ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.8 by root, Fri Mar 23 16:13:59 2007 UTC vs.
Revision 1.129 by root, Wed Nov 16 18:06:34 2016 UTC

1	#include "EXTERN.h"	1	#include "EXTERN.h"
2	#include "perl.h"	2	#include "perl.h"
3	#include "XSUB.h"	3	#include "XSUB.h"
4		4
5	#include "assert.h"	5	#include <assert.h>
6	#include "string.h"	6	#include <string.h>
7	#include "stdlib.h"	7	#include <stdlib.h>
		8	#include <stdio.h>
		9	#include <limits.h>
		10	#include <float.h>
8		11
		12	#if defined(__BORLANDC__) || defined(_MSC_VER)
		13	# define snprintf _snprintf // C compilers have this in stdio.h
		14	#endif
		15
		16	// some old perls do not have this, try to make it work, no
		17	// guarantees, though. if it breaks, you get to keep the pieces.
		18	#ifndef UTF8_MAXBYTES
		19	# define UTF8_MAXBYTES 13
		20	#endif
		21
		22	// compatibility with perl <5.18
		23	#ifndef HvNAMELEN_get
		24	# define HvNAMELEN_get(hv) strlen (HvNAME (hv))
		25	#endif
		26	#ifndef HvNAMELEN
		27	# define HvNAMELEN(hv) HvNAMELEN_get (hv)
		28	#endif
		29	#ifndef HvNAMEUTF8
		30	# define HvNAMEUTF8(hv) 0
		31	#endif
		32
		33	// three extra for rounding, sign, and end of string
		34	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
		35
9	#define F_ASCII 0x00000001	36	#define F_ASCII 0x00000001UL
		37	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	38	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	39	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	40	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	41	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	42	#define F_SPACE_AFTER 0x00000040UL
15	#define F_JSON_RPC 0x00000040
16	#define F_ALLOW_NONREF 0x00000080	43	#define F_ALLOW_NONREF 0x00000100UL
17	#define F_SHRINK 0x00000100	44	#define F_SHRINK 0x00000200UL
		45	#define F_ALLOW_BLESSED 0x00000400UL
		46	#define F_CONV_BLESSED 0x00000800UL
		47	#define F_RELAXED 0x00001000UL
		48	#define F_ALLOW_UNKNOWN 0x00002000UL
		49	#define F_ALLOW_TAGS 0x00004000UL
		50	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
18		51
19	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	52	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
20	#define F_DEFAULT 0
21		53
22	#define INIT_SIZE 32 // initial scalar size to be allocated	54	#define INIT_SIZE 32 // initial scalar size to be allocated
		55	#define INDENT_STEP 3 // spaces per indentation level
		56
		57	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
		58
		59	#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
23		60
24	#define SB do {	61	#define SB do {
25	#define SE } while (0)	62	#define SE } while (0)
26		63
27	static HV *json_stash;	64	#if __GNUC__ >= 3
		65	# define expect(expr,value) __builtin_expect ((expr), (value))
		66	# define INLINE static inline
		67	#else
		68	# define expect(expr,value) (expr)
		69	# define INLINE static
		70	#endif
		71
		72	#define expect_false(expr) expect ((expr) != 0, 0)
		73	#define expect_true(expr) expect ((expr) != 0, 1)
		74
		75	#define IN_RANGE_INC(type,val,beg,end) \
		76	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		77	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		78
		79	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
		80
		81	#ifdef USE_ITHREADS
		82	# define JSON_SLOW 1
		83	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		84	# define BOOL_STASH (bool_stash ? bool_stash : gv_stashpv ("Types::Serialiser::Boolean", 1))
		85	#else
		86	# define JSON_SLOW 0
		87	# define JSON_STASH json_stash
		88	# define BOOL_STASH bool_stash
		89	#endif
		90
		91	// the amount of HEs to allocate on the stack, when sorting keys
		92	#define STACK_HES 64
		93
		94	static HV *json_stash, *bool_stash; // JSON::XS::, Types::Serialiser::Boolean::
		95	static SV *bool_true, *bool_false, *sv_json;
		96
		97	enum {
		98	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		99	INCR_M_STR, // inside string
		100	INCR_M_BS, // inside backslash
		101	INCR_M_C0, // inside comment in initial whitespace sequence
		102	INCR_M_C1, // inside comment in other places
		103	INCR_M_JSON // outside anything, count nesting
		104	};
		105
		106	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		107
		108	typedef struct {
		109	U32 flags;
		110	U32 max_depth;
		111	STRLEN max_size;
		112
		113	SV *cb_object;
		114	HV *cb_sk_object;
		115
		116	// for the incremental parser
		117	SV *incr_text; // the source text so far
		118	STRLEN incr_pos; // the current offset into the text
		119	int incr_nest; // {[]}-nesting level
		120	unsigned char incr_mode;
		121	} JSON;
		122
		123	INLINE void
		124	json_init (JSON *json)
		125	{
		126	Zero (json, 1, JSON);
		127	json->max_depth = 512;
		128	}
		129
		130	/////////////////////////////////////////////////////////////////////////////
		131	// utility functions
		132
		133	INLINE SV *
		134	get_bool (const char *name)
		135	{
		136	SV *sv = get_sv (name, 1);
		137
		138	SvREADONLY_on (sv);
		139	SvREADONLY_on (SvRV (sv));
		140
		141	return sv;
		142	}
		143
		144	INLINE void
		145	shrink (SV *sv)
		146	{
		147	sv_utf8_downgrade (sv, 1);
		148
		149	if (SvLEN (sv) > SvCUR (sv) + 1)
		150	{
		151	#ifdef SvPV_shrink_to_cur
		152	SvPV_shrink_to_cur (sv);
		153	#elif defined (SvPV_renew)
		154	SvPV_renew (sv, SvCUR (sv) + 1);
		155	#endif
		156	}
		157	}
		158
		159	// decode an utf-8 character and return it, or (UV)-1 in
		160	// case of an error.
		161	// we special-case "safe" characters from U+80 .. U+7FF,
		162	// but use the very good perl function to parse anything else.
		163	// note that we never call this function for a ascii codepoints
		164	INLINE UV
		165	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		166	{
		167	if (expect_true (len >= 2
		168	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
		169	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
		170	{
		171	*clen = 2;
		172	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		173	}
		174	else
		175	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		176	}
		177
		178	// likewise for encoding, also never called for ascii codepoints
		179	// this function takes advantage of this fact, although current gccs
		180	// seem to optimise the check for >= 0x80 away anyways
		181	INLINE unsigned char *
		182	encode_utf8 (unsigned char *s, UV ch)
		183	{
		184	if (expect_false (ch < 0x000080))
		185	*s++ = ch;
		186	else if (expect_true (ch < 0x000800))
		187	*s++ = 0xc0 | ( ch >> 6),
		188	*s++ = 0x80 | ( ch & 0x3f);
		189	else if ( ch < 0x010000)
		190	*s++ = 0xe0 | ( ch >> 12),
		191	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		192	*s++ = 0x80 | ( ch & 0x3f);
		193	else if ( ch < 0x110000)
		194	*s++ = 0xf0 | ( ch >> 18),
		195	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		196	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		197	*s++ = 0x80 | ( ch & 0x3f);
		198
		199	return s;
		200	}
		201
		202	// convert offset pointer to character index, sv must be string
		203	static STRLEN
		204	ptr_to_index (SV *sv, char *offset)
		205	{
		206	return SvUTF8 (sv)
		207	? utf8_distance (offset, SvPVX (sv))
		208	: offset - SvPVX (sv);
		209	}
		210
		211	/////////////////////////////////////////////////////////////////////////////
		212	// fp hell
		213
		214	// scan a group of digits, and a trailing exponent
		215	static void
		216	json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
		217	{
		218	UV uaccum = 0;
		219	int eaccum = 0;
		220
		221	// if we recurse too deep, skip all remaining digits
		222	// to avoid a stack overflow attack
		223	if (expect_false (--maxdepth <= 0))
		224	while (((U8)*s - '0') < 10)
		225	++s;
		226
		227	for (;;)
		228	{
		229	U8 dig = (U8)*s - '0';
		230
		231	if (expect_false (dig >= 10))
		232	{
		233	if (dig == (U8)((U8)'.' - (U8)'0'))
		234	{
		235	++s;
		236	json_atof_scan1 (s, accum, expo, 1, maxdepth);
		237	}
		238	else if ((dig | ' ') == 'e' - '0')
		239	{
		240	int exp2 = 0;
		241	int neg = 0;
		242
		243	++s;
		244
		245	if (*s == '-')
		246	{
		247	++s;
		248	neg = 1;
		249	}
		250	else if (*s == '+')
		251	++s;
		252
		253	while ((dig = (U8)*s - '0') < 10)
		254	exp2 = exp2 * 10 + *s++ - '0';
		255
		256	*expo += neg ? -exp2 : exp2;
		257	}
		258
		259	break;
		260	}
		261
		262	++s;
		263
		264	uaccum = uaccum * 10 + dig;
		265	++eaccum;
		266
		267	// if we have too many digits, then recurse for more
		268	// we actually do this for rather few digits
		269	if (uaccum >= (UV_MAX - 9) / 10)
		270	{
		271	if (postdp) *expo -= eaccum;
		272	json_atof_scan1 (s, accum, expo, postdp, maxdepth);
		273	if (postdp) *expo += eaccum;
		274
		275	break;
		276	}
		277	}
		278
		279	// this relies greatly on the quality of the pow ()
		280	// implementation of the platform, but a good
		281	// implementation is hard to beat.
		282	// (IEEE 754 conformant ones are required to be exact)
		283	if (postdp) *expo -= eaccum;
		284	*accum += uaccum * Perl_pow (10., *expo);
		285	*expo += eaccum;
		286	}
		287
		288	static NV
		289	json_atof (const char *s)
		290	{
		291	NV accum = 0.;
		292	int expo = 0;
		293	int neg = 0;
		294
		295	if (*s == '-')
		296	{
		297	++s;
		298	neg = 1;
		299	}
		300
		301	// a recursion depth of ten gives us >>500 bits
		302	json_atof_scan1 (s, &accum, &expo, 0, 10);
		303
		304	return neg ? -accum : accum;
		305	}
		306
		307	// target of scalar reference is bool? -1 == nope, 0 == false, 1 == true
		308	static int
		309	ref_bool_type (SV *sv)
		310	{
		311	svtype svt = SvTYPE (sv);
		312
		313	if (svt < SVt_PVAV)
		314	{
		315	STRLEN len = 0;
		316	char *pv = svt ? SvPV (sv, len) : 0;
		317
		318	if (len == 1)
		319	if (*pv == '1')
		320	return 1;
		321	else if (*pv == '0')
		322	return 0;
		323	}
		324
		325	return -1;
		326	}
		327
		328	// returns whether scalar is not a reference in the sense of allow_nonref
		329	static int
		330	json_nonref (SV *scalar)
		331	{
		332	if (!SvROK (scalar))
		333	return 1;
		334
		335	scalar = SvRV (scalar);
		336
		337	if (SvTYPE (scalar) >= SVt_PVMG)
		338	{
		339	if (SvSTASH (scalar) == bool_stash)
		340	return 1;
		341
		342	if (!SvOBJECT (scalar) && ref_bool_type (scalar) >= 0)
		343	return 1;
		344	}
		345
		346	return 0;
		347	}
		348
		349	/////////////////////////////////////////////////////////////////////////////
		350	// encoder
28		351
29	// structure used for encoding JSON	352	// structure used for encoding JSON
30	typedef struct	353	typedef struct
31	{	354	{
32	char *cur;	355	char *cur; // SvPVX (sv) + current output position
33	STRLEN len; // SvLEN (sv)
34	char *end; // SvEND (sv)	356	char *end; // SvEND (sv)
35	SV *sv;	357	SV *sv; // result scalar
36	UV flags;	358	JSON json;
37	int max_recurse;	359	U32 indent; // indentation level
38	int indent;	360	UV limit; // escape character values >= this value when encoding
39	} enc_t;	361	} enc_t;
40		362
41	// structure used for decoding JSON	363	INLINE void
42	typedef struct
43	{
44	char *cur;
45	char *end;
46	const char *err;
47	UV flags;
48	} dec_t;
49
50	static UV *
51	SvJSON (SV *sv)
52	{
53	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
54	croak ("object is not of type JSON::XS");
55
56	return &SvUVX (SvRV (sv));
57	}
58
59	static void
60	shrink (SV *sv)
61	{
62	sv_utf8_downgrade (sv, 1);
63	#ifdef SvPV_shrink_to_cur
64	SvPV_shrink_to_cur (sv);
65	#endif
66	}
67
68	/////////////////////////////////////////////////////////////////////////////
69
70	static void
71	need (enc_t *enc, STRLEN len)	364	need (enc_t *enc, STRLEN len)
72	{	365	{
73	if (enc->cur + len >= enc->end)	366	if (expect_false (enc->cur + len >= enc->end))
74	{	367	{
75	STRLEN cur = enc->cur - SvPVX (enc->sv);	368	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
76	SvGROW (enc->sv, cur + len + 1);	369	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
77	enc->cur = SvPVX (enc->sv) + cur;	370	enc->cur = SvPVX (enc->sv) + cur;
78	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	371	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
79	}	372	}
80	}	373	}
81		374
82	static void	375	INLINE void
83	encode_ch (enc_t *enc, char ch)	376	encode_ch (enc_t *enc, char ch)
84	{	377	{
85	need (enc, 1);	378	need (enc, 1);
86	*enc->cur++ = ch;	379	*enc->cur++ = ch;
87	}	380	}
…		…
95		388
96	while (str < end)	389	while (str < end)
97	{	390	{
98	unsigned char ch = *(unsigned char *)str;	391	unsigned char ch = *(unsigned char *)str;
99		392
100	if (ch >= 0x20 && ch < 0x80) // most common case	393	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
101	{	394	{
102	if (ch == '"') // but with slow exceptions	395	if (expect_false (ch == '"')) // but with slow exceptions
103	{	396	{
104	need (enc, len += 1);	397	need (enc, len += 1);
105	*enc->cur++ = '\\';	398	*enc->cur++ = '\\';
106	*enc->cur++ = '"';	399	*enc->cur++ = '"';
107	}	400	}
108	else if (ch == '\\')	401	else if (expect_false (ch == '\\'))
109	{	402	{
110	need (enc, len += 1);	403	need (enc, len += 1);
111	*enc->cur++ = '\\';	404	*enc->cur++ = '\\';
112	*enc->cur++ = '\\';	405	*enc->cur++ = '\\';
113	}	406	}
…		…
131	STRLEN clen;	424	STRLEN clen;
132	UV uch;	425	UV uch;
133		426
134	if (is_utf8)	427	if (is_utf8)
135	{	428	{
136	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	429	uch = decode_utf8 (str, end - str, &clen);
137	if (clen == (STRLEN)-1)	430	if (clen == (STRLEN)-1)
138	croak ("malformed UTF-8 character in string, cannot convert to JSON");	431	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
139	}	432	}
140	else	433	else
141	{	434	{
142	uch = ch;	435	uch = ch;
143	clen = 1;	436	clen = 1;
144	}	437	}
145		438
146	if (uch < 0x80 || enc->flags & F_ASCII)	439	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
147	{	440	{
148	if (uch > 0xFFFFUL)	441	if (uch >= 0x10000UL)
149	{	442	{
		443	if (uch >= 0x110000UL)
		444	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		445
150	need (enc, len += 11);	446	need (enc, len += 11);
151	sprintf (enc->cur, "\\u%04x\\u%04x",	447	sprintf (enc->cur, "\\u%04x\\u%04x",
152	(uch - 0x10000) / 0x400 + 0xD800,	448	(int)((uch - 0x10000) / 0x400 + 0xD800),
153	(uch - 0x10000) % 0x400 + 0xDC00);	449	(int)((uch - 0x10000) % 0x400 + 0xDC00));
154	enc->cur += 12;	450	enc->cur += 12;
155	}	451	}
156	else	452	else
157	{	453	{
158	static char hexdigit [16] = "0123456789abcdef";
159	need (enc, len += 5);	454	need (enc, len += 5);
160	*enc->cur++ = '\\';	455	*enc->cur++ = '\\';
161	*enc->cur++ = 'u';	456	*enc->cur++ = 'u';
162	*enc->cur++ = hexdigit [ uch >> 12 ];	457	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
163	*enc->cur++ = hexdigit [(uch >> 8) & 15];	458	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
164	*enc->cur++ = hexdigit [(uch >> 4) & 15];	459	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
165	*enc->cur++ = hexdigit [(uch >> 0) & 15];	460	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
166	}	461	}
167		462
		463	str += clen;
		464	}
		465	else if (enc->json.flags & F_LATIN1)
		466	{
		467	*enc->cur++ = uch;
168	str += clen;	468	str += clen;
169	}	469	}
170	else if (is_utf8)	470	else if (is_utf8)
171	{	471	{
172	need (enc, len += clen);	472	need (enc, len += clen);
…		…
176	}	476	}
177	while (--clen);	477	while (--clen);
178	}	478	}
179	else	479	else
180	{	480	{
181	need (enc, len += 10); // never more than 11 bytes needed	481	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
182	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	482	enc->cur = encode_utf8 (enc->cur, uch);
183	++str;	483	++str;
184	}	484	}
185	}	485	}
186	}	486	}
187	}	487	}
188		488
189	--len;	489	--len;
190	}	490	}
191	}	491	}
192		492
193	#define INDENT SB \	493	INLINE void
		494	encode_indent (enc_t *enc)
		495	{
194	if (enc->flags & F_INDENT) \	496	if (enc->json.flags & F_INDENT)
195	{ \	497	{
196	int i_; \	498	int spaces = enc->indent * INDENT_STEP;
197	need (enc, enc->indent); \	499
198	for (i_ = enc->indent * 3; i_--; )\	500	need (enc, spaces);
		501	memset (enc->cur, ' ', spaces);
		502	enc->cur += spaces;
		503	}
		504	}
		505
		506	INLINE void
		507	encode_space (enc_t *enc)
		508	{
		509	need (enc, 1);
		510	encode_ch (enc, ' ');
		511	}
		512
		513	INLINE void
		514	encode_nl (enc_t *enc)
		515	{
		516	if (enc->json.flags & F_INDENT)
		517	{
		518	need (enc, 1);
199	encode_ch (enc, ' '); \	519	encode_ch (enc, '\n');
200	} \	520	}
201	SE	521	}
202		522
203	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	523	INLINE void
204	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	524	encode_comma (enc_t *enc)
205	#define COMMA SB \	525	{
206	encode_ch (enc, ','); \	526	encode_ch (enc, ',');
		527
207	if (enc->flags & F_INDENT) \	528	if (enc->json.flags & F_INDENT)
208	NL; \	529	encode_nl (enc);
209	else if (enc->flags & F_SPACE_AFTER) \	530	else if (enc->json.flags & F_SPACE_AFTER)
210	SPACE; \	531	encode_space (enc);
211	SE	532	}
212		533
213	static void encode_sv (enc_t *enc, SV *sv);	534	static void encode_sv (enc_t *enc, SV *sv);
214		535
215	static void	536	static void
216	encode_av (enc_t *enc, AV *av)	537	encode_av (enc_t *enc, AV *av)
217	{	538	{
218	int i, len = av_len (av);	539	int i, len = av_len (av);
219		540
		541	if (enc->indent >= enc->json.max_depth)
		542	croak (ERR_NESTING_EXCEEDED);
		543
220	encode_ch (enc, '['); NL;	544	encode_ch (enc, '[');
221	++enc->indent;
222		545
		546	if (len >= 0)
		547	{
		548	encode_nl (enc); ++enc->indent;
		549
223	for (i = 0; i <= len; ++i)	550	for (i = 0; i <= len; ++i)
224	{	551	{
225	INDENT;	552	SV **svp = av_fetch (av, i, 0);
226	encode_sv (enc, *av_fetch (av, i, 0));
227		553
		554	encode_indent (enc);
		555
		556	if (svp)
		557	encode_sv (enc, *svp);
		558	else
		559	encode_str (enc, "null", 4, 0);
		560
228	if (i < len)	561	if (i < len)
229	COMMA;	562	encode_comma (enc);
230	}	563	}
231		564
232	NL;	565	encode_nl (enc); --enc->indent; encode_indent (enc);
		566	}
233		567
234	--enc->indent;
235	INDENT; encode_ch (enc, ']');	568	encode_ch (enc, ']');
236	}	569	}
237		570
238	static void	571	static void
239	encode_he (enc_t *enc, HE *he)	572	encode_hk (enc_t *enc, HE *he)
240	{	573	{
241	encode_ch (enc, '"');	574	encode_ch (enc, '"');
242		575
243	if (HeKLEN (he) == HEf_SVKEY)	576	if (HeKLEN (he) == HEf_SVKEY)
244	{	577	{
245	SV *sv = HeSVKEY (he);	578	SV *sv = HeSVKEY (he);
246	STRLEN len;	579	STRLEN len;
247	char *str;	580	char *str;
248		581
249	SvGETMAGIC (sv);	582	SvGETMAGIC (sv);
250	str = SvPV (sv, len);	583	str = SvPV (sv, len);
251		584
252	encode_str (enc, str, len, SvUTF8 (sv));	585	encode_str (enc, str, len, SvUTF8 (sv));
253	}	586	}
254	else	587	else
255	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	588	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
256		589
257	encode_ch (enc, '"');	590	encode_ch (enc, '"');
258		591
259	if (enc->flags & F_SPACE_BEFORE) SPACE;	592	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
260	encode_ch (enc, ':');	593	encode_ch (enc, ':');
261	if (enc->flags & F_SPACE_AFTER ) SPACE;	594	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
262	encode_sv (enc, HeVAL (he));
263	}	595	}
264		596
265	// compare hash entries, used when all keys are bytestrings	597	// compare hash entries, used when all keys are bytestrings
266	static int	598	static int
267	he_cmp_fast (const void *a_, const void *b_)	599	he_cmp_fast (const void *a_, const void *b_)
…		…
272	HE *b = *(HE **)b_;	604	HE *b = *(HE **)b_;
273		605
274	STRLEN la = HeKLEN (a);	606	STRLEN la = HeKLEN (a);
275	STRLEN lb = HeKLEN (b);	607	STRLEN lb = HeKLEN (b);
276		608
277	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	609	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
278	cmp = la < lb ? -1 : la == lb ? 0 : 1;	610	cmp = lb - la;
279		611
280	return cmp;	612	return cmp;
281	}	613	}
282		614
283	// compare hash entries, used when some keys are sv's or utf-x	615	// compare hash entries, used when some keys are sv's or utf-x
284	static int	616	static int
285	he_cmp_slow (const void *a, const void *b)	617	he_cmp_slow (const void *a, const void *b)
286	{	618	{
287	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	619	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
288	}	620	}
289		621
290	static void	622	static void
291	encode_hv (enc_t *enc, HV *hv)	623	encode_hv (enc_t *enc, HV *hv)
292	{	624	{
293	int count, i;	625	HE *he;
294		626
295	encode_ch (enc, '{'); NL; ++enc->indent;	627	if (enc->indent >= enc->json.max_depth)
		628	croak (ERR_NESTING_EXCEEDED);
296		629
297	if ((count = hv_iterinit (hv)))	630	encode_ch (enc, '{');
298	{	631
299	// for canonical output we have to sort by keys first	632	// for canonical output we have to sort by keys first
300	// actually, this is mostly due to the stupid so-called	633	// actually, this is mostly due to the stupid so-called
301	// security workaround added somewhere in 5.8.x.	634	// security workaround added somewhere in 5.8.x
302	// that randomises hash orderings	635	// that randomises hash orderings
303	if (enc->flags & F_CANONICAL)	636	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
		637	{
		638	int count = hv_iterinit (hv);
		639
		640	if (SvMAGICAL (hv))
304	{	641	{
305	HE *he, *hes [count];	642	// need to count by iterating. could improve by dynamically building the vector below
		643	// but I don't care for the speed of this special case.
		644	// note also that we will run into undefined behaviour when the two iterations
		645	// do not result in the same count, something I might care for in some later release.
		646
		647	count = 0;
		648	while (hv_iternext (hv))
		649	++count;
		650
		651	hv_iterinit (hv);
		652	}
		653
		654	if (count)
		655	{
306	int fast = 1;	656	int i, fast = 1;
		657	HE *hes_stack [STACK_HES];
		658	HE **hes = hes_stack;
		659
		660	// allocate larger arrays on the heap
		661	if (count > STACK_HES)
		662	{
		663	SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
		664	hes = (HE **)SvPVX (sv);
		665	}
307		666
308	i = 0;	667	i = 0;
309	while ((he = hv_iternext (hv)))	668	while ((he = hv_iternext (hv)))
310	{	669	{
311	hes [i++] = he;	670	hes [i++] = he;
…		…
333		692
334	FREETMPS;	693	FREETMPS;
335	LEAVE;	694	LEAVE;
336	}	695	}
337		696
338	for (i = 0; i < count; ++i)	697	encode_nl (enc); ++enc->indent;
		698
		699	while (count--)
339	{	700	{
340	INDENT;	701	encode_indent (enc);
		702	he = hes [count];
341	encode_he (enc, hes [i]);	703	encode_hk (enc, he);
		704	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
342		705
343	if (i < count - 1)	706	if (count)
344	COMMA;	707	encode_comma (enc);
345	}	708	}
346		709
347	NL;	710	encode_nl (enc); --enc->indent; encode_indent (enc);
348	}	711	}
		712	}
		713	else
		714	{
		715	if (hv_iterinit (hv) || SvMAGICAL (hv))
		716	if ((he = hv_iternext (hv)))
		717	{
		718	encode_nl (enc); ++enc->indent;
		719
		720	for (;;)
		721	{
		722	encode_indent (enc);
		723	encode_hk (enc, he);
		724	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		725
		726	if (!(he = hv_iternext (hv)))
		727	break;
		728
		729	encode_comma (enc);
		730	}
		731
		732	encode_nl (enc); --enc->indent; encode_indent (enc);
		733	}
		734	}
		735
		736	encode_ch (enc, '}');
		737	}
		738
		739	// encode objects, arrays and special \0=false and \1=true values.
		740	static void
		741	encode_rv (enc_t *enc, SV *sv)
		742	{
		743	svtype svt;
		744	GV *method;
		745
		746	SvGETMAGIC (sv);
		747	svt = SvTYPE (sv);
		748
		749	if (expect_false (SvOBJECT (sv)))
		750	{
		751	HV *stash = SvSTASH (sv);
		752
		753	if (stash == bool_stash)
		754	{
		755	if (SvIV (sv))
		756	encode_str (enc, "true", 4, 0);
		757	else
		758	encode_str (enc, "false", 5, 0);
		759	}
		760	else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
		761	{
		762	int count;
		763	dSP;
		764
		765	ENTER; SAVETMPS;
		766	SAVESTACK_POS ();
		767	PUSHMARK (SP);
		768	EXTEND (SP, 2);
		769	// we re-bless the reference to get overload and other niceties right
		770	PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		771	PUSHs (sv_json);
		772
		773	PUTBACK;
		774	count = call_sv ((SV *)GvCV (method), G_ARRAY);
		775	SPAGAIN;
		776
		777	// catch this surprisingly common error
		778	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		779	croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		780
		781	encode_ch (enc, '(');
		782	encode_ch (enc, '"');
		783	encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
		784	encode_ch (enc, '"');
		785	encode_ch (enc, ')');
		786	encode_ch (enc, '[');
		787
		788	while (count)
		789	{
		790	encode_sv (enc, SP[1 - count--]);
		791
		792	if (count)
		793	encode_ch (enc, ',');
		794	}
		795
		796	encode_ch (enc, ']');
		797
		798	FREETMPS; LEAVE;
		799	}
		800	else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
		801	{
		802	dSP;
		803
		804	ENTER; SAVETMPS;
		805	PUSHMARK (SP);
		806	// we re-bless the reference to get overload and other niceties right
		807	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		808
		809	// calling with G_SCALAR ensures that we always get a 1 return value
		810	PUTBACK;
		811	call_sv ((SV *)GvCV (method), G_SCALAR);
		812	SPAGAIN;
		813
		814	// catch this surprisingly common error
		815	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		816	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		817
		818	sv = POPs;
		819	PUTBACK;
		820
		821	encode_sv (enc, sv);
		822
		823	FREETMPS; LEAVE;
		824	}
		825	else if (enc->json.flags & F_ALLOW_BLESSED)
		826	encode_str (enc, "null", 4, 0);
349	else	827	else
350	{	828	croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
351	SV *sv;	829	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
352	HE *he = hv_iternext (hv);	830	}
353		831	else if (svt == SVt_PVHV)
354	for (;;)	832	encode_hv (enc, (HV *)sv);
355	{	833	else if (svt == SVt_PVAV)
356	INDENT;	834	encode_av (enc, (AV *)sv);
357	encode_he (enc, he);	835	else if (svt < SVt_PVAV)
358
359	if (!(he = hv_iternext (hv)))
360	break;
361
362	COMMA;
363	}
364
365	NL;
366	}
367	}	836	{
		837	int bool_type = ref_bool_type (sv);
368		838
369	--enc->indent; INDENT; encode_ch (enc, '}');	839	if (bool_type == 1)
		840	encode_str (enc, "true", 4, 0);
		841	else if (bool_type == 0)
		842	encode_str (enc, "false", 5, 0);
		843	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		844	encode_str (enc, "null", 4, 0);
		845	else
		846	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		847	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		848	}
		849	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		850	encode_str (enc, "null", 4, 0);
		851	else
		852	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		853	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
370	}	854	}
371		855
372	static void	856	static void
373	encode_sv (enc_t *enc, SV *sv)	857	encode_sv (enc_t *enc, SV *sv)
374	{	858	{
…		…
382	encode_str (enc, str, len, SvUTF8 (sv));	866	encode_str (enc, str, len, SvUTF8 (sv));
383	encode_ch (enc, '"');	867	encode_ch (enc, '"');
384	}	868	}
385	else if (SvNOKp (sv))	869	else if (SvNOKp (sv))
386	{	870	{
		871	// trust that perl will do the right thing w.r.t. JSON syntax.
387	need (enc, NV_DIG + 32);	872	need (enc, NV_DIG + 32);
388	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	873	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
389	enc->cur += strlen (enc->cur);	874	enc->cur += strlen (enc->cur);
390	}	875	}
391	else if (SvIOKp (sv))	876	else if (SvIOKp (sv))
392	{	877	{
		878	// we assume we can always read an IV as a UV and vice versa
		879	// we assume two's complement
		880	// we assume no aliasing issues in the union
		881	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		882	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		883	{
		884	// optimise the "small number case"
		885	// code will likely be branchless and use only a single multiplication
		886	// works for numbers up to 59074
		887	I32 i = SvIVX (sv);
		888	U32 u;
		889	char digit, nz = 0;
		890
393	need (enc, 64);	891	need (enc, 6);
		892
		893	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		894	u = i < 0 ? -i : i;
		895
		896	// convert to 4.28 fixed-point representation
		897	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		898
		899	// now output digit by digit, each time masking out the integer part
		900	// and multiplying by 5 while moving the decimal point one to the right,
		901	// resulting in a net multiplication by 10.
		902	// we always write the digit to memory but conditionally increment
		903	// the pointer, to enable the use of conditional move instructions.
		904	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		905	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		906	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		907	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		908	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		909	}
		910	else
		911	{
		912	// large integer, use the (rather slow) snprintf way.
		913	need (enc, IVUV_MAXCHARS);
394	enc->cur +=	914	enc->cur +=
395	SvIsUV(sv)	915	SvIsUV(sv)
396	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	916	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
397	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	917	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		918	}
398	}	919	}
399	else if (SvROK (sv))	920	else if (SvROK (sv))
400	{	921	encode_rv (enc, SvRV (sv));
401	if (!--enc->max_recurse)	922	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
402	croak ("data structure too deep (hit recursion limit)");
403
404	sv = SvRV (sv);
405
406	switch (SvTYPE (sv))
407	{
408	case SVt_PVAV: encode_av (enc, (AV *)sv); break;
409	case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
410
411	default:
412	croak ("JSON can only represent references to arrays or hashes");
413	}
414	}
415	else if (!SvOK (sv))
416	encode_str (enc, "null", 4, 0);	923	encode_str (enc, "null", 4, 0);
417	else	924	else
418	croak ("encountered perl type that JSON cannot handle");	925	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
		926	SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
419	}	927	}
420		928
421	static SV *	929	static SV *
422	encode_json (SV *scalar, UV flags)	930	encode_json (SV *scalar, JSON *json)
423	{	931	{
424	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
425	croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");
426
427	enc_t enc;	932	enc_t enc;
428	enc.flags = flags;	933
		934	if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar))
		935	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
		936
		937	enc.json = *json;
429	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	938	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
430	enc.cur = SvPVX (enc.sv);	939	enc.cur = SvPVX (enc.sv);
431	enc.end = SvEND (enc.sv);	940	enc.end = SvEND (enc.sv);
432	enc.max_recurse = 0;
433	enc.indent = 0;	941	enc.indent = 0;
		942	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		943	: enc.json.flags & F_LATIN1 ? 0x000100UL
		944	: 0x110000UL;
434		945
435	SvPOK_only (enc.sv);	946	SvPOK_only (enc.sv);
436	encode_sv (&enc, scalar);	947	encode_sv (&enc, scalar);
		948	encode_nl (&enc);
437		949
		950	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		951	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		952
438	if (!(flags & (F_ASCII | F_UTF8)))	953	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
439	SvUTF8_on (enc.sv);	954	SvUTF8_on (enc.sv);
440		955
441	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
442
443	if (enc.flags & F_SHRINK)	956	if (enc.json.flags & F_SHRINK)
444	shrink (enc.sv);	957	shrink (enc.sv);
445		958
446	return enc.sv;	959	return enc.sv;
447	}	960	}
448		961
449	/////////////////////////////////////////////////////////////////////////////	962	/////////////////////////////////////////////////////////////////////////////
		963	// decoder
450		964
451	#define WS \	965	// structure used for decoding JSON
		966	typedef struct
		967	{
		968	char *cur; // current parser pointer
		969	char *end; // end of input string
		970	const char *err; // parse error, if != 0
		971	JSON json;
		972	U32 depth; // recursion depth
		973	U32 maxdepth; // recursion depth limit
		974	} dec_t;
		975
		976	INLINE void
		977	decode_comment (dec_t *dec)
		978	{
		979	// only '#'-style comments allowed a.t.m.
		980
		981	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		982	++dec->cur;
		983	}
		984
		985	INLINE void
		986	decode_ws (dec_t *dec)
		987	{
452	for (;;) \	988	for (;;)
453	{ \	989	{
454	char ch = *dec->cur; \	990	char ch = *dec->cur;
		991
455	if (ch > 0x20 \	992	if (ch > 0x20)
		993	{
		994	if (expect_false (ch == '#'))
		995	{
		996	if (dec->json.flags & F_RELAXED)
		997	decode_comment (dec);
		998	else
		999	break;
		1000	}
		1001	else
		1002	break;
		1003	}
456	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	1004	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
457	break; \	1005	break; // parse error, but let higher level handle it, gives better error messages
		1006
458	++dec->cur; \	1007	++dec->cur;
459	}	1008	}
		1009	}
460		1010
461	#define ERR(reason) SB dec->err = reason; goto fail; SE	1011	#define ERR(reason) SB dec->err = reason; goto fail; SE
		1012
462	#define EXPECT_CH(ch) SB \	1013	#define EXPECT_CH(ch) SB \
463	if (*dec->cur != ch) \	1014	if (*dec->cur != ch) \
464	ERR (# ch " expected"); \	1015	ERR (# ch " expected"); \
465	++dec->cur; \	1016	++dec->cur; \
466	SE	1017	SE
467		1018
		1019	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
		1020	#define DEC_DEC_DEPTH --dec->depth
		1021
468	static SV *decode_sv (dec_t *dec);	1022	static SV *decode_sv (dec_t *dec);
469		1023
470	static signed char decode_hexdigit[256];	1024	static signed char decode_hexdigit[256];
471		1025
472	static UV	1026	static UV
473	decode_4hex (dec_t *dec)	1027	decode_4hex (dec_t *dec)
474	{	1028	{
475	signed char d1, d2, d3, d4;	1029	signed char d1, d2, d3, d4;
		1030	unsigned char *cur = (unsigned char *)dec->cur;
476		1031
477	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	1032	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
478	if (d1 < 0) ERR ("four hexadecimal digits expected");	1033	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
479	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	1034	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
480	if (d2 < 0) ERR ("four hexadecimal digits expected");	1035	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
481	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
482	if (d3 < 0) ERR ("four hexadecimal digits expected");
483	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
484	if (d4 < 0) ERR ("four hexadecimal digits expected");
485		1036
486	dec->cur += 4;	1037	dec->cur += 4;
487		1038
488	return ((UV)d1) << 12	1039	return ((UV)d1) << 12
489	| ((UV)d2) << 8	1040	| ((UV)d2) << 8
…		…
492		1043
493	fail:	1044	fail:
494	return (UV)-1;	1045	return (UV)-1;
495	}	1046	}
496		1047
497	#define APPEND_GROW(n) SB \
498	if (cur + (n) >= end) \
499	{ \
500	STRLEN ofs = cur - SvPVX (sv); \
501	SvGROW (sv, ofs + (n) + 1); \
502	cur = SvPVX (sv) + ofs; \
503	end = SvEND (sv); \
504	} \
505	SE
506
507	#define APPEND_CH(ch) SB \
508	APPEND_GROW (1); \
509	*cur++ = (ch); \
510	SE
511
512	static SV *	1048	static SV *
513	decode_str (dec_t *dec)	1049	decode_str (dec_t *dec)
514	{	1050	{
515	SV *sv = NEWSV (0,2);	1051	SV *sv = 0;
516	int utf8 = 0;	1052	int utf8 = 0;
517	char *cur = SvPVX (sv);	1053	char *dec_cur = dec->cur;
518	char *end = SvEND (sv);
519		1054
520	for (;;)	1055	do
521	{	1056	{
522	unsigned char ch = *(unsigned char *)dec->cur;	1057	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		1058	char *cur = buf;
523		1059
524	if (ch == '"')	1060	do
525	break;
526	else if (ch == '\\')
527	{	1061	{
528	switch (*++dec->cur)	1062	unsigned char ch = *(unsigned char *)dec_cur++;
		1063
		1064	if (expect_false (ch == '"'))
529	{	1065	{
530	case '\\':	1066	--dec_cur;
531	case '/':	1067	break;
532	case '"': APPEND_CH (*dec->cur++); break;	1068	}
533		1069	else if (expect_false (ch == '\\'))
534	case 'b': APPEND_CH ('\010'); ++dec->cur; break;	1070	{
535	case 't': APPEND_CH ('\011'); ++dec->cur; break;	1071	switch (*dec_cur)
536	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
537	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
538	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
539
540	case 'u':
541	{	1072	{
542	UV lo, hi;	1073	case '\\':
543	++dec->cur;	1074	case '/':
		1075	case '"': *cur++ = *dec_cur++; break;
544		1076
545	hi = decode_4hex (dec);	1077	case 'b': ++dec_cur; *cur++ = '\010'; break;
546	if (hi == (UV)-1)	1078	case 't': ++dec_cur; *cur++ = '\011'; break;
547	goto fail;	1079	case 'n': ++dec_cur; *cur++ = '\012'; break;
		1080	case 'f': ++dec_cur; *cur++ = '\014'; break;
		1081	case 'r': ++dec_cur; *cur++ = '\015'; break;
548		1082
549	// possibly a surrogate pair	1083	case 'u':
550	if (hi >= 0xd800 && hi < 0xdc00)
551	{	1084	{
552	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	1085	UV lo, hi;
553	ERR ("missing low surrogate character in surrogate pair");	1086	++dec_cur;
554		1087
555	dec->cur += 2;	1088	dec->cur = dec_cur;
556
557	lo = decode_4hex (dec);	1089	hi = decode_4hex (dec);
		1090	dec_cur = dec->cur;
558	if (lo == (UV)-1)	1091	if (hi == (UV)-1)
559	goto fail;	1092	goto fail;
560		1093
		1094	// possibly a surrogate pair
		1095	if (hi >= 0xd800)
		1096	if (hi < 0xdc00)
		1097	{
		1098	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		1099	ERR ("missing low surrogate character in surrogate pair");
		1100
		1101	dec_cur += 2;
		1102
		1103	dec->cur = dec_cur;
		1104	lo = decode_4hex (dec);
		1105	dec_cur = dec->cur;
		1106	if (lo == (UV)-1)
		1107	goto fail;
		1108
561	if (lo < 0xdc00 || lo >= 0xe000)	1109	if (lo < 0xdc00 || lo >= 0xe000)
562	ERR ("surrogate pair expected");	1110	ERR ("surrogate pair expected");
563		1111
564	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	1112	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		1113	}
		1114	else if (hi < 0xe000)
		1115	ERR ("missing high surrogate character in surrogate pair");
		1116
		1117	if (hi >= 0x80)
		1118	{
		1119	utf8 = 1;
		1120
		1121	cur = encode_utf8 (cur, hi);
		1122	}
		1123	else
		1124	*cur++ = hi;
565	}	1125	}
566	else if (hi >= 0xdc00 && hi < 0xe000)
567	ERR ("missing high surrogate character in surrogate pair");
568
569	if (hi >= 0x80)
570	{	1126	break;
571	utf8 = 1;
572		1127
573	APPEND_GROW (4); // at most 4 bytes for 21 bits
574	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
575	}
576	else	1128	default:
577	APPEND_CH (hi);	1129	--dec_cur;
		1130	ERR ("illegal backslash escape sequence in string");
578	}	1131	}
579	break;	1132	}
		1133	else if (expect_true (ch >= 0x20 && ch < 0x80))
		1134	*cur++ = ch;
		1135	else if (ch >= 0x80)
		1136	{
		1137	STRLEN clen;
580		1138
581	default:
582	--dec->cur;	1139	--dec_cur;
583	ERR ("illegal backslash escape sequence in string");	1140
		1141	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		1142	if (clen == (STRLEN)-1)
		1143	ERR ("malformed UTF-8 character in JSON string");
		1144
		1145	do
		1146	*cur++ = *dec_cur++;
		1147	while (--clen);
		1148
		1149	utf8 = 1;
		1150	}
		1151	else if (ch == '\t' && dec->json.flags & F_RELAXED)
		1152	*cur++ = ch;
		1153	else
		1154	{
		1155	--dec_cur;
		1156
		1157	if (!ch)
		1158	ERR ("unexpected end of string while parsing JSON string");
		1159	else
		1160	ERR ("invalid character encountered while parsing JSON string");
584	}	1161	}
585	}	1162	}
586	else if (ch >= 0x20 && ch <= 0x7f)	1163	while (cur < buf + SHORT_STRING_LEN);
587	APPEND_CH (*dec->cur++);	1164
588	else if (ch >= 0x80)
589	{	1165	{
590	STRLEN clen;	1166	STRLEN len = cur - buf;
591	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
592	if (clen == (STRLEN)-1)
593	ERR ("malformed UTF-8 character in JSON string");
594		1167
595	APPEND_GROW (clen);	1168	if (sv)
596	do
597	{	1169	{
598	*cur++ = *dec->cur++;	1170	STRLEN cur = SvCUR (sv);
		1171
		1172	if (SvLEN (sv) <= cur + len)
		1173	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1174
		1175	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		1176	SvCUR_set (sv, SvCUR (sv) + len);
599	}	1177	}
600	while (--clen);
601
602	utf8 = 1;
603	}
604	else if (dec->cur == dec->end)
605	ERR ("unexpected end of string while parsing json string");
606	else	1178	else
607	ERR ("invalid character encountered");	1179	sv = newSVpvn (buf, len);
608	}	1180	}
		1181	}
		1182	while (*dec_cur != '"');
609		1183
610	++dec->cur;	1184	++dec_cur;
611		1185
612	SvCUR_set (sv, cur - SvPVX (sv));	1186	if (sv)
613		1187	{
614	SvPOK_only (sv);	1188	SvPOK_only (sv);
615	*SvEND (sv) = 0;	1189	*SvEND (sv) = 0;
616		1190
617	if (utf8)	1191	if (utf8)
618	SvUTF8_on (sv);	1192	SvUTF8_on (sv);
		1193	}
		1194	else
		1195	sv = newSVpvn ("", 0);
619		1196
620	if (dec->flags & F_SHRINK)	1197	dec->cur = dec_cur;
621	shrink (sv);
622
623	return sv;	1198	return sv;
624		1199
625	fail:	1200	fail:
626	SvREFCNT_dec (sv);	1201	dec->cur = dec_cur;
627	return 0;	1202	return 0;
628	}	1203	}
629		1204
630	static SV *	1205	static SV *
631	decode_num (dec_t *dec)	1206	decode_num (dec_t *dec)
…		…
689	is_nv = 1;	1264	is_nv = 1;
690	}	1265	}
691		1266
692	if (!is_nv)	1267	if (!is_nv)
693	{	1268	{
694	UV uv;	1269	int len = dec->cur - start;
695	int numtype = grok_number (start, dec->cur - start, &uv);	1270
696	if (numtype & IS_NUMBER_IN_UV)	1271	// special case the rather common 1..5-digit-int case
697	if (numtype & IS_NUMBER_NEG)	1272	if (*start == '-')
		1273	switch (len)
698	{	1274	{
699	if (uv < (UV)IV_MIN)	1275	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
700	return newSViv (-(IV)uv);	1276	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
		1277	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1278	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1279	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
701	}	1280	}
		1281	else
		1282	switch (len)
		1283	{
		1284	case 1: return newSViv ( start [0] - '0' * 1);
		1285	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1286	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1287	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1288	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1289	}
		1290
		1291	{
		1292	UV uv;
		1293	int numtype = grok_number (start, len, &uv);
		1294	if (numtype & IS_NUMBER_IN_UV)
		1295	if (numtype & IS_NUMBER_NEG)
		1296	{
		1297	if (uv < (UV)IV_MIN)
		1298	return newSViv (-(IV)uv);
		1299	}
702	else	1300	else
703	return newSVuv (uv);	1301	return newSVuv (uv);
704	}	1302	}
705		1303
		1304	len -= *start == '-' ? 1 : 0;
		1305
		1306	// does not fit into IV or UV, try NV
		1307	if (len <= NV_DIG)
		1308	// fits into NV without loss of precision
		1309	return newSVnv (json_atof (start));
		1310
		1311	// everything else fails, convert it to a string
		1312	return newSVpvn (start, dec->cur - start);
		1313	}
		1314
		1315	// loss of precision here
706	return newSVnv (Atof (start));	1316	return newSVnv (json_atof (start));
707		1317
708	fail:	1318	fail:
709	return 0;	1319	return 0;
710	}	1320	}
711		1321
712	static SV *	1322	static SV *
713	decode_av (dec_t *dec)	1323	decode_av (dec_t *dec)
714	{	1324	{
715	AV *av = newAV ();	1325	AV *av = newAV ();
716		1326
717	WS;	1327	DEC_INC_DEPTH;
		1328	decode_ws (dec);
		1329
718	if (*dec->cur == ']')	1330	if (*dec->cur == ']')
719	++dec->cur;	1331	++dec->cur;
720	else	1332	else
721	for (;;)	1333	for (;;)
722	{	1334	{
…		…
726	if (!value)	1338	if (!value)
727	goto fail;	1339	goto fail;
728		1340
729	av_push (av, value);	1341	av_push (av, value);
730		1342
731	WS;	1343	decode_ws (dec);
732		1344
733	if (*dec->cur == ']')	1345	if (*dec->cur == ']')
734	{	1346	{
735	++dec->cur;	1347	++dec->cur;
736	break;	1348	break;
737	}	1349	}
738		1350
739	if (*dec->cur != ',')	1351	if (*dec->cur != ',')
740	ERR (", or ] expected while parsing array");	1352	ERR (", or ] expected while parsing array");
741		1353
742	++dec->cur;	1354	++dec->cur;
		1355
		1356	decode_ws (dec);
		1357
		1358	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1359	{
		1360	++dec->cur;
		1361	break;
		1362	}
743	}	1363	}
744		1364
		1365	DEC_DEC_DEPTH;
745	return newRV_noinc ((SV *)av);	1366	return newRV_noinc ((SV *)av);
746		1367
747	fail:	1368	fail:
748	SvREFCNT_dec (av);	1369	SvREFCNT_dec (av);
		1370	DEC_DEC_DEPTH;
749	return 0;	1371	return 0;
750	}	1372	}
751		1373
752	static SV *	1374	static SV *
753	decode_hv (dec_t *dec)	1375	decode_hv (dec_t *dec)
754	{	1376	{
		1377	SV *sv;
755	HV *hv = newHV ();	1378	HV *hv = newHV ();
756		1379
757	WS;	1380	DEC_INC_DEPTH;
		1381	decode_ws (dec);
		1382
758	if (*dec->cur == '}')	1383	if (*dec->cur == '}')
759	++dec->cur;	1384	++dec->cur;
760	else	1385	else
761	for (;;)	1386	for (;;)
762	{	1387	{
763	SV *key, *value;
764
765	WS; EXPECT_CH ('"');	1388	EXPECT_CH ('"');
766		1389
767	key = decode_str (dec);	1390	// heuristic: assume that
768	if (!key)	1391	// a) decode_str + hv_store_ent are abysmally slow.
769	goto fail;	1392	// b) most hash keys are short, simple ascii text.
		1393	// => try to "fast-match" such strings to avoid
		1394	// the overhead of decode_str + hv_store_ent.
		1395	{
		1396	SV *value;
		1397	char *p = dec->cur;
		1398	char *e = p + 24; // only try up to 24 bytes
770		1399
771	WS; EXPECT_CH (':');	1400	for (;;)
772
773	value = decode_sv (dec);
774	if (!value)
775	{	1401	{
		1402	// the >= 0x80 is false on most architectures
		1403	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1404	{
		1405	// slow path, back up and use decode_str
		1406	SV *key = decode_str (dec);
		1407	if (!key)
		1408	goto fail;
		1409
		1410	decode_ws (dec); EXPECT_CH (':');
		1411
		1412	decode_ws (dec);
		1413	value = decode_sv (dec);
		1414	if (!value)
		1415	{
		1416	SvREFCNT_dec (key);
		1417	goto fail;
		1418	}
		1419
		1420	hv_store_ent (hv, key, value, 0);
776	SvREFCNT_dec (key);	1421	SvREFCNT_dec (key);
		1422
		1423	break;
		1424	}
		1425	else if (*p == '"')
		1426	{
		1427	// fast path, got a simple key
		1428	char *key = dec->cur;
		1429	int len = p - key;
		1430	dec->cur = p + 1;
		1431
		1432	decode_ws (dec); EXPECT_CH (':');
		1433
		1434	decode_ws (dec);
		1435	value = decode_sv (dec);
		1436	if (!value)
777	goto fail;	1437	goto fail;
		1438
		1439	hv_store (hv, key, len, value, 0);
		1440
		1441	break;
		1442	}
		1443
		1444	++p;
778	}	1445	}
779
780	//TODO: optimise
781	hv_store_ent (hv, key, value, 0);
782
783	WS;	1446	}
		1447
		1448	decode_ws (dec);
784		1449
785	if (*dec->cur == '}')	1450	if (*dec->cur == '}')
786	{	1451	{
787	++dec->cur;	1452	++dec->cur;
788	break;	1453	break;
…		…
790		1455
791	if (*dec->cur != ',')	1456	if (*dec->cur != ',')
792	ERR (", or } expected while parsing object/hash");	1457	ERR (", or } expected while parsing object/hash");
793		1458
794	++dec->cur;	1459	++dec->cur;
		1460
		1461	decode_ws (dec);
		1462
		1463	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1464	{
		1465	++dec->cur;
		1466	break;
		1467	}
795	}	1468	}
796		1469
		1470	DEC_DEC_DEPTH;
797	return newRV_noinc ((SV *)hv);	1471	sv = newRV_noinc ((SV *)hv);
		1472
		1473	// check filter callbacks
		1474	if (dec->json.flags & F_HOOK)
		1475	{
		1476	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1477	{
		1478	HE *cb, *he;
		1479
		1480	hv_iterinit (hv);
		1481	he = hv_iternext (hv);
		1482	hv_iterinit (hv);
		1483
		1484	// the next line creates a mortal sv each time it's called.
		1485	// might want to optimise this for common cases.
		1486	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1487
		1488	if (cb)
		1489	{
		1490	dSP;
		1491	int count;
		1492
		1493	ENTER; SAVETMPS;
		1494	SAVESTACK_POS ();
		1495	PUSHMARK (SP);
		1496	XPUSHs (HeVAL (he));
		1497	sv_2mortal (sv);
		1498
		1499	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1500
		1501	if (count == 1)
		1502	{
		1503	sv = newSVsv (POPs);
		1504	FREETMPS; LEAVE;
		1505	return sv;
		1506	}
		1507
		1508	SvREFCNT_inc (sv);
		1509	FREETMPS; LEAVE;
		1510	}
		1511	}
		1512
		1513	if (dec->json.cb_object)
		1514	{
		1515	dSP;
		1516	int count;
		1517
		1518	ENTER; SAVETMPS;
		1519	SAVESTACK_POS ();
		1520	PUSHMARK (SP);
		1521	XPUSHs (sv_2mortal (sv));
		1522
		1523	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1524
		1525	if (count == 1)
		1526	{
		1527	sv = newSVsv (POPs);
		1528	FREETMPS; LEAVE;
		1529	return sv;
		1530	}
		1531
		1532	SvREFCNT_inc (sv);
		1533	FREETMPS; LEAVE;
		1534	}
		1535	}
		1536
		1537	return sv;
798		1538
799	fail:	1539	fail:
800	SvREFCNT_dec (hv);	1540	SvREFCNT_dec (hv);
		1541	DEC_DEC_DEPTH;
		1542	return 0;
		1543	}
		1544
		1545	static SV *
		1546	decode_tag (dec_t *dec)
		1547	{
		1548	SV *tag = 0;
		1549	SV *val = 0;
		1550
		1551	if (!(dec->json.flags & F_ALLOW_TAGS))
		1552	ERR ("malformed JSON string, neither array, object, number, string or atom");
		1553
		1554	++dec->cur;
		1555
		1556	decode_ws (dec);
		1557
		1558	tag = decode_sv (dec);
		1559	if (!tag)
		1560	goto fail;
		1561
		1562	if (!SvPOK (tag))
		1563	ERR ("malformed JSON string, (tag) must be a string");
		1564
		1565	decode_ws (dec);
		1566
		1567	if (*dec->cur != ')')
		1568	ERR (") expected after tag");
		1569
		1570	++dec->cur;
		1571
		1572	decode_ws (dec);
		1573
		1574	val = decode_sv (dec);
		1575	if (!val)
		1576	goto fail;
		1577
		1578	if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
		1579	ERR ("malformed JSON string, tag value must be an array");
		1580
		1581	{
		1582	AV *av = (AV *)SvRV (val);
		1583	int i, len = av_len (av) + 1;
		1584	HV *stash = gv_stashsv (tag, 0);
		1585	SV *sv;
		1586
		1587	if (!stash)
		1588	ERR ("cannot decode perl-object (package does not exist)");
		1589
		1590	GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
		1591
		1592	if (!method)
		1593	ERR ("cannot decode perl-object (package does not have a THAW method)");
		1594
		1595	dSP;
		1596
		1597	ENTER; SAVETMPS;
		1598	PUSHMARK (SP);
		1599	EXTEND (SP, len + 2);
		1600	// we re-bless the reference to get overload and other niceties right
		1601	PUSHs (tag);
		1602	PUSHs (sv_json);
		1603
		1604	for (i = 0; i < len; ++i)
		1605	PUSHs (*av_fetch (av, i, 1));
		1606
		1607	PUTBACK;
		1608	call_sv ((SV *)GvCV (method), G_SCALAR);
		1609	SPAGAIN;
		1610
		1611	SvREFCNT_dec (tag);
		1612	SvREFCNT_dec (val);
		1613	sv = SvREFCNT_inc (POPs);
		1614
		1615	PUTBACK;
		1616
		1617	FREETMPS; LEAVE;
		1618
		1619	return sv;
		1620	}
		1621
		1622	fail:
		1623	SvREFCNT_dec (tag);
		1624	SvREFCNT_dec (val);
801	return 0;	1625	return 0;
802	}	1626	}
803		1627
804	static SV *	1628	static SV *
805	decode_sv (dec_t *dec)	1629	decode_sv (dec_t *dec)
806	{	1630	{
807	WS;	1631	// the beauty of JSON: you need exactly one character lookahead
		1632	// to parse everything.
808	switch (*dec->cur)	1633	switch (*dec->cur)
809	{	1634	{
810	case '"': ++dec->cur; return decode_str (dec);	1635	case '"': ++dec->cur; return decode_str (dec);
811	case '[': ++dec->cur; return decode_av (dec);	1636	case '[': ++dec->cur; return decode_av (dec);
812	case '{': ++dec->cur; return decode_hv (dec);	1637	case '{': ++dec->cur; return decode_hv (dec);
		1638	case '(': return decode_tag (dec);
813		1639
814	case '-':	1640	case '-':
815	case '0': case '1': case '2': case '3': case '4':	1641	case '0': case '1': case '2': case '3': case '4':
816	case '5': case '6': case '7': case '8': case '9':	1642	case '5': case '6': case '7': case '8': case '9':
817	return decode_num (dec);	1643	return decode_num (dec);
818		1644
819	case 't':	1645	case 't':
820	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1646	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
821	{	1647	{
822	dec->cur += 4;	1648	dec->cur += 4;
		1649	#if JSON_SLOW
		1650	bool_true = get_bool ("Types::Serialiser::true");
		1651	#endif
823	return newSViv (1);	1652	return newSVsv (bool_true);
824	}	1653	}
825	else	1654	else
826	ERR ("'true' expected");	1655	ERR ("'true' expected");
827		1656
828	break;	1657	break;
829		1658
830	case 'f':	1659	case 'f':
831	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1660	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
832	{	1661	{
833	dec->cur += 5;	1662	dec->cur += 5;
		1663	#if JSON_SLOW
		1664	bool_false = get_bool ("Types::Serialiser::false");
		1665	#endif
834	return newSViv (0);	1666	return newSVsv (bool_false);
835	}	1667	}
836	else	1668	else
837	ERR ("'false' expected");	1669	ERR ("'false' expected");
838		1670
839	break;	1671	break;
…		…
848	ERR ("'null' expected");	1680	ERR ("'null' expected");
849		1681
850	break;	1682	break;
851		1683
852	default:	1684	default:
853	ERR ("malformed json string, neither array, object, number, string or atom");	1685	ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
854	break;	1686	break;
855	}	1687	}
856		1688
857	fail:	1689	fail:
858	return 0;	1690	return 0;
859	}	1691	}
860		1692
861	static SV *	1693	static SV *
862	decode_json (SV *string, UV flags)	1694	decode_json (SV *string, JSON *json, STRLEN *offset_return)
863	{	1695	{
		1696	dec_t dec;
864	SV *sv;	1697	SV *sv;
865		1698
866	if (flags & F_UTF8)	1699	/* work around bugs in 5.10 where manipulating magic values
		1700	* makes perl ignore the magic in subsequent accesses.
		1701	* also make a copy of non-PV values, to get them into a clean
		1702	* state (SvPV should do that, but it's buggy, see below).
		1703	*
		1704	* SvIsCOW_shared_hash works around a bug in perl (possibly 5.16),
		1705	* as reported by Reini Urban.
		1706	*/
		1707	/*SvGETMAGIC (string);*/
		1708	if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
		1709	string = sv_2mortal (newSVsv (string));
		1710
		1711	SvUPGRADE (string, SVt_PV);
		1712
		1713	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1714	* assertion with -DDEBUGGING, although SvCUR is documented to
		1715	* return the xpv_cur field which certainly exists after upgrading.
		1716	* according to nicholas clark, calling SvPOK fixes this.
		1717	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1718	* and hope for the best.
		1719	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1720	* assertion business is seriously broken, try yet another workaround
		1721	* for the broken -DDEBUGGING.
		1722	*/
		1723	{
		1724	#ifdef DEBUGGING
		1725	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1726	#else
		1727	STRLEN offset = SvCUR (string);
		1728	#endif
		1729
		1730	if (offset > json->max_size && json->max_size)
		1731	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1732	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1733	}
		1734
		1735	if (DECODE_WANTS_OCTETS (json))
867	sv_utf8_downgrade (string, 0);	1736	sv_utf8_downgrade (string, 0);
868	else	1737	else
869	sv_utf8_upgrade (string);	1738	sv_utf8_upgrade (string);
870		1739
871	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1740	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
872		1741
873	dec_t dec;	1742	dec.json = *json;
874	dec.flags = flags;
875	dec.cur = SvPVX (string);	1743	dec.cur = SvPVX (string);
876	dec.end = SvEND (string);	1744	dec.end = SvEND (string);
877	dec.err = 0;	1745	dec.err = 0;
		1746	dec.depth = 0;
878		1747
		1748	if (dec.json.cb_object || dec.json.cb_sk_object)
		1749	dec.json.flags |= F_HOOK;
		1750
		1751	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1752
		1753	decode_ws (&dec);
879	sv = decode_sv (&dec);	1754	sv = decode_sv (&dec);
880		1755
		1756	if (offset_return)
		1757	*offset_return = dec.cur - SvPVX (string);
		1758	else if (sv)
		1759	{
		1760	// check for trailing garbage
		1761	decode_ws (&dec);
		1762
		1763	if (*dec.cur)
		1764	{
		1765	dec.err = "garbage after JSON object";
		1766	SvREFCNT_dec (sv);
		1767	sv = 0;
		1768	}
		1769	}
		1770
881	if (!sv)	1771	if (!sv)
882	{	1772	{
883	IV offset = dec.flags & F_UTF8
884	? dec.cur - SvPVX (string)
885	: utf8_distance (dec.cur, SvPVX (string));
886	SV *uni = sv_newmortal ();	1773	SV *uni = sv_newmortal ();
887		1774
888	// horrible hack to silence warning inside pv_uni_display	1775	// horrible hack to silence warning inside pv_uni_display
889	COP cop = *PL_curcop;	1776	COP cop = *PL_curcop;
890	cop.cop_warnings = pWARN_NONE;	1777	cop.cop_warnings = pWARN_NONE;
…		…
892	SAVEVPTR (PL_curcop);	1779	SAVEVPTR (PL_curcop);
893	PL_curcop = &cop;	1780	PL_curcop = &cop;
894	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1781	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
895	LEAVE;	1782	LEAVE;
896		1783
897	croak ("%s, at character offset %d (%s)",	1784	croak ("%s, at character offset %d (before \"%s\")",
898	dec.err,	1785	dec.err,
899	(int)offset,	1786	(int)ptr_to_index (string, dec.cur),
900	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1787	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
901	}	1788	}
902		1789
903	sv = sv_2mortal (sv);	1790	sv = sv_2mortal (sv);
904		1791
905	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1792	if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv))
906	croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)");	1793	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
907		1794
908	return sv;	1795	return sv;
909	}	1796	}
910		1797
		1798	/////////////////////////////////////////////////////////////////////////////
		1799	// incremental parser
		1800
		1801	static void
		1802	incr_parse (JSON *self)
		1803	{
		1804	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1805
		1806	// the state machine here is a bit convoluted and could be simplified a lot
		1807	// but this would make it slower, so...
		1808
		1809	for (;;)
		1810	{
		1811	//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
		1812	switch (self->incr_mode)
		1813	{
		1814	// only used for initial whitespace skipping
		1815	case INCR_M_WS:
		1816	for (;;)
		1817	{
		1818	if (*p > 0x20)
		1819	{
		1820	if (*p == '#')
		1821	{
		1822	self->incr_mode = INCR_M_C0;
		1823	goto incr_m_c;
		1824	}
		1825	else
		1826	{
		1827	self->incr_mode = INCR_M_JSON;
		1828	goto incr_m_json;
		1829	}
		1830	}
		1831	else if (!*p)
		1832	goto interrupt;
		1833
		1834	++p;
		1835	}
		1836
		1837	// skip a single char inside a string (for \\-processing)
		1838	case INCR_M_BS:
		1839	if (!*p)
		1840	goto interrupt;
		1841
		1842	++p;
		1843	self->incr_mode = INCR_M_STR;
		1844	goto incr_m_str;
		1845
		1846	// inside #-style comments
		1847	case INCR_M_C0:
		1848	case INCR_M_C1:
		1849	incr_m_c:
		1850	for (;;)
		1851	{
		1852	if (*p == '\n')
		1853	{
		1854	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1855	break;
		1856	}
		1857	else if (!*p)
		1858	goto interrupt;
		1859
		1860	++p;
		1861	}
		1862
		1863	break;
		1864
		1865	// inside a string
		1866	case INCR_M_STR:
		1867	incr_m_str:
		1868	for (;;)
		1869	{
		1870	if (*p == '"')
		1871	{
		1872	++p;
		1873	self->incr_mode = INCR_M_JSON;
		1874
		1875	if (!self->incr_nest)
		1876	goto interrupt;
		1877
		1878	goto incr_m_json;
		1879	}
		1880	else if (*p == '\\')
		1881	{
		1882	++p; // "virtually" consumes character after \
		1883
		1884	if (!*p) // if at end of string we have to switch modes
		1885	{
		1886	self->incr_mode = INCR_M_BS;
		1887	goto interrupt;
		1888	}
		1889	}
		1890	else if (!*p)
		1891	goto interrupt;
		1892
		1893	++p;
		1894	}
		1895
		1896	// after initial ws, outside string
		1897	case INCR_M_JSON:
		1898	incr_m_json:
		1899	for (;;)
		1900	{
		1901	switch (*p++)
		1902	{
		1903	case 0:
		1904	--p;
		1905	goto interrupt;
		1906
		1907	case 0x09:
		1908	case 0x0a:
		1909	case 0x0d:
		1910	case 0x20:
		1911	if (!self->incr_nest)
		1912	{
		1913	--p; // do not eat the whitespace, let the next round do it
		1914	goto interrupt;
		1915	}
		1916	break;
		1917
		1918	case '"':
		1919	self->incr_mode = INCR_M_STR;
		1920	goto incr_m_str;
		1921
		1922	case '[':
		1923	case '{':
		1924	case '(':
		1925	if (++self->incr_nest > self->max_depth)
		1926	croak (ERR_NESTING_EXCEEDED);
		1927	break;
		1928
		1929	case ']':
		1930	case '}':
		1931	if (--self->incr_nest <= 0)
		1932	goto interrupt;
		1933	break;
		1934
		1935	case ')':
		1936	--self->incr_nest;
		1937	break;
		1938
		1939	case '#':
		1940	self->incr_mode = INCR_M_C1;
		1941	goto incr_m_c;
		1942	}
		1943	}
		1944	}
		1945
		1946	modechange:
		1947	;
		1948	}
		1949
		1950	interrupt:
		1951	self->incr_pos = p - SvPVX (self->incr_text);
		1952	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1953	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
		1954	}
		1955
		1956	/////////////////////////////////////////////////////////////////////////////
		1957	// XS interface functions
		1958
911	MODULE = JSON::XS PACKAGE = JSON::XS	1959	MODULE = JSON::XS PACKAGE = JSON::XS
912		1960
913	BOOT:	1961	BOOT:
914	{	1962	{
915	int i;	1963	int i;
916		1964
917	memset (decode_hexdigit, 0xff, 256);
918	for (i = 10; i--; )	1965	for (i = 0; i < 256; ++i)
919	decode_hexdigit ['0' + i] = i;	1966	decode_hexdigit [i] =
		1967	i >= '0' && i <= '9' ? i - '0'
		1968	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1969	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1970	: -1;
920		1971
921	for (i = 7; i--; )
922	{
923	decode_hexdigit ['a' + i] = 10 + i;
924	decode_hexdigit ['A' + i] = 10 + i;
925	}
926
927	json_stash = gv_stashpv ("JSON::XS", 1);	1972	json_stash = gv_stashpv ("JSON::XS" , 1);
		1973	bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
		1974	bool_true = get_bool ("Types::Serialiser::true");
		1975	bool_false = get_bool ("Types::Serialiser::false");
		1976
		1977	sv_json = newSVpv ("JSON", 0);
		1978	SvREADONLY_on (sv_json);
		1979
		1980	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
928	}	1981	}
929		1982
930	PROTOTYPES: DISABLE	1983	PROTOTYPES: DISABLE
931		1984
932	SV *new (char *dummy)	1985	void CLONE (...)
933	CODE:	1986	CODE:
934	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1987	json_stash = 0;
		1988	bool_stash = 0;
		1989
		1990	void new (char *klass)
		1991	PPCODE:
		1992	{
		1993	SV *pv = NEWSV (0, sizeof (JSON));
		1994	SvPOK_only (pv);
		1995	json_init ((JSON *)SvPVX (pv));
		1996	XPUSHs (sv_2mortal (sv_bless (
		1997	newRV_noinc (pv),
		1998	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1999	)));
		2000	}
		2001
		2002	void ascii (JSON *self, int enable = 1)
		2003	ALIAS:
		2004	ascii = F_ASCII
		2005	latin1 = F_LATIN1
		2006	utf8 = F_UTF8
		2007	indent = F_INDENT
		2008	canonical = F_CANONICAL
		2009	space_before = F_SPACE_BEFORE
		2010	space_after = F_SPACE_AFTER
		2011	pretty = F_PRETTY
		2012	allow_nonref = F_ALLOW_NONREF
		2013	shrink = F_SHRINK
		2014	allow_blessed = F_ALLOW_BLESSED
		2015	convert_blessed = F_CONV_BLESSED
		2016	relaxed = F_RELAXED
		2017	allow_unknown = F_ALLOW_UNKNOWN
		2018	allow_tags = F_ALLOW_TAGS
		2019	PPCODE:
		2020	{
		2021	if (enable)
		2022	self->flags |= ix;
		2023	else
		2024	self->flags &= ~ix;
		2025
		2026	XPUSHs (ST (0));
		2027	}
		2028
		2029	void get_ascii (JSON *self)
		2030	ALIAS:
		2031	get_ascii = F_ASCII
		2032	get_latin1 = F_LATIN1
		2033	get_utf8 = F_UTF8
		2034	get_indent = F_INDENT
		2035	get_canonical = F_CANONICAL
		2036	get_space_before = F_SPACE_BEFORE
		2037	get_space_after = F_SPACE_AFTER
		2038	get_allow_nonref = F_ALLOW_NONREF
		2039	get_shrink = F_SHRINK
		2040	get_allow_blessed = F_ALLOW_BLESSED
		2041	get_convert_blessed = F_CONV_BLESSED
		2042	get_relaxed = F_RELAXED
		2043	get_allow_unknown = F_ALLOW_UNKNOWN
		2044	get_allow_tags = F_ALLOW_TAGS
		2045	PPCODE:
		2046	XPUSHs (boolSV (self->flags & ix));
		2047
		2048	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		2049	PPCODE:
		2050	self->max_depth = max_depth;
		2051	XPUSHs (ST (0));
		2052
		2053	U32 get_max_depth (JSON *self)
		2054	CODE:
		2055	RETVAL = self->max_depth;
935	OUTPUT:	2056	OUTPUT:
936	RETVAL	2057	RETVAL
937		2058
938	SV *ascii (SV *self, int enable = 1)	2059	void max_size (JSON *self, U32 max_size = 0)
939	ALIAS:	2060	PPCODE:
940	ascii = F_ASCII	2061	self->max_size = max_size;
941	utf8 = F_UTF8	2062	XPUSHs (ST (0));
942	indent = F_INDENT	2063
943	canonical = F_CANONICAL	2064	int get_max_size (JSON *self)
944	space_before = F_SPACE_BEFORE
945	space_after = F_SPACE_AFTER
946	json_rpc = F_JSON_RPC
947	pretty = F_PRETTY
948	allow_nonref = F_ALLOW_NONREF
949	shrink = F_SHRINK
950	CODE:	2065	CODE:
951	{	2066	RETVAL = self->max_size;
952	UV *uv = SvJSON (self);
953	if (enable)
954	*uv |= ix;
955	else
956	*uv &= ~ix;
957
958	RETVAL = newSVsv (self);
959	}
960	OUTPUT:	2067	OUTPUT:
961	RETVAL	2068	RETVAL
962		2069
963	void encode (SV *self, SV *scalar)	2070	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
964	PPCODE:	2071	PPCODE:
965	XPUSHs (encode_json (scalar, *SvJSON (self)));	2072	{
		2073	SvREFCNT_dec (self->cb_object);
		2074	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
966		2075
967	void decode (SV *self, SV *jsonstr)	2076	XPUSHs (ST (0));
		2077	}
		2078
		2079	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
968	PPCODE:	2080	PPCODE:
969	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	2081	{
		2082	if (!self->cb_sk_object)
		2083	self->cb_sk_object = newHV ();
		2084
		2085	if (SvOK (cb))
		2086	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		2087	else
		2088	{
		2089	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		2090
		2091	if (!HvKEYS (self->cb_sk_object))
		2092	{
		2093	SvREFCNT_dec (self->cb_sk_object);
		2094	self->cb_sk_object = 0;
		2095	}
		2096	}
		2097
		2098	XPUSHs (ST (0));
		2099	}
		2100
		2101	void encode (JSON *self, SV *scalar)
		2102	PPCODE:
		2103	PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
		2104	XPUSHs (scalar);
		2105
		2106	void decode (JSON *self, SV *jsonstr)
		2107	PPCODE:
		2108	PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
		2109	XPUSHs (jsonstr);
		2110
		2111	void decode_prefix (JSON *self, SV *jsonstr)
		2112	PPCODE:
		2113	{
		2114	SV *sv;
		2115	STRLEN offset;
		2116	PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
		2117	EXTEND (SP, 2);
		2118	PUSHs (sv);
		2119	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset))));
		2120	}
		2121
		2122	void incr_parse (JSON *self, SV *jsonstr = 0)
		2123	PPCODE:
		2124	{
		2125	if (!self->incr_text)
		2126	self->incr_text = newSVpvn ("", 0);
		2127
		2128	/* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
		2129	if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
		2130	if (DECODE_WANTS_OCTETS (self))
		2131	{
		2132	if (self->incr_pos)
		2133	self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
		2134	(U8 *)SvPVX (self->incr_text) + self->incr_pos);
		2135
		2136	sv_utf8_downgrade (self->incr_text, 0);
		2137	}
		2138	else
		2139	{
		2140	sv_utf8_upgrade (self->incr_text);
		2141
		2142	if (self->incr_pos)
		2143	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		2144	- (U8 *)SvPVX (self->incr_text);
		2145	}
		2146
		2147	// append data, if any
		2148	if (jsonstr)
		2149	{
		2150	/* make sure both strings have same encoding */
		2151	if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
		2152	if (SvUTF8 (jsonstr))
		2153	sv_utf8_downgrade (jsonstr, 0);
		2154	else
		2155	sv_utf8_upgrade (jsonstr);
		2156
		2157	/* and then just blindly append */
		2158	{
		2159	STRLEN len;
		2160	const char *str = SvPV (jsonstr, len);
		2161	STRLEN cur = SvCUR (self->incr_text);
		2162
		2163	if (SvLEN (self->incr_text) <= cur + len)
		2164	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		2165
		2166	Move (str, SvEND (self->incr_text), len, char);
		2167	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		2168	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		2169	}
		2170	}
		2171
		2172	if (GIMME_V != G_VOID)
		2173	do
		2174	{
		2175	SV *sv;
		2176	STRLEN offset;
		2177
		2178	if (!INCR_DONE (self))
		2179	{
		2180	incr_parse (self);
		2181
		2182	if (self->incr_pos > self->max_size && self->max_size)
		2183	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		2184	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		2185
		2186	if (!INCR_DONE (self))
		2187	{
		2188	// as an optimisation, do not accumulate white space in the incr buffer
		2189	if (self->incr_mode == INCR_M_WS && self->incr_pos)
		2190	{
		2191	self->incr_pos = 0;
		2192	SvCUR_set (self->incr_text, 0);
		2193	}
		2194
		2195	break;
		2196	}
		2197	}
		2198
		2199	PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
		2200	XPUSHs (sv);
		2201
		2202	self->incr_pos -= offset;
		2203	self->incr_nest = 0;
		2204	self->incr_mode = 0;
		2205
		2206	sv_chop (self->incr_text, SvPVX (self->incr_text) + offset);
		2207	}
		2208	while (GIMME_V == G_ARRAY);
		2209	}
		2210
		2211	SV *incr_text (JSON *self)
		2212	ATTRS: lvalue
		2213	CODE:
		2214	{
		2215	if (self->incr_pos)
		2216	croak ("incr_text can not be called when the incremental parser already started parsing");
		2217
		2218	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		2219	}
		2220	OUTPUT:
		2221	RETVAL
		2222
		2223	void incr_skip (JSON *self)
		2224	CODE:
		2225	{
		2226	if (self->incr_pos)
		2227	{
		2228	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2229	self->incr_pos = 0;
		2230	self->incr_nest = 0;
		2231	self->incr_mode = 0;
		2232	}
		2233	}
		2234
		2235	void incr_reset (JSON *self)
		2236	CODE:
		2237	{
		2238	SvREFCNT_dec (self->incr_text);
		2239	self->incr_text = 0;
		2240	self->incr_pos = 0;
		2241	self->incr_nest = 0;
		2242	self->incr_mode = 0;
		2243	}
		2244
		2245	void DESTROY (JSON *self)
		2246	CODE:
		2247	SvREFCNT_dec (self->cb_sk_object);
		2248	SvREFCNT_dec (self->cb_object);
		2249	SvREFCNT_dec (self->incr_text);
970		2250
971	PROTOTYPES: ENABLE	2251	PROTOTYPES: ENABLE
972		2252
973	void to_json (SV *scalar)	2253	void encode_json (SV *scalar)
974	PPCODE:	2254	PPCODE:
975	XPUSHs (encode_json (scalar, F_UTF8));	2255	{
		2256	JSON json;
		2257	json_init (&json);
		2258	json.flags |= F_UTF8;
		2259	PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
		2260	XPUSHs (scalar);
		2261	}
976		2262
977	void from_json (SV *jsonstr)	2263	void decode_json (SV *jsonstr)
978	PPCODE:	2264	PPCODE:
979	XPUSHs (decode_json (jsonstr, F_UTF8));	2265	{
		2266	JSON json;
		2267	json_init (&json);
		2268	json.flags |= F_UTF8;
		2269	PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
		2270	XPUSHs (jsonstr);
		2271	}
980		2272

Diff Legend

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