	{
172	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
173	uch = decode_utf8 (str, end - str, &clen);	459	uch = decode_utf8 (str, end - str, &clen);
174	if (clen == (STRLEN)-1)	460	if (clen == (STRLEN)-1)
175	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	461	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
176	}	462	}
177	else	463	else
178	{	464	{
179	uch = ch;	465	uch = ch;
180	clen = 1;	466	clen = 1;
181	}	467	}
182		468
183	if (uch > 0x10FFFFUL)	469	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
184	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
185
186	if (uch < 0x80 || enc->flags & F_ASCII || (enc->flags & F_LATIN1 && uch > 0xFF))
187	{	470	{
188	if (uch > 0xFFFFUL)	471	if (uch >= 0x10000UL)
189	{	472	{
		473	if (uch >= 0x110000UL)
		474	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		475
190	need (enc, len += 11);	476	need (enc, len + 11);
191	sprintf (enc->cur, "\\u%04x\\u%04x",	477	sprintf (enc->cur, "\\u%04x\\u%04x",
192	(int)((uch - 0x10000) / 0x400 + 0xD800),	478	(int)((uch - 0x10000) / 0x400 + 0xD800),
193	(int)((uch - 0x10000) % 0x400 + 0xDC00));	479	(int)((uch - 0x10000) % 0x400 + 0xDC00));
194	enc->cur += 12;	480	enc->cur += 12;
195	}	481	}
196	else	482	else
197	{	483	{
198	static char hexdigit [16] = "0123456789abcdef";
199	need (enc, len += 5);	484	need (enc, len + 5);
200	*enc->cur++ = '\\';	485	*enc->cur++ = '\\';
201	*enc->cur++ = 'u';	486	*enc->cur++ = 'u';
202	*enc->cur++ = hexdigit [ uch >> 12 ];	487	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
203	*enc->cur++ = hexdigit [(uch >> 8) & 15];	488	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
204	*enc->cur++ = hexdigit [(uch >> 4) & 15];	489	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
205	*enc->cur++ = hexdigit [(uch >> 0) & 15];	490	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
206	}	491	}
207		492
208	str += clen;	493	str += clen;
209	}	494	}
210	else if (enc->flags & F_LATIN1)	495	else if (enc->json.flags & F_LATIN1)
211	{	496	{
212	*enc->cur++ = uch;	497	*enc->cur++ = uch;
213	str += clen;	498	str += clen;
214	}	499	}
215	else if (is_utf8)	500	else if (is_utf8)
216	{	501	{
217	need (enc, len += clen);	502	need (enc, len + clen);
218	do	503	do
219	{	504	{
220	*enc->cur++ = *str++;	505	*enc->cur++ = *str++;
221	}	506	}
222	while (--clen);	507	while (--clen);
223	}	508	}
224	else	509	else
225	{	510	{
226	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	511	need (enc, len + UTF8_MAXBYTES - 1); // never more than 11 bytes needed
227	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	512	enc->cur = encode_utf8 (enc->cur, uch);
228	++str;	513	++str;
229	}	514	}
230	}	515	}
231	}	516	}
232	}	517	}
233		518
234	--len;	519	--len;
235	}	520	}
236	}	521	}
237		522
238	static void	523	INLINE void
239	encode_indent (enc_t *enc)	524	encode_indent (enc_t *enc)
240	{	525	{
241	if (enc->flags & F_INDENT)	526	if (enc->json.flags & F_INDENT)
242	{	527	{
243	int spaces = enc->indent * INDENT_STEP;	528	int spaces = enc->indent * INDENT_STEP;
244		529
245	need (enc, spaces);	530	need (enc, spaces);
246	memset (enc->cur, ' ', spaces);	531	memset (enc->cur, ' ', spaces);
247	enc->cur += spaces;	532	enc->cur += spaces;
248	}	533	}
249	}	534	}
250		535
251	static void	536	INLINE void
252	encode_space (enc_t *enc)	537	encode_space (enc_t *enc)
253	{	538	{
254	need (enc, 1);	539	need (enc, 1);
255	encode_ch (enc, ' ');	540	encode_ch (enc, ' ');
256	}	541	}
257		542
258	static void	543	INLINE void
259	encode_nl (enc_t *enc)	544	encode_nl (enc_t *enc)
260	{	545	{
261	if (enc->flags & F_INDENT)	546	if (enc->json.flags & F_INDENT)
262	{	547	{
263	need (enc, 1);	548	need (enc, 1);
264	encode_ch (enc, '\n');	549	encode_ch (enc, '\n');
265	}	550	}
266	}	551	}
267		552
268	static void	553	INLINE void
269	encode_comma (enc_t *enc)	554	encode_comma (enc_t *enc)
270	{	555	{
271	encode_ch (enc, ',');	556	encode_ch (enc, ',');
272		557
273	if (enc->flags & F_INDENT)	558	if (enc->json.flags & F_INDENT)
274	encode_nl (enc);	559	encode_nl (enc);
275	else if (enc->flags & F_SPACE_AFTER)	560	else if (enc->json.flags & F_SPACE_AFTER)
276	encode_space (enc);	561	encode_space (enc);
277	}	562	}
278		563
279	static void encode_sv (enc_t *enc, SV *sv);	564	static void encode_sv (enc_t *enc, SV *sv);
280		565
281	static void	566	static void
282	encode_av (enc_t *enc, AV *av)	567	encode_av (enc_t *enc, AV *av)
283	{	568	{
284	int i, len = av_len (av);	569	int i, len = av_len (av);
285		570
286	if (enc->indent >= enc->maxdepth)	571	if (enc->indent >= enc->json.max_depth)
287	croak ("data structure too deep (hit recursion limit)");	572	croak (ERR_NESTING_EXCEEDED);
288		573
289	encode_ch (enc, '['); encode_nl (enc);	574	encode_ch (enc, '[');
290	++enc->indent;
291		575
		576	if (len >= 0)
		577	{
		578	encode_nl (enc); ++enc->indent;
		579
292	for (i = 0; i <= len; ++i)	580	for (i = 0; i <= len; ++i)
293	{	581	{
		582	SV **svp = av_fetch (av, i, 0);
		583
294	encode_indent (enc);	584	encode_indent (enc);
295	encode_sv (enc, *av_fetch (av, i, 0));
296		585
		586	if (svp)
		587	encode_sv (enc, *svp);
		588	else
		589	encode_str (enc, "null", 4, 0);
		590
297	if (i < len)	591	if (i < len)
298	encode_comma (enc);	592	encode_comma (enc);
299	}	593	}
300		594
		595	encode_nl (enc); --enc->indent; encode_indent (enc);
		596	}
		597
301	encode_nl (enc);	598	encode_ch (enc, ']');
302
303	--enc->indent;
304	encode_indent (enc); encode_ch (enc, ']');
305	}	599	}
306		600
307	static void	601	static void
308	encode_he (enc_t *enc, HE *he)	602	encode_hk (enc_t *enc, HE *he)
309	{	603	{
310	encode_ch (enc, '"');	604	encode_ch (enc, '"');
311		605
312	if (HeKLEN (he) == HEf_SVKEY)	606	if (HeKLEN (he) == HEf_SVKEY)
313	{	607	{
314	SV *sv = HeSVKEY (he);	608	SV *sv = HeSVKEY (he);
315	STRLEN len;	609	STRLEN len;
316	char *str;	610	char *str;
317		611
318	SvGETMAGIC (sv);	612	SvGETMAGIC (sv);
319	str = SvPV (sv, len);	613	str = SvPV (sv, len);
320		614
321	encode_str (enc, str, len, SvUTF8 (sv));	615	encode_str (enc, str, len, SvUTF8 (sv));
322	}	616	}
323	else	617	else
324	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	618	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
325		619
326	encode_ch (enc, '"');	620	encode_ch (enc, '"');
327		621
328	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	622	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
329	encode_ch (enc, ':');	623	encode_ch (enc, ':');
330	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	624	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
331	encode_sv (enc, HeVAL (he));
332	}	625	}
333		626
334	// compare hash entries, used when all keys are bytestrings	627	// compare hash entries, used when all keys are bytestrings
335	static int	628	static int
336	he_cmp_fast (const void *a_, const void *b_)	629	he_cmp_fast (const void *a_, const void *b_)
…		…
341	HE *b = *(HE **)b_;	634	HE *b = *(HE **)b_;
342		635
343	STRLEN la = HeKLEN (a);	636	STRLEN la = HeKLEN (a);
344	STRLEN lb = HeKLEN (b);	637	STRLEN lb = HeKLEN (b);
345		638
346	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	639	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
347	cmp = la - lb;	640	cmp = lb - la;
348		641
349	return cmp;	642	return cmp;
350	}	643	}
351		644
352	// compare hash entries, used when some keys are sv's or utf-x	645	// compare hash entries, used when some keys are sv's or utf-x
353	static int	646	static int
354	he_cmp_slow (const void *a, const void *b)	647	he_cmp_slow (const void *a, const void *b)
355	{	648	{
356	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	649	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
357	}	650	}
358		651
359	static void	652	static void
360	encode_hv (enc_t *enc, HV *hv)	653	encode_hv (enc_t *enc, HV *hv)
361	{	654	{
362	int count, i;	655	HE *he;
363		656
364	if (enc->indent >= enc->maxdepth)	657	if (enc->indent >= enc->json.max_depth)
365	croak ("data structure too deep (hit recursion limit)");	658	croak (ERR_NESTING_EXCEEDED);
366		659
367	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	660	encode_ch (enc, '{');
368		661
369	if ((count = hv_iterinit (hv)))
370	{
371	// for canonical output we have to sort by keys first	662	// for canonical output we have to sort by keys first
372	// actually, this is mostly due to the stupid so-called	663	// actually, this is mostly due to the stupid so-called
373	// security workaround added somewhere in 5.8.x.	664	// security workaround added somewhere in 5.8.x
374	// that randomises hash orderings	665	// that randomises hash orderings
375	if (enc->flags & F_CANONICAL)	666	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
		667	{
		668	int count = hv_iterinit (hv);
		669
		670	if (SvMAGICAL (hv))
376	{	671	{
377	HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode	672	// need to count by iterating. could improve by dynamically building the vector below
		673	// but I don't care for the speed of this special case.
		674	// note also that we will run into undefined behaviour when the two iterations
		675	// do not result in the same count, something I might care for in some later release.
		676
		677	count = 0;
		678	while (hv_iternext (hv))
		679	++count;
		680
		681	hv_iterinit (hv);
		682	}
		683
		684	if (count)
		685	{
378	int fast = 1;	686	int i, fast = 1;
		687	HE *hes_stack [STACK_HES];
		688	HE **hes = hes_stack;
		689
		690	// allocate larger arrays on the heap
		691	if (count > STACK_HES)
		692	{
		693	SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
		694	hes = (HE **)SvPVX (sv);
		695	}
379		696
380	i = 0;	697	i = 0;
381	while ((he = hv_iternext (hv)))	698	while ((he = hv_iternext (hv)))
382	{	699	{
383	hes [i++] = he;	700	hes [i++] = he;
…		…
405		722
406	FREETMPS;	723	FREETMPS;
407	LEAVE;	724	LEAVE;
408	}	725	}
409		726
410	for (i = 0; i < count; ++i)	727	encode_nl (enc); ++enc->indent;
		728
		729	while (count--)
411	{	730	{
412	encode_indent (enc);	731	encode_indent (enc);
		732	he = hes [count];
413	encode_he (enc, hes [i]);	733	encode_hk (enc, he);
		734	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
414		735
415	if (i < count - 1)	736	if (count)
416	encode_comma (enc);	737	encode_comma (enc);
417	}	738	}
418		739
419	encode_nl (enc);	740	encode_nl (enc); --enc->indent; encode_indent (enc);
420	}	741	}
		742	}
421	else	743	else
		744	{
		745	if (hv_iterinit (hv) || SvMAGICAL (hv))
		746	if ((he = hv_iternext (hv)))
422	{	747	{
423	HE *he = hv_iternext (hv);	748	encode_nl (enc); ++enc->indent;
424		749
425	for (;;)	750	for (;;)
426	{	751	{
427	encode_indent (enc);	752	encode_indent (enc);
428	encode_he (enc, he);	753	encode_hk (enc, he);
		754	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
429		755
430	if (!(he = hv_iternext (hv)))	756	if (!(he = hv_iternext (hv)))
431	break;	757	break;
432		758
433	encode_comma (enc);	759	encode_comma (enc);
434	}	760	}
435		761
436	encode_nl (enc);	762	encode_nl (enc); --enc->indent; encode_indent (enc);
437	}	763	}
438	}	764	}
439		765
440	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	766	encode_ch (enc, '}');
441	}	767	}
442		768
443	// encode objects, arrays and special \0=false and \1=true values.	769	// encode objects, arrays and special \0=false and \1=true values.
444	static void	770	static void
445	encode_rv (enc_t *enc, SV *sv)	771	encode_rv (enc_t *enc, SV *sv)
446	{	772	{
447	svtype svt;	773	svtype svt;
		774	GV *method;
448		775
449	SvGETMAGIC (sv);	776	SvGETMAGIC (sv);
450	svt = SvTYPE (sv);	777	svt = SvTYPE (sv);
451		778
		779	if (expect_false (SvOBJECT (sv)))
		780	{
		781	HV *stash = SvSTASH (sv);
		782
		783	if (stash == bool_stash)
		784	{
		785	if (SvIV (sv)) encode_str (enc, "true" , 4, 0);
		786	else encode_str (enc, "false", 5, 0);
		787	}
		788	else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
		789	{
		790	int count;
		791	dSP;
		792
		793	ENTER; SAVETMPS;
		794	PUSHMARK (SP);
		795	EXTEND (SP, 2);
		796	// we re-bless the reference to get overload and other niceties right
		797	PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		798	PUSHs (sv_json);
		799
		800	PUTBACK;
		801	count = call_sv ((SV *)GvCV (method), G_ARRAY);
		802	SPAGAIN;
		803
		804	// catch this surprisingly common error
		805	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		806	croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		807
		808	encode_ch (enc, '(');
		809	encode_ch (enc, '"');
		810	encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
		811	encode_ch (enc, '"');
		812	encode_ch (enc, ')');
		813	encode_ch (enc, '[');
		814
		815	if (count)
		816	{
		817	int i;
		818
		819	for (i = 0; i < count - 1; ++i)
		820	{
		821	encode_sv (enc, SP[i + 1 - count]);
		822	encode_ch (enc, ',');
		823	}
		824
		825	encode_sv (enc, TOPs);
		826	SP -= count;
		827	}
		828
		829	encode_ch (enc, ']');
		830
		831	FREETMPS; LEAVE;
		832	}
		833	else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
		834	{
		835	dSP;
		836
		837	ENTER; SAVETMPS;
		838	PUSHMARK (SP);
		839	// we re-bless the reference to get overload and other niceties right
		840	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		841
		842	// calling with G_SCALAR ensures that we always get a 1 return value
		843	PUTBACK;
		844	call_sv ((SV *)GvCV (method), G_SCALAR);
		845	SPAGAIN;
		846
		847	// catch this surprisingly common error
		848	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		849	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		850
		851	sv = POPs;
		852	PUTBACK;
		853
		854	encode_sv (enc, sv);
		855
		856	FREETMPS; LEAVE;
		857	}
		858	else if (enc->json.flags & F_ALLOW_BLESSED)
		859	encode_str (enc, "null", 4, 0);
		860	else
		861	croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
		862	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		863	}
452	if (svt == SVt_PVHV)	864	else if (svt == SVt_PVHV)
453	encode_hv (enc, (HV *)sv);	865	encode_hv (enc, (HV *)sv);
454	else if (svt == SVt_PVAV)	866	else if (svt == SVt_PVAV)
455	encode_av (enc, (AV *)sv);	867	encode_av (enc, (AV *)sv);
456	else if (svt < SVt_PVAV)	868	else if (svt < SVt_PVAV)
457	{	869	{
458	if (SvNIOK (sv) && SvIV (sv) == 0)	870	int bool_type = ref_bool_type (sv);
		871
		872	if (bool_type == 1)
		873	encode_str (enc, "true", 4, 0);
		874	else if (bool_type == 0)
459	encode_str (enc, "false", 5, 0);	875	encode_str (enc, "false", 5, 0);
460	else if (SvNIOK (sv) && SvIV (sv) == 1)	876	else if (enc->json.flags & F_ALLOW_UNKNOWN)
461	encode_str (enc, "true", 4, 0);	877	encode_str (enc, "null", 4, 0);
462	else	878	else
463	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	879	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
464	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	880	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
465	}	881	}
		882	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		883	encode_str (enc, "null", 4, 0);
466	else	884	else
467	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	885	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
468	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	886	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
469	}	887	}
470		888
…		…
481	encode_str (enc, str, len, SvUTF8 (sv));	899	encode_str (enc, str, len, SvUTF8 (sv));
482	encode_ch (enc, '"');	900	encode_ch (enc, '"');
483	}	901	}
484	else if (SvNOKp (sv))	902	else if (SvNOKp (sv))
485	{	903	{
		904	// trust that perl will do the right thing w.r.t. JSON syntax.
486	need (enc, NV_DIG + 32);	905	need (enc, NV_DIG + 32);
487	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	906	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
488	enc->cur += strlen (enc->cur);	907	enc->cur += strlen (enc->cur);
489	}	908	}
490	else if (SvIOKp (sv))	909	else if (SvIOKp (sv))
491	{	910	{
		911	// we assume we can always read an IV as a UV and vice versa
		912	// we assume two's complement
		913	// we assume no aliasing issues in the union
		914	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		915	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		916	{
		917	// optimise the "small number case"
		918	// code will likely be branchless and use only a single multiplication
		919	// works for numbers up to 59074
		920	I32 i = SvIVX (sv);
		921	U32 u;
		922	char digit, nz = 0;
		923
492	need (enc, 64);	924	need (enc, 6);
		925
		926	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		927	u = i < 0 ? -i : i;
		928
		929	// convert to 4.28 fixed-point representation
		930	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		931
		932	// now output digit by digit, each time masking out the integer part
		933	// and multiplying by 5 while moving the decimal point one to the right,
		934	// resulting in a net multiplication by 10.
		935	// we always write the digit to memory but conditionally increment
		936	// the pointer, to enable the use of conditional move instructions.
		937	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		938	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		939	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		940	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		941	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		942	}
		943	else
		944	{
		945	// large integer, use the (rather slow) snprintf way.
		946	need (enc, IVUV_MAXCHARS);
493	enc->cur +=	947	enc->cur +=
494	SvIsUV(sv)	948	SvIsUV(sv)
495	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	949	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
496	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	950	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		951	}
497	}	952	}
498	else if (SvROK (sv))	953	else if (SvROK (sv))
499	encode_rv (enc, SvRV (sv));	954	encode_rv (enc, SvRV (sv));
500	else if (!SvOK (sv))	955	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
501	encode_str (enc, "null", 4, 0);	956	encode_str (enc, "null", 4, 0);
502	else	957	else
503	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	958	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
504	SvPV_nolen (sv), SvFLAGS (sv));	959	SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
505	}	960	}
506		961
507	static SV *	962	static SV *
508	encode_json (SV *scalar, U32 flags)	963	encode_json (SV *scalar, JSON *json)
509	{	964	{
510	enc_t enc;	965	enc_t enc;
511		966
512	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	967	if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar))
513	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	968	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
514		969
515	enc.flags = flags;	970	enc.json = *json;
516	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	971	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
517	enc.cur = SvPVX (enc.sv);	972	enc.cur = SvPVX (enc.sv);
518	enc.end = SvEND (enc.sv);	973	enc.end = SvEND (enc.sv);
519	enc.indent = 0;	974	enc.indent = 0;
520	enc.maxdepth = DEC_DEPTH (flags);	975	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		976	: enc.json.flags & F_LATIN1 ? 0x000100UL
		977	: 0x110000UL;
521		978
522	SvPOK_only (enc.sv);	979	SvPOK_only (enc.sv);
523	encode_sv (&enc, scalar);	980	encode_sv (&enc, scalar);
		981	encode_nl (&enc);
524		982
525	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	983	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
526	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	984	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
527		985
528	if (!(flags & (F_ASCII | F_LATIN1 | F_UTF8)))	986	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
529	SvUTF8_on (enc.sv);	987	SvUTF8_on (enc.sv);
530		988
531	if (enc.flags & F_SHRINK)	989	if (enc.json.flags & F_SHRINK)
532	shrink (enc.sv);	990	shrink (enc.sv);
533		991
534	return enc.sv;	992	return enc.sv;
535	}	993	}
536		994
…		…
541	typedef struct	999	typedef struct
542	{	1000	{
543	char *cur; // current parser pointer	1001	char *cur; // current parser pointer
544	char *end; // end of input string	1002	char *end; // end of input string
545	const char *err; // parse error, if != 0	1003	const char *err; // parse error, if != 0
546	U32 flags; // F_*	1004	JSON json;
547	U32 depth; // recursion depth	1005	U32 depth; // recursion depth
548	U32 maxdepth; // recursion depth limit	1006	U32 maxdepth; // recursion depth limit
549	} dec_t;	1007	} dec_t;
550		1008
551	static void	1009	INLINE void
		1010	decode_comment (dec_t *dec)
		1011	{
		1012	// only '#'-style comments allowed a.t.m.
		1013
		1014	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		1015	++dec->cur;
		1016	}
		1017
		1018	INLINE void
552	decode_ws (dec_t *dec)	1019	decode_ws (dec_t *dec)
553	{	1020	{
554	for (;;)	1021	for (;;)
555	{	1022	{
556	char ch = *dec->cur;	1023	char ch = *dec->cur;
557		1024
558	if (ch > 0x20	1025	if (ch > 0x20)
		1026	{
		1027	if (expect_false (ch == '#'))
		1028	{
		1029	if (dec->json.flags & F_RELAXED)
		1030	decode_comment (dec);
		1031	else
		1032	break;
		1033	}
		1034	else
		1035	break;
		1036	}
559	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	1037	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
560	break;	1038	break; // parse error, but let higher level handle it, gives better error messages
561		1039
562	++dec->cur;	1040	++dec->cur;
563	}	1041	}
564	}	1042	}
565		1043
…		…
569	if (*dec->cur != ch) \	1047	if (*dec->cur != ch) \
570	ERR (# ch " expected"); \	1048	ERR (# ch " expected"); \
571	++dec->cur; \	1049	++dec->cur; \
572	SE	1050	SE
573		1051
574	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	1052	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
575	#define DEC_DEC_DEPTH --dec->depth	1053	#define DEC_DEC_DEPTH --dec->depth
576		1054
577	static SV *decode_sv (dec_t *dec);	1055	static SV *decode_sv (dec_t *dec);
578		1056
579	static signed char decode_hexdigit[256];	1057	static signed char decode_hexdigit[256];
…		…
582	decode_4hex (dec_t *dec)	1060	decode_4hex (dec_t *dec)
583	{	1061	{
584	signed char d1, d2, d3, d4;	1062	signed char d1, d2, d3, d4;
585	unsigned char *cur = (unsigned char *)dec->cur;	1063	unsigned char *cur = (unsigned char *)dec->cur;
586		1064
587	d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected");	1065	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
588	d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected");	1066	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
589	d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected");	1067	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
590	d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected");	1068	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
591		1069
592	dec->cur += 4;	1070	dec->cur += 4;
593		1071
594	return ((UV)d1) << 12	1072	return ((UV)d1) << 12
595	| ((UV)d2) << 8	1073	| ((UV)d2) << 8
…		…
603	static SV *	1081	static SV *
604	decode_str (dec_t *dec)	1082	decode_str (dec_t *dec)
605	{	1083	{
606	SV *sv = 0;	1084	SV *sv = 0;
607	int utf8 = 0;	1085	int utf8 = 0;
		1086	char *dec_cur = dec->cur;
608		1087
609	do	1088	do
610	{	1089	{
611	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];	1090	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
612	char *cur = buf;	1091	char *cur = buf;
613		1092
614	do	1093	do
615	{	1094	{
616	unsigned char ch = *(unsigned char *)dec->cur++;	1095	unsigned char ch = *(unsigned char *)dec_cur++;
617		1096
618	if (ch == '"')	1097	if (expect_false (ch == '"'))
619	{	1098	{
620	--dec->cur;	1099	--dec_cur;
621	break;	1100	break;
622	}	1101	}
623	else if (ch == '\\')	1102	else if (expect_false (ch == '\\'))
624	{	1103	{
625	switch (*dec->cur)	1104	switch (*dec_cur)
626	{	1105	{
627	case '\\':	1106	case '\\':
628	case '/':	1107	case '/':
629	case '"': *cur++ = *dec->cur++; break;	1108	case '"': *cur++ = *dec_cur++; break;
630		1109
631	case 'b': ++dec->cur; *cur++ = '\010'; break;	1110	case 'b': ++dec_cur; *cur++ = '\010'; break;
632	case 't': ++dec->cur; *cur++ = '\011'; break;	1111	case 't': ++dec_cur; *cur++ = '\011'; break;
633	case 'n': ++dec->cur; *cur++ = '\012'; break;	1112	case 'n': ++dec_cur; *cur++ = '\012'; break;
634	case 'f': ++dec->cur; *cur++ = '\014'; break;	1113	case 'f': ++dec_cur; *cur++ = '\014'; break;
635	case 'r': ++dec->cur; *cur++ = '\015'; break;	1114	case 'r': ++dec_cur; *cur++ = '\015'; break;
636		1115
637	case 'u':	1116	case 'u':
638	{	1117	{
639	UV lo, hi;	1118	UV lo, hi;
640	++dec->cur;	1119	++dec_cur;
641		1120
		1121	dec->cur = dec_cur;
642	hi = decode_4hex (dec);	1122	hi = decode_4hex (dec);
		1123	dec_cur = dec->cur;
643	if (hi == (UV)-1)	1124	if (hi == (UV)-1)
644	goto fail;	1125	goto fail;
645		1126
646	// possibly a surrogate pair	1127	// possibly a surrogate pair
647	if (hi >= 0xd800)	1128	if (hi >= 0xd800)
648	if (hi < 0xdc00)	1129	if (hi < 0xdc00)
649	{	1130	{
650	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	1131	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
651	ERR ("missing low surrogate character in surrogate pair");	1132	ERR ("missing low surrogate character in surrogate pair");
652		1133
653	dec->cur += 2;	1134	dec_cur += 2;
654		1135
		1136	dec->cur = dec_cur;
655	lo = decode_4hex (dec);	1137	lo = decode_4hex (dec);
		1138	dec_cur = dec->cur;
656	if (lo == (UV)-1)	1139	if (lo == (UV)-1)
657	goto fail;	1140	goto fail;
658		1141
659	if (lo < 0xdc00 || lo >= 0xe000)	1142	if (lo < 0xdc00 || lo >= 0xe000)
660	ERR ("surrogate pair expected");	1143	ERR ("surrogate pair expected");
…		…
666		1149
667	if (hi >= 0x80)	1150	if (hi >= 0x80)
668	{	1151	{
669	utf8 = 1;	1152	utf8 = 1;
670		1153
671	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	1154	cur = encode_utf8 (cur, hi);
672	}	1155	}
673	else	1156	else
674	*cur++ = hi;	1157	*cur++ = hi;
675	}	1158	}
676	break;	1159	break;
677		1160
678	default:	1161	default:
679	--dec->cur;	1162	--dec_cur;
680	ERR ("illegal backslash escape sequence in string");	1163	ERR ("illegal backslash escape sequence in string");
681	}	1164	}
682	}	1165	}
683	else if (ch >= 0x20 && ch <= 0x7f)	1166	else if (expect_true (ch >= 0x20 && ch < 0x80))
684	*cur++ = ch;	1167	*cur++ = ch;
685	else if (ch >= 0x80)	1168	else if (ch >= 0x80)
686	{	1169	{
687	STRLEN clen;	1170	STRLEN clen;
688	UV uch;
689		1171
690	--dec->cur;	1172	--dec_cur;
691		1173
692	uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	1174	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
693	if (clen == (STRLEN)-1)	1175	if (clen == (STRLEN)-1)
694	ERR ("malformed UTF-8 character in JSON string");	1176	ERR ("malformed UTF-8 character in JSON string");
695		1177
696	do	1178	do
697	*cur++ = *dec->cur++;	1179	*cur++ = *dec_cur++;
698	while (--clen);	1180	while (--clen);
699		1181
700	utf8 = 1;	1182	utf8 = 1;
701	}	1183	}
		1184	else if (ch == '\t' && dec->json.flags & F_RELAXED)
		1185	*cur++ = ch;
702	else	1186	else
703	{	1187	{
704	--dec->cur;	1188	--dec_cur;
705		1189
706	if (!ch)	1190	if (!ch)
707	ERR ("unexpected end of string while parsing JSON string");	1191	ERR ("unexpected end of string while parsing JSON string");
708	else	1192	else
709	ERR ("invalid character encountered while parsing JSON string");	1193	ERR ("invalid character encountered while parsing JSON string");
…		…
714	{	1198	{
715	STRLEN len = cur - buf;	1199	STRLEN len = cur - buf;
716		1200
717	if (sv)	1201	if (sv)
718	{	1202	{
719	SvGROW (sv, SvCUR (sv) + len + 1);	1203	STRLEN cur = SvCUR (sv);
		1204
		1205	if (SvLEN (sv) - cur <= len)
		1206	json_sv_grow (sv, cur, len);
		1207
720	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	1208	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
721	SvCUR_set (sv, SvCUR (sv) + len);	1209	SvCUR_set (sv, SvCUR (sv) + len);
722	}	1210	}
723	else	1211	else
724	sv = newSVpvn (buf, len);	1212	sv = newSVpvn (buf, len);
725	}	1213	}
726	}	1214	}
727	while (*dec->cur != '"');	1215	while (*dec_cur != '"');
728		1216
729	++dec->cur;	1217	++dec_cur;
730		1218
731	if (sv)	1219	if (sv)
732	{	1220	{
733	SvPOK_only (sv);	1221	SvPOK_only (sv);
734	*SvEND (sv) = 0;	1222	*SvEND (sv) = 0;
…		…
737	SvUTF8_on (sv);	1225	SvUTF8_on (sv);
738	}	1226	}
739	else	1227	else
740	sv = newSVpvn ("", 0);	1228	sv = newSVpvn ("", 0);
741		1229
		1230	dec->cur = dec_cur;
742	return sv;	1231	return sv;
743		1232
744	fail:	1233	fail:
		1234	dec->cur = dec_cur;
745	return 0;	1235	return 0;
746	}	1236	}
747		1237
748	static SV *	1238	static SV *
749	decode_num (dec_t *dec)	1239	decode_num (dec_t *dec)
…		…
807	is_nv = 1;	1297	is_nv = 1;
808	}	1298	}
809		1299
810	if (!is_nv)	1300	if (!is_nv)
811	{	1301	{
812	UV uv;	1302	int len = dec->cur - start;
813	int numtype = grok_number (start, dec->cur - start, &uv);	1303
814	if (numtype & IS_NUMBER_IN_UV)	1304	// special case the rather common 1..5-digit-int case
815	if (numtype & IS_NUMBER_NEG)	1305	if (*start == '-')
		1306	switch (len)
816	{	1307	{
817	if (uv < (UV)IV_MIN)	1308	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
818	return newSViv (-(IV)uv);	1309	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
		1310	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1311	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1312	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
819	}	1313	}
		1314	else
		1315	switch (len)
		1316	{
		1317	case 1: return newSViv ( start [0] - '0' * 1);
		1318	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1319	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1320	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1321	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1322	}
		1323
		1324	{
		1325	UV uv;
		1326	int numtype = grok_number (start, len, &uv);
		1327	if (numtype & IS_NUMBER_IN_UV)
		1328	if (numtype & IS_NUMBER_NEG)
		1329	{
		1330	if (uv < (UV)IV_MIN)
		1331	return newSViv (-(IV)uv);
		1332	}
820	else	1333	else
821	return newSVuv (uv);	1334	return newSVuv (uv);
822	}	1335	}
823		1336
		1337	len -= *start == '-' ? 1 : 0;
		1338
		1339	// does not fit into IV or UV, try NV
		1340	if (len <= NV_DIG)
		1341	// fits into NV without loss of precision
		1342	return newSVnv (json_atof (start));
		1343
		1344	// everything else fails, convert it to a string
		1345	return newSVpvn (start, dec->cur - start);
		1346	}
		1347
		1348	// loss of precision here
824	return newSVnv (Atof (start));	1349	return newSVnv (json_atof (start));
825		1350
826	fail:	1351	fail:
827	return 0;	1352	return 0;
828	}	1353	}
829		1354
…		…
853	if (*dec->cur == ']')	1378	if (*dec->cur == ']')
854	{	1379	{
855	++dec->cur;	1380	++dec->cur;
856	break;	1381	break;
857	}	1382	}
858		1383
859	if (*dec->cur != ',')	1384	if (*dec->cur != ',')
860	ERR (", or ] expected while parsing array");	1385	ERR (", or ] expected while parsing array");
861		1386
862	++dec->cur;	1387	++dec->cur;
		1388
		1389	decode_ws (dec);
		1390
		1391	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1392	{
		1393	++dec->cur;
		1394	break;
		1395	}
863	}	1396	}
864		1397
865	DEC_DEC_DEPTH;	1398	DEC_DEC_DEPTH;
866	return newRV_noinc ((SV *)av);	1399	return newRV_noinc ((SV *)av);
867		1400
…		…
872	}	1405	}
873		1406
874	static SV *	1407	static SV *
875	decode_hv (dec_t *dec)	1408	decode_hv (dec_t *dec)
876	{	1409	{
		1410	SV *sv;
877	HV *hv = newHV ();	1411	HV *hv = newHV ();
878		1412
879	DEC_INC_DEPTH;	1413	DEC_INC_DEPTH;
880	decode_ws (dec);	1414	decode_ws (dec);
881		1415
882	if (*dec->cur == '}')	1416	if (*dec->cur == '}')
883	++dec->cur;	1417	++dec->cur;
884	else	1418	else
885	for (;;)	1419	for (;;)
886	{	1420	{
		1421	EXPECT_CH ('"');
		1422
		1423	// heuristic: assume that
		1424	// a) decode_str + hv_store_ent are abysmally slow.
		1425	// b) most hash keys are short, simple ascii text.
		1426	// => try to "fast-match" such strings to avoid
		1427	// the overhead of decode_str + hv_store_ent.
		1428	{
887	SV *key, *value;	1429	SV *value;
		1430	char *p = dec->cur;
		1431	char *e = p + 24; // only try up to 24 bytes
888		1432
889	decode_ws (dec); EXPECT_CH ('"');	1433	for (;;)
890
891	key = decode_str (dec);
892	if (!key)
893	goto fail;
894
895	decode_ws (dec); EXPECT_CH (':');
896
897	value = decode_sv (dec);
898	if (!value)
899	{	1434	{
		1435	// the >= 0x80 is false on most architectures
		1436	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1437	{
		1438	// slow path, back up and use decode_str
		1439	SV *key = decode_str (dec);
		1440	if (!key)
		1441	goto fail;
		1442
		1443	decode_ws (dec); EXPECT_CH (':');
		1444
		1445	decode_ws (dec);
		1446	value = decode_sv (dec);
		1447	if (!value)
		1448	{
		1449	SvREFCNT_dec (key);
		1450	goto fail;
		1451	}
		1452
		1453	hv_store_ent (hv, key, value, 0);
900	SvREFCNT_dec (key);	1454	SvREFCNT_dec (key);
		1455
		1456	break;
		1457	}
		1458	else if (*p == '"')
		1459	{
		1460	// fast path, got a simple key
		1461	char *key = dec->cur;
		1462	int len = p - key;
		1463	dec->cur = p + 1;
		1464
		1465	decode_ws (dec); EXPECT_CH (':');
		1466
		1467	decode_ws (dec);
		1468	value = decode_sv (dec);
		1469	if (!value)
901	goto fail;	1470	goto fail;
		1471
		1472	hv_store (hv, key, len, value, 0);
		1473
		1474	break;
		1475	}
		1476
		1477	++p;
902	}	1478	}
903		1479	}
904	hv_store_ent (hv, key, value, 0);
905	SvREFCNT_dec (key);
906		1480
907	decode_ws (dec);	1481	decode_ws (dec);
908		1482
909	if (*dec->cur == '}')	1483	if (*dec->cur == '}')
910	{	1484	{
…		…
914		1488
915	if (*dec->cur != ',')	1489	if (*dec->cur != ',')
916	ERR (", or } expected while parsing object/hash");	1490	ERR (", or } expected while parsing object/hash");
917		1491
918	++dec->cur;	1492	++dec->cur;
		1493
		1494	decode_ws (dec);
		1495
		1496	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1497	{
		1498	++dec->cur;
		1499	break;
		1500	}
919	}	1501	}
920		1502
921	DEC_DEC_DEPTH;	1503	DEC_DEC_DEPTH;
922	return newRV_noinc ((SV *)hv);	1504	sv = newRV_noinc ((SV *)hv);
		1505
		1506	// check filter callbacks
		1507	if (dec->json.flags & F_HOOK)
		1508	{
		1509	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1510	{
		1511	HE *cb, *he;
		1512
		1513	hv_iterinit (hv);
		1514	he = hv_iternext (hv);
		1515	hv_iterinit (hv);
		1516
		1517	// the next line creates a mortal sv each time it's called.
		1518	// might want to optimise this for common cases.
		1519	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1520
		1521	if (cb)
		1522	{
		1523	dSP;
		1524	int count;
		1525
		1526	ENTER; SAVETMPS;
		1527	PUSHMARK (SP);
		1528	XPUSHs (HeVAL (he));
		1529	sv_2mortal (sv);
		1530
		1531	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1532
		1533	if (count == 1)
		1534	{
		1535	sv = newSVsv (POPs);
		1536	FREETMPS; LEAVE;
		1537	return sv;
		1538	}
		1539	else if (count)
		1540	croak ("filter_json_single_key_object callbacks must not return more than one scalar");
		1541
		1542	SvREFCNT_inc (sv);
		1543	FREETMPS; LEAVE;
		1544	}
		1545	}
		1546
		1547	if (dec->json.cb_object)
		1548	{
		1549	dSP;
		1550	int count;
		1551
		1552	ENTER; SAVETMPS;
		1553	PUSHMARK (SP);
		1554	XPUSHs (sv_2mortal (sv));
		1555
		1556	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1557
		1558	if (count == 1)
		1559	{
		1560	sv = newSVsv (POPs);
		1561	FREETMPS; LEAVE;
		1562	return sv;
		1563	}
		1564	else if (count)
		1565	croak ("filter_json_object callbacks must not return more than one scalar");
		1566
		1567	SvREFCNT_inc (sv);
		1568	FREETMPS; LEAVE;
		1569	}
		1570	}
		1571
		1572	return sv;
923		1573
924	fail:	1574	fail:
925	SvREFCNT_dec (hv);	1575	SvREFCNT_dec (hv);
926	DEC_DEC_DEPTH;	1576	DEC_DEC_DEPTH;
927	return 0;	1577	return 0;
928	}	1578	}
929		1579
930	static SV *	1580	static SV *
		1581	decode_tag (dec_t *dec)
		1582	{
		1583	SV *tag = 0;
		1584	SV *val = 0;
		1585
		1586	if (!(dec->json.flags & F_ALLOW_TAGS))
		1587	ERR ("malformed JSON string, neither array, object, number, string or atom");
		1588
		1589	++dec->cur;
		1590
		1591	decode_ws (dec);
		1592
		1593	tag = decode_sv (dec);
		1594	if (!tag)
		1595	goto fail;
		1596
		1597	if (!SvPOK (tag))
		1598	ERR ("malformed JSON string, (tag) must be a string");
		1599
		1600	decode_ws (dec);
		1601
		1602	if (*dec->cur != ')')
		1603	ERR (") expected after tag");
		1604
		1605	++dec->cur;
		1606
		1607	decode_ws (dec);
		1608
		1609	val = decode_sv (dec);
		1610	if (!val)
		1611	goto fail;
		1612
		1613	if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
		1614	ERR ("malformed JSON string, tag value must be an array");
		1615
		1616	{
		1617	AV *av = (AV *)SvRV (val);
		1618	int i, len = av_len (av) + 1;
		1619	HV *stash = gv_stashsv (tag, 0);
		1620	SV *sv;
		1621
		1622	if (!stash)
		1623	ERR ("cannot decode perl-object (package does not exist)");
		1624
		1625	GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
		1626
		1627	if (!method)
		1628	ERR ("cannot decode perl-object (package does not have a THAW method)");
		1629
		1630	dSP;
		1631
		1632	ENTER; SAVETMPS;
		1633	PUSHMARK (SP);
		1634	EXTEND (SP, len + 2);
		1635	// we re-bless the reference to get overload and other niceties right
		1636	PUSHs (tag);
		1637	PUSHs (sv_json);
		1638
		1639	for (i = 0; i < len; ++i)
		1640	PUSHs (*av_fetch (av, i, 1));
		1641
		1642	PUTBACK;
		1643	call_sv ((SV *)GvCV (method), G_SCALAR);
		1644	SPAGAIN;
		1645
		1646	SvREFCNT_dec (tag);
		1647	SvREFCNT_dec (val);
		1648	sv = SvREFCNT_inc (POPs);
		1649
		1650	PUTBACK;
		1651
		1652	FREETMPS; LEAVE;
		1653
		1654	return sv;
		1655	}
		1656
		1657	fail:
		1658	SvREFCNT_dec (tag);
		1659	SvREFCNT_dec (val);
		1660	return 0;
		1661	}
		1662
		1663	static SV *
931	decode_sv (dec_t *dec)	1664	decode_sv (dec_t *dec)
932	{	1665	{
933	decode_ws (dec);	1666	// the beauty of JSON: you need exactly one character lookahead
		1667	// to parse everything.
934	switch (*dec->cur)	1668	switch (*dec->cur)
935	{	1669	{
936	case '"': ++dec->cur; return decode_str (dec);	1670	case '"': ++dec->cur; return decode_str (dec);
937	case '[': ++dec->cur; return decode_av (dec);	1671	case '[': ++dec->cur; return decode_av (dec);
938	case '{': ++dec->cur; return decode_hv (dec);	1672	case '{': ++dec->cur; return decode_hv (dec);
		1673	case '(': return decode_tag (dec);
939		1674
940	case '-':	1675	case '-':
941	case '0': case '1': case '2': case '3': case '4':	1676	case '0': case '1': case '2': case '3': case '4':
942	case '5': case '6': case '7': case '8': case '9':	1677	case '5': case '6': case '7': case '8': case '9':
943	return decode_num (dec);	1678	return decode_num (dec);
944		1679
		1680	case 'f':
		1681	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
		1682	{
		1683	dec->cur += 5;
		1684
		1685	if (expect_false (!dec->json.v_false))
		1686	dec->json.v_false = GET_BOOL (false);
		1687
		1688	return newSVsv (dec->json.v_false);
		1689	}
		1690	else
		1691	ERR ("'false' expected");
		1692
		1693	break;
		1694
945	case 't':	1695	case 't':
946	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1696	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
947	{	1697	{
948	dec->cur += 4;	1698	dec->cur += 4;
		1699
		1700	if (expect_false (!dec->json.v_true))
		1701	dec->json.v_true = GET_BOOL (true);
		1702
949	return newSViv (1);	1703	return newSVsv (dec->json.v_true);
950	}	1704	}
951	else	1705	else
952	ERR ("'true' expected");	1706	ERR ("'true' expected");
953
954	break;
955
956	case 'f':
957	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
958	{
959	dec->cur += 5;
960	return newSViv (0);
961	}
962	else
963	ERR ("'false' expected");
964		1707
965	break;	1708	break;
966		1709
967	case 'n':	1710	case 'n':
968	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))	1711	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
…		…
974	ERR ("'null' expected");	1717	ERR ("'null' expected");
975		1718
976	break;	1719	break;
977		1720
978	default:	1721	default:
979	ERR ("malformed JSON string, neither array, object, number, string or atom");	1722	ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
980	break;	1723	break;
981	}	1724	}
982		1725
983	fail:	1726	fail:
984	return 0;	1727	return 0;
985	}	1728	}
986		1729
987	static SV *	1730	static SV *
988	decode_json (SV *string, U32 flags, UV *offset_return)	1731	decode_json (SV *string, JSON *json, STRLEN *offset_return)
989	{	1732	{
990	dec_t dec;	1733	dec_t dec;
991	UV offset;
992	SV *sv;	1734	SV *sv;
993		1735
		1736	/* work around bugs in 5.10 where manipulating magic values
		1737	* makes perl ignore the magic in subsequent accesses.
		1738	* also make a copy of non-PV values, to get them into a clean
		1739	* state (SvPV should do that, but it's buggy, see below).
		1740	*
		1741	* SvIsCOW_shared_hash works around a bug in perl (possibly 5.16),
		1742	* as reported by Reini Urban.
		1743	*/
994	SvGETMAGIC (string);	1744	/*SvGETMAGIC (string);*/
		1745	if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
		1746	string = sv_2mortal (newSVsv (string));
		1747
995	SvUPGRADE (string, SVt_PV);	1748	SvUPGRADE (string, SVt_PV);
996		1749
997	if (flags & F_UTF8)	1750	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1751	* assertion with -DDEBUGGING, although SvCUR is documented to
		1752	* return the xpv_cur field which certainly exists after upgrading.
		1753	* according to nicholas clark, calling SvPOK fixes this.
		1754	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1755	* and hope for the best.
		1756	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1757	* assertion business is seriously broken, try yet another workaround
		1758	* for the broken -DDEBUGGING.
		1759	*/
		1760	{
		1761	#ifdef DEBUGGING
		1762	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1763	#else
		1764	STRLEN offset = SvCUR (string);
		1765	#endif
		1766
		1767	if (offset > json->max_size && json->max_size)
		1768	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1769	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1770	}
		1771
		1772	if (DECODE_WANTS_OCTETS (json))
998	sv_utf8_downgrade (string, 0);	1773	sv_utf8_downgrade (string, 0);
999	else	1774	else
1000	sv_utf8_upgrade (string);	1775	sv_utf8_upgrade (string);
1001		1776
1002	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1777	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1003		1778
1004	dec.flags = flags;	1779	dec.json = *json;
1005	dec.cur = SvPVX (string);	1780	dec.cur = SvPVX (string);
1006	dec.end = SvEND (string);	1781	dec.end = SvEND (string);
1007	dec.err = 0;	1782	dec.err = 0;
1008	dec.depth = 0;	1783	dec.depth = 0;
1009	dec.maxdepth = DEC_DEPTH (dec.flags);	1784
		1785	if (dec.json.cb_object || dec.json.cb_sk_object)
		1786	dec.json.flags |= F_HOOK;
1010		1787
1011	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1788	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1789
		1790	decode_ws (&dec);
1012	sv = decode_sv (&dec);	1791	sv = decode_sv (&dec);
1013		1792
1014	if (!(offset_return || !sv))	1793	if (offset_return)
		1794	*offset_return = dec.cur - SvPVX (string);
		1795	else if (sv)
1015	{	1796	{
1016	// check for trailing garbage	1797	// check for trailing garbage
1017	decode_ws (&dec);	1798	decode_ws (&dec);
1018		1799
1019	if (*dec.cur)	1800	if (dec.cur != dec.end)
1020	{	1801	{
1021	dec.err = "garbage after JSON object";	1802	dec.err = "garbage after JSON object";
1022	SvREFCNT_dec (sv);	1803	SvREFCNT_dec (sv);
1023	sv = 0;	1804	sv = 0;
1024	}	1805	}
1025	}
1026
1027	if (offset_return || !sv)
1028	{
1029	offset = dec.flags & F_UTF8
1030	? dec.cur - SvPVX (string)
1031	: utf8_distance (dec.cur, SvPVX (string));
1032
1033	if (offset_return)
1034	*offset_return = offset;
1035	}	1806	}
1036		1807
1037	if (!sv)	1808	if (!sv)
1038	{	1809	{
1039	SV *uni = sv_newmortal ();	1810	SV *uni = sv_newmortal ();
…		…
1045	SAVEVPTR (PL_curcop);	1816	SAVEVPTR (PL_curcop);
1046	PL_curcop = &cop;	1817	PL_curcop = &cop;
1047	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1818	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1048	LEAVE;	1819	LEAVE;
1049		1820
1050	croak ("%s, at character offset %d [\"%s\"]",	1821	croak ("%s, at character offset %d (before \"%s\")",
1051	dec.err,	1822	dec.err,
1052	(int)offset,	1823	(int)ptr_to_index (string, dec.cur),
1053	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1824	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1054	}	1825	}
1055		1826
1056	sv = sv_2mortal (sv);	1827	sv = sv_2mortal (sv);
1057		1828
1058	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1829	if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv))
1059	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1830	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1060		1831
1061	return sv;	1832	return sv;
		1833	}
		1834
		1835	/////////////////////////////////////////////////////////////////////////////
		1836	// incremental parser
		1837
		1838	static void
		1839	incr_parse (JSON *self)
		1840	{
		1841	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1842
		1843	// the state machine here is a bit convoluted and could be simplified a lot
		1844	// but this would make it slower, so...
		1845
		1846	for (;;)
		1847	{
		1848	//printf ("loop pod %d *p<%c><%s>, mode %d nest %d\n", p - SvPVX (self->incr_text), *p, p, self->incr_mode, self->incr_nest);//D
		1849	switch (self->incr_mode)
		1850	{
		1851	// only used for initial whitespace skipping
		1852	case INCR_M_WS:
		1853	for (;;)
		1854	{
		1855	if (*p > 0x20)
		1856	{
		1857	if (*p == '#')
		1858	{
		1859	self->incr_mode = INCR_M_C0;
		1860	goto incr_m_c;
		1861	}
		1862	else
		1863	{
		1864	self->incr_mode = INCR_M_JSON;
		1865	goto incr_m_json;
		1866	}
		1867	}
		1868	else if (!*p)
		1869	goto interrupt;
		1870
		1871	++p;
		1872	}
		1873
		1874	// skip a single char inside a string (for \\-processing)
		1875	case INCR_M_BS:
		1876	if (!*p)
		1877	goto interrupt;
		1878
		1879	++p;
		1880	self->incr_mode = INCR_M_STR;
		1881	goto incr_m_str;
		1882
		1883	// inside #-style comments
		1884	case INCR_M_C0:
		1885	case INCR_M_C1:
		1886	incr_m_c:
		1887	for (;;)
		1888	{
		1889	if (*p == '\n')
		1890	{
		1891	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1892	break;
		1893	}
		1894	else if (!*p)
		1895	goto interrupt;
		1896
		1897	++p;
		1898	}
		1899
		1900	break;
		1901
		1902	// inside a string
		1903	case INCR_M_STR:
		1904	incr_m_str:
		1905	for (;;)
		1906	{
		1907	if (*p == '"')
		1908	{
		1909	++p;
		1910	self->incr_mode = INCR_M_JSON;
		1911
		1912	if (!self->incr_nest)
		1913	goto interrupt;
		1914
		1915	goto incr_m_json;
		1916	}
		1917	else if (*p == '\\')
		1918	{
		1919	++p; // "virtually" consumes character after \
		1920
		1921	if (!*p) // if at end of string we have to switch modes
		1922	{
		1923	self->incr_mode = INCR_M_BS;
		1924	goto interrupt;
		1925	}
		1926	}
		1927	else if (!*p)
		1928	goto interrupt;
		1929
		1930	++p;
		1931	}
		1932
		1933	// after initial ws, outside string
		1934	case INCR_M_JSON:
		1935	incr_m_json:
		1936	for (;;)
		1937	{
		1938	switch (*p++)
		1939	{
		1940	case 0:
		1941	--p;
		1942	goto interrupt;
		1943
		1944	case 0x09:
		1945	case 0x0a:
		1946	case 0x0d:
		1947	case 0x20:
		1948	if (!self->incr_nest)
		1949	{
		1950	--p; // do not eat the whitespace, let the next round do it
		1951	goto interrupt;
		1952	}
		1953	break;
		1954
		1955	case '"':
		1956	self->incr_mode = INCR_M_STR;
		1957	goto incr_m_str;
		1958
		1959	case '[':
		1960	case '{':
		1961	case '(':
		1962	if (++self->incr_nest > self->max_depth)
		1963	croak (ERR_NESTING_EXCEEDED);
		1964	break;
		1965
		1966	case ']':
		1967	case '}':
		1968	if (--self->incr_nest <= 0)
		1969	goto interrupt;
		1970	break;
		1971
		1972	case ')':
		1973	--self->incr_nest;
		1974	break;
		1975
		1976	case '#':
		1977	self->incr_mode = INCR_M_C1;
		1978	goto incr_m_c;
		1979	}
		1980	}
		1981	}
		1982
		1983	modechange:
		1984	;
		1985	}
		1986
		1987	interrupt:
		1988	self->incr_pos = p - SvPVX (self->incr_text);
		1989	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1990	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1062	}	1991	}
1063		1992
1064	/////////////////////////////////////////////////////////////////////////////	1993	/////////////////////////////////////////////////////////////////////////////
1065	// XS interface functions	1994	// XS interface functions
1066		1995
1067	MODULE = JSON::XS PACKAGE = JSON::XS	1996	MODULE = JSON::XS PACKAGE = JSON::XS
1068		1997
1069	BOOT:	1998	BOOT:
1070	{	1999	{
1071	int i;	2000	int i;
1072
1073	memset (decode_hexdigit, 0xff, 256);
1074		2001
1075	for (i = 0; i < 256; ++i)	2002	for (i = 0; i < 256; ++i)
1076	decode_hexdigit [i] =	2003	decode_hexdigit [i] =
1077	i >= '0' && i <= '9' ? i - '0'	2004	i >= '0' && i <= '9' ? i - '0'
1078	: i >= 'a' && i <= 'f' ? i - 'a' + 10	2005	: i >= 'a' && i <= 'f' ? i - 'a' + 10
1079	: i >= 'A' && i <= 'F' ? i - 'A' + 10	2006	: i >= 'A' && i <= 'F' ? i - 'A' + 10
1080	: -1;	2007	: -1;
1081		2008
1082	json_stash = gv_stashpv ("JSON::XS", 1);	2009	json_stash = gv_stashpv ("JSON::XS" , 1);
		2010	bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
		2011	bool_false = get_bool ("Types::Serialiser::false");
		2012	bool_true = get_bool ("Types::Serialiser::true");
		2013
		2014	sv_json = newSVpv ("JSON", 0);
		2015	SvREADONLY_on (sv_json);
		2016
		2017	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1083	}	2018	}
1084		2019
1085	PROTOTYPES: DISABLE	2020	PROTOTYPES: DISABLE
1086		2021
1087	SV *new (char *dummy)	2022	void CLONE (...)
1088	CODE:	2023	CODE:
1089	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	2024	// as long as these writes are atomic, the race should not matter
		2025	// as existing threads either already use 0, or use the old value,
		2026	// which is sitll correct for the initial thread.
		2027	json_stash = 0;
		2028	bool_stash = 0;
		2029	bool_false = 0;
		2030	bool_true = 0;
		2031
		2032	void new (char *klass)
		2033	PPCODE:
		2034	{
		2035	SV *pv = NEWSV (0, sizeof (JSON));
		2036	SvPOK_only (pv);
		2037	json_init ((JSON *)SvPVX (pv));
		2038	XPUSHs (sv_2mortal (sv_bless (
		2039	newRV_noinc (pv),
		2040	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		2041	)));
		2042	}
		2043
		2044	void boolean_values (JSON *self, SV *v_false = 0, SV *v_true = 0)
		2045	PPCODE:
		2046	self->v_false = newSVsv (v_false);
		2047	self->v_true = newSVsv (v_true);
		2048	XPUSHs (ST (0));
		2049
		2050	void get_boolean_values (JSON *self)
		2051	PPCODE:
		2052	if (self->v_false && self->v_true)
		2053	{
		2054	EXTEND (SP, 2);
		2055	PUSHs (self->v_false);
		2056	PUSHs (self->v_true);
		2057	}
		2058
		2059	void ascii (JSON *self, int enable = 1)
		2060	ALIAS:
		2061	ascii = F_ASCII
		2062	latin1 = F_LATIN1
		2063	utf8 = F_UTF8
		2064	indent = F_INDENT
		2065	canonical = F_CANONICAL
		2066	space_before = F_SPACE_BEFORE
		2067	space_after = F_SPACE_AFTER
		2068	pretty = F_PRETTY
		2069	allow_nonref = F_ALLOW_NONREF
		2070	shrink = F_SHRINK
		2071	allow_blessed = F_ALLOW_BLESSED
		2072	convert_blessed = F_CONV_BLESSED
		2073	relaxed = F_RELAXED
		2074	allow_unknown = F_ALLOW_UNKNOWN
		2075	allow_tags = F_ALLOW_TAGS
		2076	PPCODE:
		2077	{
		2078	if (enable)
		2079	self->flags |= ix;
		2080	else
		2081	self->flags &= ~ix;
		2082
		2083	XPUSHs (ST (0));
		2084	}
		2085
		2086	void get_ascii (JSON *self)
		2087	ALIAS:
		2088	get_ascii = F_ASCII
		2089	get_latin1 = F_LATIN1
		2090	get_utf8 = F_UTF8
		2091	get_indent = F_INDENT
		2092	get_canonical = F_CANONICAL
		2093	get_space_before = F_SPACE_BEFORE
		2094	get_space_after = F_SPACE_AFTER
		2095	get_allow_nonref = F_ALLOW_NONREF
		2096	get_shrink = F_SHRINK
		2097	get_allow_blessed = F_ALLOW_BLESSED
		2098	get_convert_blessed = F_CONV_BLESSED
		2099	get_relaxed = F_RELAXED
		2100	get_allow_unknown = F_ALLOW_UNKNOWN
		2101	get_allow_tags = F_ALLOW_TAGS
		2102	PPCODE:
		2103	XPUSHs (boolSV (self->flags & ix));
		2104
		2105	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		2106	PPCODE:
		2107	self->max_depth = max_depth;
		2108	XPUSHs (ST (0));
		2109
		2110	U32 get_max_depth (JSON *self)
		2111	CODE:
		2112	RETVAL = self->max_depth;
1090	OUTPUT:	2113	OUTPUT:
1091	RETVAL	2114	RETVAL
1092		2115
1093	SV *ascii (SV *self, int enable = 1)	2116	void max_size (JSON *self, U32 max_size = 0)
1094	ALIAS:	2117	PPCODE:
1095	ascii = F_ASCII	2118	self->max_size = max_size;
1096	latin1 = F_LATIN1	2119	XPUSHs (ST (0));
1097	utf8 = F_UTF8	2120
1098	indent = F_INDENT	2121	int get_max_size (JSON *self)
1099	canonical = F_CANONICAL
1100	space_before = F_SPACE_BEFORE
1101	space_after = F_SPACE_AFTER
1102	pretty = F_PRETTY
1103	allow_nonref = F_ALLOW_NONREF
1104	shrink = F_SHRINK
1105	CODE:	2122	CODE:
1106	{	2123	RETVAL = self->max_size;
1107	UV *uv = SvJSON (self);
1108	if (enable)
1109	*uv |= ix;
1110	else
1111	*uv &= ~ix;
1112
1113	RETVAL = newSVsv (self);
1114	}
1115	OUTPUT:	2124	OUTPUT:
1116	RETVAL	2125	RETVAL
1117		2126
1118	SV *max_depth (SV *self, UV max_depth = 0x80000000UL)	2127	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
		2128	PPCODE:
		2129	{
		2130	SvREFCNT_dec (self->cb_object);
		2131	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		2132
		2133	XPUSHs (ST (0));
		2134	}
		2135
		2136	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
		2137	PPCODE:
		2138	{
		2139	if (!self->cb_sk_object)
		2140	self->cb_sk_object = newHV ();
		2141
		2142	if (SvOK (cb))
		2143	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		2144	else
		2145	{
		2146	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		2147
		2148	if (!HvKEYS (self->cb_sk_object))
		2149	{
		2150	SvREFCNT_dec (self->cb_sk_object);
		2151	self->cb_sk_object = 0;
		2152	}
		2153	}
		2154
		2155	XPUSHs (ST (0));
		2156	}
		2157
		2158	void encode (JSON *self, SV *scalar)
		2159	PPCODE:
		2160	PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
		2161	XPUSHs (scalar);
		2162
		2163	void decode (JSON *self, SV *jsonstr)
		2164	PPCODE:
		2165	PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
		2166	XPUSHs (jsonstr);
		2167
		2168	void decode_prefix (JSON *self, SV *jsonstr)
		2169	PPCODE:
		2170	{
		2171	SV *sv;
		2172	STRLEN offset;
		2173	PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
		2174	EXTEND (SP, 2);
		2175	PUSHs (sv);
		2176	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset))));
		2177	}
		2178
		2179	void incr_parse (JSON *self, SV *jsonstr = 0)
		2180	PPCODE:
		2181	{
		2182	if (!self->incr_text)
		2183	self->incr_text = newSVpvn ("", 0);
		2184
		2185	/* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
		2186	if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
		2187	if (DECODE_WANTS_OCTETS (self))
		2188	{
		2189	if (self->incr_pos)
		2190	self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
		2191	(U8 *)SvPVX (self->incr_text) + self->incr_pos);
		2192
		2193	sv_utf8_downgrade (self->incr_text, 0);
		2194	}
		2195	else
		2196	{
		2197	sv_utf8_upgrade (self->incr_text);
		2198
		2199	if (self->incr_pos)
		2200	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		2201	- (U8 *)SvPVX (self->incr_text);
		2202	}
		2203
		2204	// append data, if any
		2205	if (jsonstr)
		2206	{
		2207	/* make sure both strings have same encoding */
		2208	if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
		2209	if (SvUTF8 (jsonstr))
		2210	sv_utf8_downgrade (jsonstr, 0);
		2211	else
		2212	sv_utf8_upgrade (jsonstr);
		2213
		2214	/* and then just blindly append */
		2215	{
		2216	STRLEN len;
		2217	const char *str = SvPV (jsonstr, len);
		2218	STRLEN cur = SvCUR (self->incr_text);
		2219
		2220	if (SvLEN (self->incr_text) - cur <= len)
		2221	json_sv_grow (self->incr_text, cur, len);
		2222
		2223	Move (str, SvEND (self->incr_text), len, char);
		2224	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		2225	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		2226	}
		2227	}
		2228
		2229	if (GIMME_V != G_VOID)
		2230	do
		2231	{
		2232	SV *sv;
		2233	STRLEN offset;
		2234
		2235	if (!INCR_DONE (self))
		2236	{
		2237	incr_parse (self);
		2238
		2239	if (self->incr_pos > self->max_size && self->max_size)
		2240	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		2241	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		2242
		2243	if (!INCR_DONE (self))
		2244	{
		2245	// as an optimisation, do not accumulate white space in the incr buffer
		2246	if (self->incr_mode == INCR_M_WS && self->incr_pos)
		2247	{
		2248	self->incr_pos = 0;
		2249	SvCUR_set (self->incr_text, 0);
		2250	}
		2251
		2252	break;
		2253	}
		2254	}
		2255
		2256	PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
		2257	XPUSHs (sv);
		2258
		2259	self->incr_pos -= offset;
		2260	self->incr_nest = 0;
		2261	self->incr_mode = 0;
		2262
		2263	sv_chop (self->incr_text, SvPVX (self->incr_text) + offset);
		2264	}
		2265	while (GIMME_V == G_ARRAY);
		2266	}
		2267
		2268	SV *incr_text (JSON *self)
		2269	ATTRS: lvalue
1119	CODE:	2270	CODE:
1120	{	2271	{
1121	UV *uv = SvJSON (self);	2272	if (self->incr_pos)
1122	UV log2 = 0;	2273	croak ("incr_text can not be called when the incremental parser already started parsing");
1123		2274
1124	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	2275	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
1125
1126	while ((1UL << log2) < max_depth)
1127	++log2;
1128
1129	*uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1130
1131	RETVAL = newSVsv (self);
1132	}	2276	}
1133	OUTPUT:	2277	OUTPUT:
1134	RETVAL	2278	RETVAL
1135		2279
		2280	void incr_skip (JSON *self)
		2281	CODE:
		2282	{
		2283	if (self->incr_pos)
		2284	{
		2285	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2286	self->incr_pos = 0;
		2287	self->incr_nest = 0;
		2288	self->incr_mode = 0;
		2289	}
		2290	}
		2291
		2292	void incr_reset (JSON *self)
		2293	CODE:
		2294	{
		2295	SvREFCNT_dec (self->incr_text);
		2296	self->incr_text = 0;
		2297	self->incr_pos = 0;
		2298	self->incr_nest = 0;
		2299	self->incr_mode = 0;
		2300	}
		2301
		2302	void DESTROY (JSON *self)
		2303	CODE:
		2304	SvREFCNT_dec (self->v_false);
		2305	SvREFCNT_dec (self->v_true);
		2306	SvREFCNT_dec (self->cb_sk_object);
		2307	SvREFCNT_dec (self->cb_object);
		2308	SvREFCNT_dec (self->incr_text);
		2309
		2310	PROTOTYPES: ENABLE
		2311
1136	void encode (SV *self, SV *scalar)	2312	void encode_json (SV *scalar)
1137	PPCODE:	2313	PPCODE:
1138	XPUSHs (encode_json (scalar, *SvJSON (self)));	2314	{
		2315	JSON json;
		2316	json_init (&json);
		2317	json.flags |= F_UTF8;
		2318	PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
		2319	XPUSHs (scalar);
		2320	}
1139		2321
1140	void decode (SV *self, SV *jsonstr)	2322	void decode_json (SV *jsonstr)
1141	PPCODE:	2323	PPCODE:
1142	XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));
1143
1144	void decode_prefix (SV *self, SV *jsonstr)
1145	PPCODE:
1146	{	2324	{
1147	UV offset;	2325	JSON json;
1148	EXTEND (SP, 2);	2326	json_init (&json);
1149	PUSHs (decode_json (jsonstr, *SvJSON (self), &offset));	2327	json.flags |= F_UTF8;
1150	PUSHs (sv_2mortal (newSVuv (offset)));	2328	PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
		2329	XPUSHs (jsonstr);
1151	}	2330	}
1152		2331
1153	PROTOTYPES: ENABLE
1154
1155	void to_json (SV *scalar)
1156	ALIAS:
1157	objToJson = 0
1158	PPCODE:
1159	XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8));
1160
1161	void from_json (SV *jsonstr)
1162	ALIAS:
1163	jsonToObj = 0
1164	PPCODE:
1165	XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8, 0));
1166

Diff Legend

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