ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.7 by root, Fri Mar 23 15:57:18 2007 UTC vs.
Revision 1.125 by root, Sun Feb 21 15:37:53 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	#else
		85	# define JSON_SLOW 0
		86	# define JSON_STASH json_stash
		87	#endif
		88
		89	// the amount of HEs to allocate on the stack, when sorting keys
		90	#define STACK_HES 64
		91
		92	static HV *json_stash, *types_boolean_stash; // JSON::XS::
		93	static SV *types_true, *types_false, *sv_json;
		94
		95	enum {
		96	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		97	INCR_M_STR, // inside string
		98	INCR_M_BS, // inside backslash
		99	INCR_M_C0, // inside comment in initial whitespace sequence
		100	INCR_M_C1, // inside comment in other places
		101	INCR_M_JSON // outside anything, count nesting
		102	};
		103
		104	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		105
		106	typedef struct {
		107	U32 flags;
		108	U32 max_depth;
		109	STRLEN max_size;
		110
		111	SV *cb_object;
		112	HV *cb_sk_object;
		113
		114	// for the incremental parser
		115	SV *incr_text; // the source text so far
		116	STRLEN incr_pos; // the current offset into the text
		117	int incr_nest; // {[]}-nesting level
		118	unsigned char incr_mode;
		119	} JSON;
		120
		121	INLINE void
		122	json_init (JSON *json)
		123	{
		124	Zero (json, 1, JSON);
		125	json->max_depth = 512;
		126	}
		127
		128	/////////////////////////////////////////////////////////////////////////////
		129	// utility functions
		130
		131	INLINE SV *
		132	get_bool (const char *name)
		133	{
		134	SV *sv = get_sv (name, 1);
		135
		136	SvREADONLY_on (sv);
		137	SvREADONLY_on (SvRV (sv));
		138
		139	return sv;
		140	}
		141
		142	INLINE void
		143	shrink (SV *sv)
		144	{
		145	sv_utf8_downgrade (sv, 1);
		146
		147	if (SvLEN (sv) > SvCUR (sv) + 1)
		148	{
		149	#ifdef SvPV_shrink_to_cur
		150	SvPV_shrink_to_cur (sv);
		151	#elif defined (SvPV_renew)
		152	SvPV_renew (sv, SvCUR (sv) + 1);
		153	#endif
		154	}
		155	}
		156
		157	// decode an utf-8 character and return it, or (UV)-1 in
		158	// case of an error.
		159	// we special-case "safe" characters from U+80 .. U+7FF,
		160	// but use the very good perl function to parse anything else.
		161	// note that we never call this function for a ascii codepoints
		162	INLINE UV
		163	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		164	{
		165	if (expect_true (len >= 2
		166	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
		167	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
		168	{
		169	*clen = 2;
		170	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		171	}
		172	else
		173	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		174	}
		175
		176	// likewise for encoding, also never called for ascii codepoints
		177	// this function takes advantage of this fact, although current gccs
		178	// seem to optimise the check for >= 0x80 away anyways
		179	INLINE unsigned char *
		180	encode_utf8 (unsigned char *s, UV ch)
		181	{
		182	if (expect_false (ch < 0x000080))
		183	*s++ = ch;
		184	else if (expect_true (ch < 0x000800))
		185	*s++ = 0xc0 | ( ch >> 6),
		186	*s++ = 0x80 | ( ch & 0x3f);
		187	else if ( ch < 0x010000)
		188	*s++ = 0xe0 | ( ch >> 12),
		189	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		190	*s++ = 0x80 | ( ch & 0x3f);
		191	else if ( ch < 0x110000)
		192	*s++ = 0xf0 | ( ch >> 18),
		193	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		194	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		195	*s++ = 0x80 | ( ch & 0x3f);
		196
		197	return s;
		198	}
		199
		200	// convert offset pointer to character index, sv must be string
		201	static STRLEN
		202	ptr_to_index (SV *sv, char *offset)
		203	{
		204	return SvUTF8 (sv)
		205	? utf8_distance (offset, SvPVX (sv))
		206	: offset - SvPVX (sv);
		207	}
		208
		209	/////////////////////////////////////////////////////////////////////////////
		210	// fp hell
		211
		212	// scan a group of digits, and a trailing exponent
		213	static void
		214	json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
		215	{
		216	UV uaccum = 0;
		217	int eaccum = 0;
		218
		219	// if we recurse too deep, skip all remaining digits
		220	// to avoid a stack overflow attack
		221	if (expect_false (--maxdepth <= 0))
		222	while (((U8)*s - '0') < 10)
		223	++s;
		224
		225	for (;;)
		226	{
		227	U8 dig = (U8)*s - '0';
		228
		229	if (expect_false (dig >= 10))
		230	{
		231	if (dig == (U8)((U8)'.' - (U8)'0'))
		232	{
		233	++s;
		234	json_atof_scan1 (s, accum, expo, 1, maxdepth);
		235	}
		236	else if ((dig | ' ') == 'e' - '0')
		237	{
		238	int exp2 = 0;
		239	int neg = 0;
		240
		241	++s;
		242
		243	if (*s == '-')
		244	{
		245	++s;
		246	neg = 1;
		247	}
		248	else if (*s == '+')
		249	++s;
		250
		251	while ((dig = (U8)*s - '0') < 10)
		252	exp2 = exp2 * 10 + *s++ - '0';
		253
		254	*expo += neg ? -exp2 : exp2;
		255	}
		256
		257	break;
		258	}
		259
		260	++s;
		261
		262	uaccum = uaccum * 10 + dig;
		263	++eaccum;
		264
		265	// if we have too many digits, then recurse for more
		266	// we actually do this for rather few digits
		267	if (uaccum >= (UV_MAX - 9) / 10)
		268	{
		269	if (postdp) *expo -= eaccum;
		270	json_atof_scan1 (s, accum, expo, postdp, maxdepth);
		271	if (postdp) *expo += eaccum;
		272
		273	break;
		274	}
		275	}
		276
		277	// this relies greatly on the quality of the pow ()
		278	// implementation of the platform, but a good
		279	// implementation is hard to beat.
		280	// (IEEE 754 conformant ones are required to be exact)
		281	if (postdp) *expo -= eaccum;
		282	*accum += uaccum * Perl_pow (10., *expo);
		283	*expo += eaccum;
		284	}
		285
		286	static NV
		287	json_atof (const char *s)
		288	{
		289	NV accum = 0.;
		290	int expo = 0;
		291	int neg = 0;
		292
		293	if (*s == '-')
		294	{
		295	++s;
		296	neg = 1;
		297	}
		298
		299	// a recursion depth of ten gives us >>500 bits
		300	json_atof_scan1 (s, &accum, &expo, 0, 10);
		301
		302	return neg ? -accum : accum;
		303	}
		304	/////////////////////////////////////////////////////////////////////////////
		305	// encoder
28		306
29	// structure used for encoding JSON	307	// structure used for encoding JSON
30	typedef struct	308	typedef struct
31	{	309	{
32	char *cur;	310	char *cur; // SvPVX (sv) + current output position
33	STRLEN len; // SvLEN (sv)
34	char *end; // SvEND (sv)	311	char *end; // SvEND (sv)
35	SV *sv;	312	SV *sv; // result scalar
36	UV flags;	313	JSON json;
37	int max_recurse;	314	U32 indent; // indentation level
38	int indent;	315	UV limit; // escape character values >= this value when encoding
39	} enc_t;	316	} enc_t;
40		317
41	// structure used for decoding JSON	318	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)	319	need (enc_t *enc, STRLEN len)
72	{	320	{
73	if (enc->cur + len >= enc->end)	321	if (expect_false (enc->cur + len >= enc->end))
74	{	322	{
75	STRLEN cur = enc->cur - SvPVX (enc->sv);	323	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
76	SvGROW (enc->sv, cur + len + 1);	324	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
77	enc->cur = SvPVX (enc->sv) + cur;	325	enc->cur = SvPVX (enc->sv) + cur;
78	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	326	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
79	}	327	}
80	}	328	}
81		329
82	static void	330	INLINE void
83	encode_ch (enc_t *enc, char ch)	331	encode_ch (enc_t *enc, char ch)
84	{	332	{
85	need (enc, 1);	333	need (enc, 1);
86	*enc->cur++ = ch;	334	*enc->cur++ = ch;
87	}	335	}
…		…
95		343
96	while (str < end)	344	while (str < end)
97	{	345	{
98	unsigned char ch = *(unsigned char *)str;	346	unsigned char ch = *(unsigned char *)str;
99		347
100	if (ch >= 0x20 && ch < 0x80) // most common case	348	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
101	{	349	{
102	if (ch == '"') // but with slow exceptions	350	if (expect_false (ch == '"')) // but with slow exceptions
103	{	351	{
104	need (enc, len += 1);	352	need (enc, len += 1);
105	*enc->cur++ = '\\';	353	*enc->cur++ = '\\';
106	*enc->cur++ = '"';	354	*enc->cur++ = '"';
107	}	355	}
108	else if (ch == '\\')	356	else if (expect_false (ch == '\\'))
109	{	357	{
110	need (enc, len += 1);	358	need (enc, len += 1);
111	*enc->cur++ = '\\';	359	*enc->cur++ = '\\';
112	*enc->cur++ = '\\';	360	*enc->cur++ = '\\';
113	}	361	}
…		…
131	STRLEN clen;	379	STRLEN clen;
132	UV uch;	380	UV uch;
133		381
134	if (is_utf8)	382	if (is_utf8)
135	{	383	{
136	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	384	uch = decode_utf8 (str, end - str, &clen);
137	if (clen == (STRLEN)-1)	385	if (clen == (STRLEN)-1)
138	croak ("malformed UTF-8 character in string, cannot convert to JSON");	386	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
139	}	387	}
140	else	388	else
141	{	389	{
142	uch = ch;	390	uch = ch;
143	clen = 1;	391	clen = 1;
144	}	392	}
145		393
146	if (uch < 0x80 || enc->flags & F_ASCII)	394	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
147	{	395	{
148	if (uch > 0xFFFFUL)	396	if (uch >= 0x10000UL)
149	{	397	{
		398	if (uch >= 0x110000UL)
		399	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		400
150	need (enc, len += 11);	401	need (enc, len += 11);
151	sprintf (enc->cur, "\\u%04x\\u%04x",	402	sprintf (enc->cur, "\\u%04x\\u%04x",
152	(uch - 0x10000) / 0x400 + 0xD800,	403	(int)((uch - 0x10000) / 0x400 + 0xD800),
153	(uch - 0x10000) % 0x400 + 0xDC00);	404	(int)((uch - 0x10000) % 0x400 + 0xDC00));
154	enc->cur += 12;	405	enc->cur += 12;
155	}	406	}
156	else	407	else
157	{	408	{
158	static char hexdigit [16] = "0123456789abcdef";
159	need (enc, len += 5);	409	need (enc, len += 5);
160	*enc->cur++ = '\\';	410	*enc->cur++ = '\\';
161	*enc->cur++ = 'u';	411	*enc->cur++ = 'u';
162	*enc->cur++ = hexdigit [ uch >> 12 ];	412	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
163	*enc->cur++ = hexdigit [(uch >> 8) & 15];	413	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
164	*enc->cur++ = hexdigit [(uch >> 4) & 15];	414	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
165	*enc->cur++ = hexdigit [(uch >> 0) & 15];	415	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
166	}	416	}
167		417
		418	str += clen;
		419	}
		420	else if (enc->json.flags & F_LATIN1)
		421	{
		422	*enc->cur++ = uch;
168	str += clen;	423	str += clen;
169	}	424	}
170	else if (is_utf8)	425	else if (is_utf8)
171	{	426	{
172	need (enc, len += clen);	427	need (enc, len += clen);
…		…
176	}	431	}
177	while (--clen);	432	while (--clen);
178	}	433	}
179	else	434	else
180	{	435	{
181	need (enc, len += 10); // never more than 11 bytes needed	436	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
182	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	437	enc->cur = encode_utf8 (enc->cur, uch);
183	++str;	438	++str;
184	}	439	}
185	}	440	}
186	}	441	}
187	}	442	}
188		443
189	--len;	444	--len;
190	}	445	}
191	}	446	}
192		447
193	#define INDENT SB \	448	INLINE void
		449	encode_indent (enc_t *enc)
		450	{
194	if (enc->flags & F_INDENT) \	451	if (enc->json.flags & F_INDENT)
195	{ \	452	{
196	int i_; \	453	int spaces = enc->indent * INDENT_STEP;
197	need (enc, enc->indent); \	454
198	for (i_ = enc->indent * 3; i_--; )\	455	need (enc, spaces);
		456	memset (enc->cur, ' ', spaces);
		457	enc->cur += spaces;
		458	}
		459	}
		460
		461	INLINE void
		462	encode_space (enc_t *enc)
		463	{
		464	need (enc, 1);
		465	encode_ch (enc, ' ');
		466	}
		467
		468	INLINE void
		469	encode_nl (enc_t *enc)
		470	{
		471	if (enc->json.flags & F_INDENT)
		472	{
		473	need (enc, 1);
199	encode_ch (enc, ' '); \	474	encode_ch (enc, '\n');
200	} \	475	}
201	SE	476	}
202		477
203	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	478	INLINE void
204	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	479	encode_comma (enc_t *enc)
205	#define COMMA SB \	480	{
206	encode_ch (enc, ','); \	481	encode_ch (enc, ',');
		482
207	if (enc->flags & F_INDENT) \	483	if (enc->json.flags & F_INDENT)
208	NL; \	484	encode_nl (enc);
209	else if (enc->flags & F_SPACE_AFTER) \	485	else if (enc->json.flags & F_SPACE_AFTER)
210	SPACE; \	486	encode_space (enc);
211	SE	487	}
212		488
213	static void encode_sv (enc_t *enc, SV *sv);	489	static void encode_sv (enc_t *enc, SV *sv);
214		490
215	static void	491	static void
216	encode_av (enc_t *enc, AV *av)	492	encode_av (enc_t *enc, AV *av)
217	{	493	{
218	int i, len = av_len (av);	494	int i, len = av_len (av);
219		495
		496	if (enc->indent >= enc->json.max_depth)
		497	croak (ERR_NESTING_EXCEEDED);
		498
220	encode_ch (enc, '['); NL;	499	encode_ch (enc, '[');
221	++enc->indent;
222		500
		501	if (len >= 0)
		502	{
		503	encode_nl (enc); ++enc->indent;
		504
223	for (i = 0; i <= len; ++i)	505	for (i = 0; i <= len; ++i)
224	{	506	{
225	INDENT;	507	SV **svp = av_fetch (av, i, 0);
226	encode_sv (enc, *av_fetch (av, i, 0));
227		508
		509	encode_indent (enc);
		510
		511	if (svp)
		512	encode_sv (enc, *svp);
		513	else
		514	encode_str (enc, "null", 4, 0);
		515
228	if (i < len)	516	if (i < len)
229	COMMA;	517	encode_comma (enc);
230	}	518	}
231		519
232	NL;	520	encode_nl (enc); --enc->indent; encode_indent (enc);
		521	}
233		522
234	--enc->indent;
235	INDENT; encode_ch (enc, ']');	523	encode_ch (enc, ']');
236	}	524	}
237		525
238	static void	526	static void
239	encode_he (enc_t *enc, HE *he)	527	encode_hk (enc_t *enc, HE *he)
240	{	528	{
241	encode_ch (enc, '"');	529	encode_ch (enc, '"');
242		530
243	if (HeKLEN (he) == HEf_SVKEY)	531	if (HeKLEN (he) == HEf_SVKEY)
244	{	532	{
245	SV *sv = HeSVKEY (he);	533	SV *sv = HeSVKEY (he);
246	STRLEN len;	534	STRLEN len;
247	char *str;	535	char *str;
248		536
249	SvGETMAGIC (sv);	537	SvGETMAGIC (sv);
250	str = SvPV (sv, len);	538	str = SvPV (sv, len);
251		539
252	encode_str (enc, str, len, SvUTF8 (sv));	540	encode_str (enc, str, len, SvUTF8 (sv));
253	}	541	}
254	else	542	else
255	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	543	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
256		544
257	encode_ch (enc, '"');	545	encode_ch (enc, '"');
258		546
259	if (enc->flags & F_SPACE_BEFORE) SPACE;	547	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
260	encode_ch (enc, ':');	548	encode_ch (enc, ':');
261	if (enc->flags & F_SPACE_AFTER ) SPACE;	549	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
262	encode_sv (enc, HeVAL (he));
263	}	550	}
264		551
265	// compare hash entries, used when all keys are bytestrings	552	// compare hash entries, used when all keys are bytestrings
266	static int	553	static int
267	he_cmp_fast (const void *a_, const void *b_)	554	he_cmp_fast (const void *a_, const void *b_)
…		…
272	HE *b = *(HE **)b_;	559	HE *b = *(HE **)b_;
273		560
274	STRLEN la = HeKLEN (a);	561	STRLEN la = HeKLEN (a);
275	STRLEN lb = HeKLEN (b);	562	STRLEN lb = HeKLEN (b);
276		563
277	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	564	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
278	cmp = la < lb ? -1 : la == lb ? 0 : 1;	565	cmp = lb - la;
279		566
280	return cmp;	567	return cmp;
281	}	568	}
282		569
283	// compare hash entries, used when some keys are sv's or utf-x	570	// compare hash entries, used when some keys are sv's or utf-x
284	static int	571	static int
285	he_cmp_slow (const void *a, const void *b)	572	he_cmp_slow (const void *a, const void *b)
286	{	573	{
287	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	574	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
288	}	575	}
289		576
290	static void	577	static void
291	encode_hv (enc_t *enc, HV *hv)	578	encode_hv (enc_t *enc, HV *hv)
292	{	579	{
293	int count, i;	580	HE *he;
294		581
295	encode_ch (enc, '{'); NL; ++enc->indent;	582	if (enc->indent >= enc->json.max_depth)
		583	croak (ERR_NESTING_EXCEEDED);
296		584
297	if ((count = hv_iterinit (hv)))	585	encode_ch (enc, '{');
298	{	586
299	// for canonical output we have to sort by keys first	587	// for canonical output we have to sort by keys first
300	// actually, this is mostly due to the stupid so-called	588	// actually, this is mostly due to the stupid so-called
301	// security workaround added somewhere in 5.8.x.	589	// security workaround added somewhere in 5.8.x
302	// that randomises hash orderings	590	// that randomises hash orderings
303	if (enc->flags & F_CANONICAL)	591	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
		592	{
		593	int count = hv_iterinit (hv);
		594
		595	if (SvMAGICAL (hv))
304	{	596	{
305	HE *he, *hes [count];	597	// need to count by iterating. could improve by dynamically building the vector below
		598	// but I don't care for the speed of this special case.
		599	// note also that we will run into undefined behaviour when the two iterations
		600	// do not result in the same count, something I might care for in some later release.
		601
		602	count = 0;
		603	while (hv_iternext (hv))
		604	++count;
		605
		606	hv_iterinit (hv);
		607	}
		608
		609	if (count)
		610	{
306	int fast = 1;	611	int i, fast = 1;
		612	HE *hes_stack [STACK_HES];
		613	HE **hes = hes_stack;
		614
		615	// allocate larger arrays on the heap
		616	if (count > STACK_HES)
		617	{
		618	SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
		619	hes = (HE **)SvPVX (sv);
		620	}
307		621
308	i = 0;	622	i = 0;
309	while ((he = hv_iternext (hv)))	623	while ((he = hv_iternext (hv)))
310	{	624	{
311	hes [i++] = he;	625	hes [i++] = he;
…		…
317		631
318	if (fast)	632	if (fast)
319	qsort (hes, count, sizeof (HE *), he_cmp_fast);	633	qsort (hes, count, sizeof (HE *), he_cmp_fast);
320	else	634	else
321	{	635	{
322	// hack to disable "use bytes"	636	// hack to forcefully disable "use bytes"
323	COP *oldcop = PL_curcop, cop;	637	COP cop = *PL_curcop;
324	cop.op_private = 0;	638	cop.op_private = 0;
		639
		640	ENTER;
		641	SAVETMPS;
		642
		643	SAVEVPTR (PL_curcop);
325	PL_curcop = &cop;	644	PL_curcop = &cop;
326		645
327	SAVETMPS;
328	qsort (hes, count, sizeof (HE *), he_cmp_slow);	646	qsort (hes, count, sizeof (HE *), he_cmp_slow);
		647
329	FREETMPS;	648	FREETMPS;
330		649	LEAVE;
331	PL_curcop = oldcop;
332	}	650	}
333		651
334	for (i = 0; i < count; ++i)	652	encode_nl (enc); ++enc->indent;
		653
		654	while (count--)
335	{	655	{
336	INDENT;	656	encode_indent (enc);
		657	he = hes [count];
337	encode_he (enc, hes [i]);	658	encode_hk (enc, he);
		659	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
338		660
339	if (i < count - 1)	661	if (count)
340	COMMA;	662	encode_comma (enc);
341	}	663	}
342		664
343	NL;	665	encode_nl (enc); --enc->indent; encode_indent (enc);
344	}	666	}
		667	}
		668	else
		669	{
		670	if (hv_iterinit (hv) || SvMAGICAL (hv))
		671	if ((he = hv_iternext (hv)))
		672	{
		673	encode_nl (enc); ++enc->indent;
		674
		675	for (;;)
		676	{
		677	encode_indent (enc);
		678	encode_hk (enc, he);
		679	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		680
		681	if (!(he = hv_iternext (hv)))
		682	break;
		683
		684	encode_comma (enc);
		685	}
		686
		687	encode_nl (enc); --enc->indent; encode_indent (enc);
		688	}
		689	}
		690
		691	encode_ch (enc, '}');
		692	}
		693
		694	// encode objects, arrays and special \0=false and \1=true values.
		695	static void
		696	encode_rv (enc_t *enc, SV *sv)
		697	{
		698	svtype svt;
		699	GV *method;
		700
		701	SvGETMAGIC (sv);
		702	svt = SvTYPE (sv);
		703
		704	if (expect_false (SvOBJECT (sv)))
		705	{
		706	HV *boolean_stash = !JSON_SLOW || types_boolean_stash
		707	? types_boolean_stash
		708	: gv_stashpv ("Types::Serialiser::Boolean", 1);
		709	HV *stash = SvSTASH (sv);
		710
		711	if (stash == boolean_stash)
		712	{
		713	if (SvIV (sv))
		714	encode_str (enc, "true", 4, 0);
		715	else
		716	encode_str (enc, "false", 5, 0);
		717	}
		718	else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
		719	{
		720	int count;
		721	dSP;
		722
		723	ENTER; SAVETMPS;
		724	SAVESTACK_POS ();
		725	PUSHMARK (SP);
		726	EXTEND (SP, 2);
		727	// we re-bless the reference to get overload and other niceties right
		728	PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		729	PUSHs (sv_json);
		730
		731	PUTBACK;
		732	count = call_sv ((SV *)GvCV (method), G_ARRAY);
		733	SPAGAIN;
		734
		735	// catch this surprisingly common error
		736	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		737	croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		738
		739	encode_ch (enc, '(');
		740	encode_ch (enc, '"');
		741	encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
		742	encode_ch (enc, '"');
		743	encode_ch (enc, ')');
		744	encode_ch (enc, '[');
		745
		746	while (count)
		747	{
		748	encode_sv (enc, SP[1 - count--]);
		749
		750	if (count)
		751	encode_ch (enc, ',');
		752	}
		753
		754	encode_ch (enc, ']');
		755
		756	FREETMPS; LEAVE;
		757	}
		758	else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
		759	{
		760	dSP;
		761
		762	ENTER; SAVETMPS;
		763	PUSHMARK (SP);
		764	// we re-bless the reference to get overload and other niceties right
		765	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		766
		767	// calling with G_SCALAR ensures that we always get a 1 return value
		768	PUTBACK;
		769	call_sv ((SV *)GvCV (method), G_SCALAR);
		770	SPAGAIN;
		771
		772	// catch this surprisingly common error
		773	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		774	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		775
		776	sv = POPs;
		777	PUTBACK;
		778
		779	encode_sv (enc, sv);
		780
		781	FREETMPS; LEAVE;
		782	}
		783	else if (enc->json.flags & F_ALLOW_BLESSED)
		784	encode_str (enc, "null", 4, 0);
345	else	785	else
346	{	786	croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
347	SV *sv;	787	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
348	HE *he = hv_iternext (hv);	788	}
349		789	else if (svt == SVt_PVHV)
350	for (;;)	790	encode_hv (enc, (HV *)sv);
351	{	791	else if (svt == SVt_PVAV)
352	INDENT;	792	encode_av (enc, (AV *)sv);
353	encode_he (enc, he);	793	else if (svt < SVt_PVAV)
354
355	if (!(he = hv_iternext (hv)))
356	break;
357
358	COMMA;
359	}
360
361	NL;
362	}
363	}	794	{
		795	STRLEN len = 0;
		796	char *pv = svt ? SvPV (sv, len) : 0;
364		797
365	--enc->indent; INDENT; encode_ch (enc, '}');	798	if (len == 1 && *pv == '1')
		799	encode_str (enc, "true", 4, 0);
		800	else if (len == 1 && *pv == '0')
		801	encode_str (enc, "false", 5, 0);
		802	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		803	encode_str (enc, "null", 4, 0);
		804	else
		805	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		806	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		807	}
		808	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		809	encode_str (enc, "null", 4, 0);
		810	else
		811	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		812	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
366	}	813	}
367		814
368	static void	815	static void
369	encode_sv (enc_t *enc, SV *sv)	816	encode_sv (enc_t *enc, SV *sv)
370	{	817	{
…		…
378	encode_str (enc, str, len, SvUTF8 (sv));	825	encode_str (enc, str, len, SvUTF8 (sv));
379	encode_ch (enc, '"');	826	encode_ch (enc, '"');
380	}	827	}
381	else if (SvNOKp (sv))	828	else if (SvNOKp (sv))
382	{	829	{
		830	// trust that perl will do the right thing w.r.t. JSON syntax.
383	need (enc, NV_DIG + 32);	831	need (enc, NV_DIG + 32);
384	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	832	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
385	enc->cur += strlen (enc->cur);	833	enc->cur += strlen (enc->cur);
386	}	834	}
387	else if (SvIOKp (sv))	835	else if (SvIOKp (sv))
388	{	836	{
		837	// we assume we can always read an IV as a UV and vice versa
		838	// we assume two's complement
		839	// we assume no aliasing issues in the union
		840	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		841	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		842	{
		843	// optimise the "small number case"
		844	// code will likely be branchless and use only a single multiplication
		845	// works for numbers up to 59074
		846	I32 i = SvIVX (sv);
		847	U32 u;
		848	char digit, nz = 0;
		849
389	need (enc, 64);	850	need (enc, 6);
		851
		852	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		853	u = i < 0 ? -i : i;
		854
		855	// convert to 4.28 fixed-point representation
		856	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		857
		858	// now output digit by digit, each time masking out the integer part
		859	// and multiplying by 5 while moving the decimal point one to the right,
		860	// resulting in a net multiplication by 10.
		861	// we always write the digit to memory but conditionally increment
		862	// the pointer, to enable the use of conditional move instructions.
		863	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		864	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		865	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		866	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		867	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		868	}
		869	else
		870	{
		871	// large integer, use the (rather slow) snprintf way.
		872	need (enc, IVUV_MAXCHARS);
390	enc->cur +=	873	enc->cur +=
391	SvIsUV(sv)	874	SvIsUV(sv)
392	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	875	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
393	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	876	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		877	}
394	}	878	}
395	else if (SvROK (sv))	879	else if (SvROK (sv))
396	{	880	encode_rv (enc, SvRV (sv));
397	if (!--enc->max_recurse)	881	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
398	croak ("data structure too deep (hit recursion limit)");
399
400	sv = SvRV (sv);
401
402	switch (SvTYPE (sv))
403	{
404	case SVt_PVAV: encode_av (enc, (AV *)sv); break;
405	case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
406
407	default:
408	croak ("JSON can only represent references to arrays or hashes");
409	}
410	}
411	else if (!SvOK (sv))
412	encode_str (enc, "null", 4, 0);	882	encode_str (enc, "null", 4, 0);
413	else	883	else
414	croak ("encountered perl type that JSON cannot handle");	884	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
		885	SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
		886	}
		887
		888	static int
		889	json_scalar (SV *scalar)
		890	{
		891	return 0;//D
		892	if (!SvROK (scalar))
		893	return 1;
415	}	894	}
416		895
417	static SV *	896	static SV *
418	encode_json (SV *scalar, UV flags)	897	encode_json (SV *scalar, JSON *json)
419	{	898	{
420	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
421	croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");
422
423	enc_t enc;	899	enc_t enc;
424	enc.flags = flags;	900
		901	if (!(json->flags & F_ALLOW_NONREF) && json_scalar (scalar))
		902	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
		903
		904	enc.json = *json;
425	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	905	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
426	enc.cur = SvPVX (enc.sv);	906	enc.cur = SvPVX (enc.sv);
427	enc.end = SvEND (enc.sv);	907	enc.end = SvEND (enc.sv);
428	enc.max_recurse = 0;
429	enc.indent = 0;	908	enc.indent = 0;
		909	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		910	: enc.json.flags & F_LATIN1 ? 0x000100UL
		911	: 0x110000UL;
430		912
431	SvPOK_only (enc.sv);	913	SvPOK_only (enc.sv);
432	encode_sv (&enc, scalar);	914	encode_sv (&enc, scalar);
		915	encode_nl (&enc);
433		916
		917	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		918	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		919
434	if (!(flags & (F_ASCII | F_UTF8)))	920	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
435	SvUTF8_on (enc.sv);	921	SvUTF8_on (enc.sv);
436		922
437	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
438
439	if (enc.flags & F_SHRINK)	923	if (enc.json.flags & F_SHRINK)
440	shrink (enc.sv);	924	shrink (enc.sv);
441		925
442	return enc.sv;	926	return enc.sv;
443	}	927	}
444		928
445	/////////////////////////////////////////////////////////////////////////////	929	/////////////////////////////////////////////////////////////////////////////
		930	// decoder
446		931
447	#define WS \	932	// structure used for decoding JSON
		933	typedef struct
		934	{
		935	char *cur; // current parser pointer
		936	char *end; // end of input string
		937	const char *err; // parse error, if != 0
		938	JSON json;
		939	U32 depth; // recursion depth
		940	U32 maxdepth; // recursion depth limit
		941	} dec_t;
		942
		943	INLINE void
		944	decode_comment (dec_t *dec)
		945	{
		946	// only '#'-style comments allowed a.t.m.
		947
		948	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		949	++dec->cur;
		950	}
		951
		952	INLINE void
		953	decode_ws (dec_t *dec)
		954	{
448	for (;;) \	955	for (;;)
449	{ \	956	{
450	char ch = *dec->cur; \	957	char ch = *dec->cur;
		958
451	if (ch > 0x20 \	959	if (ch > 0x20)
		960	{
		961	if (expect_false (ch == '#'))
		962	{
		963	if (dec->json.flags & F_RELAXED)
		964	decode_comment (dec);
		965	else
		966	break;
		967	}
		968	else
		969	break;
		970	}
452	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	971	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
453	break; \	972	break; // parse error, but let higher level handle it, gives better error messages
		973
454	++dec->cur; \	974	++dec->cur;
455	}	975	}
		976	}
456		977
457	#define ERR(reason) SB dec->err = reason; goto fail; SE	978	#define ERR(reason) SB dec->err = reason; goto fail; SE
		979
458	#define EXPECT_CH(ch) SB \	980	#define EXPECT_CH(ch) SB \
459	if (*dec->cur != ch) \	981	if (*dec->cur != ch) \
460	ERR (# ch " expected"); \	982	ERR (# ch " expected"); \
461	++dec->cur; \	983	++dec->cur; \
462	SE	984	SE
463		985
		986	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
		987	#define DEC_DEC_DEPTH --dec->depth
		988
464	static SV *decode_sv (dec_t *dec);	989	static SV *decode_sv (dec_t *dec);
465		990
466	static signed char decode_hexdigit[256];	991	static signed char decode_hexdigit[256];
467		992
468	static UV	993	static UV
469	decode_4hex (dec_t *dec)	994	decode_4hex (dec_t *dec)
470	{	995	{
471	signed char d1, d2, d3, d4;	996	signed char d1, d2, d3, d4;
		997	unsigned char *cur = (unsigned char *)dec->cur;
472		998
473	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	999	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
474	if (d1 < 0) ERR ("four hexadecimal digits expected");	1000	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
475	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	1001	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
476	if (d2 < 0) ERR ("four hexadecimal digits expected");	1002	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
477	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
478	if (d3 < 0) ERR ("four hexadecimal digits expected");
479	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
480	if (d4 < 0) ERR ("four hexadecimal digits expected");
481		1003
482	dec->cur += 4;	1004	dec->cur += 4;
483		1005
484	return ((UV)d1) << 12	1006	return ((UV)d1) << 12
485	| ((UV)d2) << 8	1007	| ((UV)d2) << 8
…		…
488		1010
489	fail:	1011	fail:
490	return (UV)-1;	1012	return (UV)-1;
491	}	1013	}
492		1014
493	#define APPEND_GROW(n) SB \
494	if (cur + (n) >= end) \
495	{ \
496	STRLEN ofs = cur - SvPVX (sv); \
497	SvGROW (sv, ofs + (n) + 1); \
498	cur = SvPVX (sv) + ofs; \
499	end = SvEND (sv); \
500	} \
501	SE
502
503	#define APPEND_CH(ch) SB \
504	APPEND_GROW (1); \
505	*cur++ = (ch); \
506	SE
507
508	static SV *	1015	static SV *
509	decode_str (dec_t *dec)	1016	decode_str (dec_t *dec)
510	{	1017	{
511	SV *sv = NEWSV (0,2);	1018	SV *sv = 0;
512	int utf8 = 0;	1019	int utf8 = 0;
513	char *cur = SvPVX (sv);	1020	char *dec_cur = dec->cur;
514	char *end = SvEND (sv);
515		1021
516	for (;;)	1022	do
517	{	1023	{
518	unsigned char ch = *(unsigned char *)dec->cur;	1024	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		1025	char *cur = buf;
519		1026
520	if (ch == '"')	1027	do
521	break;
522	else if (ch == '\\')
523	{	1028	{
524	switch (*++dec->cur)	1029	unsigned char ch = *(unsigned char *)dec_cur++;
		1030
		1031	if (expect_false (ch == '"'))
525	{	1032	{
526	case '\\':	1033	--dec_cur;
527	case '/':	1034	break;
528	case '"': APPEND_CH (*dec->cur++); break;	1035	}
529		1036	else if (expect_false (ch == '\\'))
530	case 'b': APPEND_CH ('\010'); ++dec->cur; break;	1037	{
531	case 't': APPEND_CH ('\011'); ++dec->cur; break;	1038	switch (*dec_cur)
532	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
533	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
534	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
535
536	case 'u':
537	{	1039	{
538	UV lo, hi;	1040	case '\\':
539	++dec->cur;	1041	case '/':
		1042	case '"': *cur++ = *dec_cur++; break;
540		1043
541	hi = decode_4hex (dec);	1044	case 'b': ++dec_cur; *cur++ = '\010'; break;
542	if (hi == (UV)-1)	1045	case 't': ++dec_cur; *cur++ = '\011'; break;
543	goto fail;	1046	case 'n': ++dec_cur; *cur++ = '\012'; break;
		1047	case 'f': ++dec_cur; *cur++ = '\014'; break;
		1048	case 'r': ++dec_cur; *cur++ = '\015'; break;
544		1049
545	// possibly a surrogate pair	1050	case 'u':
546	if (hi >= 0xd800 && hi < 0xdc00)
547	{	1051	{
548	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	1052	UV lo, hi;
549	ERR ("missing low surrogate character in surrogate pair");	1053	++dec_cur;
550		1054
551	dec->cur += 2;	1055	dec->cur = dec_cur;
552
553	lo = decode_4hex (dec);	1056	hi = decode_4hex (dec);
		1057	dec_cur = dec->cur;
554	if (lo == (UV)-1)	1058	if (hi == (UV)-1)
555	goto fail;	1059	goto fail;
556		1060
		1061	// possibly a surrogate pair
		1062	if (hi >= 0xd800)
		1063	if (hi < 0xdc00)
		1064	{
		1065	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		1066	ERR ("missing low surrogate character in surrogate pair");
		1067
		1068	dec_cur += 2;
		1069
		1070	dec->cur = dec_cur;
		1071	lo = decode_4hex (dec);
		1072	dec_cur = dec->cur;
		1073	if (lo == (UV)-1)
		1074	goto fail;
		1075
557	if (lo < 0xdc00 || lo >= 0xe000)	1076	if (lo < 0xdc00 || lo >= 0xe000)
558	ERR ("surrogate pair expected");	1077	ERR ("surrogate pair expected");
559		1078
560	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	1079	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		1080	}
		1081	else if (hi < 0xe000)
		1082	ERR ("missing high surrogate character in surrogate pair");
		1083
		1084	if (hi >= 0x80)
		1085	{
		1086	utf8 = 1;
		1087
		1088	cur = encode_utf8 (cur, hi);
		1089	}
		1090	else
		1091	*cur++ = hi;
561	}	1092	}
562	else if (hi >= 0xdc00 && hi < 0xe000)
563	ERR ("missing high surrogate character in surrogate pair");
564
565	if (hi >= 0x80)
566	{	1093	break;
567	utf8 = 1;
568		1094
569	APPEND_GROW (4); // at most 4 bytes for 21 bits
570	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
571	}
572	else	1095	default:
573	APPEND_CH (hi);	1096	--dec_cur;
		1097	ERR ("illegal backslash escape sequence in string");
574	}	1098	}
575	break;	1099	}
		1100	else if (expect_true (ch >= 0x20 && ch < 0x80))
		1101	*cur++ = ch;
		1102	else if (ch >= 0x80)
		1103	{
		1104	STRLEN clen;
576		1105
577	default:
578	--dec->cur;	1106	--dec_cur;
579	ERR ("illegal backslash escape sequence in string");	1107
		1108	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		1109	if (clen == (STRLEN)-1)
		1110	ERR ("malformed UTF-8 character in JSON string");
		1111
		1112	do
		1113	*cur++ = *dec_cur++;
		1114	while (--clen);
		1115
		1116	utf8 = 1;
		1117	}
		1118	else if (ch == '\t' && dec->json.flags & F_RELAXED)
		1119	*cur++ = ch;
		1120	else
		1121	{
		1122	--dec_cur;
		1123
		1124	if (!ch)
		1125	ERR ("unexpected end of string while parsing JSON string");
		1126	else
		1127	ERR ("invalid character encountered while parsing JSON string");
580	}	1128	}
581	}	1129	}
582	else if (ch >= 0x20 && ch <= 0x7f)	1130	while (cur < buf + SHORT_STRING_LEN);
583	APPEND_CH (*dec->cur++);	1131
584	else if (ch >= 0x80)
585	{	1132	{
586	STRLEN clen;	1133	STRLEN len = cur - buf;
587	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
588	if (clen == (STRLEN)-1)
589	ERR ("malformed UTF-8 character in JSON string");
590		1134
591	APPEND_GROW (clen);	1135	if (sv)
592	do
593	{	1136	{
594	*cur++ = *dec->cur++;	1137	STRLEN cur = SvCUR (sv);
		1138
		1139	if (SvLEN (sv) <= cur + len)
		1140	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1141
		1142	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		1143	SvCUR_set (sv, SvCUR (sv) + len);
595	}	1144	}
596	while (--clen);
597
598	utf8 = 1;
599	}
600	else if (dec->cur == dec->end)
601	ERR ("unexpected end of string while parsing json string");
602	else	1145	else
603	ERR ("invalid character encountered");	1146	sv = newSVpvn (buf, len);
604	}	1147	}
		1148	}
		1149	while (*dec_cur != '"');
605		1150
606	++dec->cur;	1151	++dec_cur;
607		1152
608	SvCUR_set (sv, cur - SvPVX (sv));	1153	if (sv)
609		1154	{
610	SvPOK_only (sv);	1155	SvPOK_only (sv);
611	*SvEND (sv) = 0;	1156	*SvEND (sv) = 0;
612		1157
613	if (utf8)	1158	if (utf8)
614	SvUTF8_on (sv);	1159	SvUTF8_on (sv);
		1160	}
		1161	else
		1162	sv = newSVpvn ("", 0);
615		1163
616	if (dec->flags & F_SHRINK)	1164	dec->cur = dec_cur;
617	shrink (sv);
618
619	return sv;	1165	return sv;
620		1166
621	fail:	1167	fail:
622	SvREFCNT_dec (sv);	1168	dec->cur = dec_cur;
623	return 0;	1169	return 0;
624	}	1170	}
625		1171
626	static SV *	1172	static SV *
627	decode_num (dec_t *dec)	1173	decode_num (dec_t *dec)
…		…
685	is_nv = 1;	1231	is_nv = 1;
686	}	1232	}
687		1233
688	if (!is_nv)	1234	if (!is_nv)
689	{	1235	{
690	UV uv;	1236	int len = dec->cur - start;
691	int numtype = grok_number (start, dec->cur - start, &uv);	1237
692	if (numtype & IS_NUMBER_IN_UV)	1238	// special case the rather common 1..5-digit-int case
693	if (numtype & IS_NUMBER_NEG)	1239	if (*start == '-')
		1240	switch (len)
694	{	1241	{
695	if (uv < (UV)IV_MIN)	1242	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
696	return newSViv (-(IV)uv);	1243	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
		1244	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1245	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1246	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
697	}	1247	}
		1248	else
		1249	switch (len)
		1250	{
		1251	case 1: return newSViv ( start [0] - '0' * 1);
		1252	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1253	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1254	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1255	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1256	}
		1257
		1258	{
		1259	UV uv;
		1260	int numtype = grok_number (start, len, &uv);
		1261	if (numtype & IS_NUMBER_IN_UV)
		1262	if (numtype & IS_NUMBER_NEG)
		1263	{
		1264	if (uv < (UV)IV_MIN)
		1265	return newSViv (-(IV)uv);
		1266	}
698	else	1267	else
699	return newSVuv (uv);	1268	return newSVuv (uv);
700	}	1269	}
701		1270
		1271	len -= *start == '-' ? 1 : 0;
		1272
		1273	// does not fit into IV or UV, try NV
		1274	if (len <= NV_DIG)
		1275	// fits into NV without loss of precision
		1276	return newSVnv (json_atof (start));
		1277
		1278	// everything else fails, convert it to a string
		1279	return newSVpvn (start, dec->cur - start);
		1280	}
		1281
		1282	// loss of precision here
702	return newSVnv (Atof (start));	1283	return newSVnv (json_atof (start));
703		1284
704	fail:	1285	fail:
705	return 0;	1286	return 0;
706	}	1287	}
707		1288
708	static SV *	1289	static SV *
709	decode_av (dec_t *dec)	1290	decode_av (dec_t *dec)
710	{	1291	{
711	AV *av = newAV ();	1292	AV *av = newAV ();
712		1293
713	WS;	1294	DEC_INC_DEPTH;
		1295	decode_ws (dec);
		1296
714	if (*dec->cur == ']')	1297	if (*dec->cur == ']')
715	++dec->cur;	1298	++dec->cur;
716	else	1299	else
717	for (;;)	1300	for (;;)
718	{	1301	{
…		…
722	if (!value)	1305	if (!value)
723	goto fail;	1306	goto fail;
724		1307
725	av_push (av, value);	1308	av_push (av, value);
726		1309
727	WS;	1310	decode_ws (dec);
728		1311
729	if (*dec->cur == ']')	1312	if (*dec->cur == ']')
730	{	1313	{
731	++dec->cur;	1314	++dec->cur;
732	break;	1315	break;
733	}	1316	}
734		1317
735	if (*dec->cur != ',')	1318	if (*dec->cur != ',')
736	ERR (", or ] expected while parsing array");	1319	ERR (", or ] expected while parsing array");
737		1320
738	++dec->cur;	1321	++dec->cur;
		1322
		1323	decode_ws (dec);
		1324
		1325	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1326	{
		1327	++dec->cur;
		1328	break;
		1329	}
739	}	1330	}
740		1331
		1332	DEC_DEC_DEPTH;
741	return newRV_noinc ((SV *)av);	1333	return newRV_noinc ((SV *)av);
742		1334
743	fail:	1335	fail:
744	SvREFCNT_dec (av);	1336	SvREFCNT_dec (av);
		1337	DEC_DEC_DEPTH;
745	return 0;	1338	return 0;
746	}	1339	}
747		1340
748	static SV *	1341	static SV *
749	decode_hv (dec_t *dec)	1342	decode_hv (dec_t *dec)
750	{	1343	{
		1344	SV *sv;
751	HV *hv = newHV ();	1345	HV *hv = newHV ();
752		1346
753	WS;	1347	DEC_INC_DEPTH;
		1348	decode_ws (dec);
		1349
754	if (*dec->cur == '}')	1350	if (*dec->cur == '}')
755	++dec->cur;	1351	++dec->cur;
756	else	1352	else
757	for (;;)	1353	for (;;)
758	{	1354	{
759	SV *key, *value;
760
761	WS; EXPECT_CH ('"');	1355	EXPECT_CH ('"');
762		1356
763	key = decode_str (dec);	1357	// heuristic: assume that
764	if (!key)	1358	// a) decode_str + hv_store_ent are abysmally slow.
765	goto fail;	1359	// b) most hash keys are short, simple ascii text.
		1360	// => try to "fast-match" such strings to avoid
		1361	// the overhead of decode_str + hv_store_ent.
		1362	{
		1363	SV *value;
		1364	char *p = dec->cur;
		1365	char *e = p + 24; // only try up to 24 bytes
766		1366
767	WS; EXPECT_CH (':');	1367	for (;;)
768
769	value = decode_sv (dec);
770	if (!value)
771	{	1368	{
		1369	// the >= 0x80 is false on most architectures
		1370	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1371	{
		1372	// slow path, back up and use decode_str
		1373	SV *key = decode_str (dec);
		1374	if (!key)
		1375	goto fail;
		1376
		1377	decode_ws (dec); EXPECT_CH (':');
		1378
		1379	decode_ws (dec);
		1380	value = decode_sv (dec);
		1381	if (!value)
		1382	{
		1383	SvREFCNT_dec (key);
		1384	goto fail;
		1385	}
		1386
		1387	hv_store_ent (hv, key, value, 0);
772	SvREFCNT_dec (key);	1388	SvREFCNT_dec (key);
		1389
		1390	break;
		1391	}
		1392	else if (*p == '"')
		1393	{
		1394	// fast path, got a simple key
		1395	char *key = dec->cur;
		1396	int len = p - key;
		1397	dec->cur = p + 1;
		1398
		1399	decode_ws (dec); EXPECT_CH (':');
		1400
		1401	decode_ws (dec);
		1402	value = decode_sv (dec);
		1403	if (!value)
773	goto fail;	1404	goto fail;
		1405
		1406	hv_store (hv, key, len, value, 0);
		1407
		1408	break;
		1409	}
		1410
		1411	++p;
774	}	1412	}
775
776	//TODO: optimise
777	hv_store_ent (hv, key, value, 0);
778
779	WS;	1413	}
		1414
		1415	decode_ws (dec);
780		1416
781	if (*dec->cur == '}')	1417	if (*dec->cur == '}')
782	{	1418	{
783	++dec->cur;	1419	++dec->cur;
784	break;	1420	break;
…		…
786		1422
787	if (*dec->cur != ',')	1423	if (*dec->cur != ',')
788	ERR (", or } expected while parsing object/hash");	1424	ERR (", or } expected while parsing object/hash");
789		1425
790	++dec->cur;	1426	++dec->cur;
		1427
		1428	decode_ws (dec);
		1429
		1430	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1431	{
		1432	++dec->cur;
		1433	break;
		1434	}
791	}	1435	}
792		1436
		1437	DEC_DEC_DEPTH;
793	return newRV_noinc ((SV *)hv);	1438	sv = newRV_noinc ((SV *)hv);
		1439
		1440	// check filter callbacks
		1441	if (dec->json.flags & F_HOOK)
		1442	{
		1443	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1444	{
		1445	HE *cb, *he;
		1446
		1447	hv_iterinit (hv);
		1448	he = hv_iternext (hv);
		1449	hv_iterinit (hv);
		1450
		1451	// the next line creates a mortal sv each time it's called.
		1452	// might want to optimise this for common cases.
		1453	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1454
		1455	if (cb)
		1456	{
		1457	dSP;
		1458	int count;
		1459
		1460	ENTER; SAVETMPS;
		1461	SAVESTACK_POS ();
		1462	PUSHMARK (SP);
		1463	XPUSHs (HeVAL (he));
		1464	sv_2mortal (sv);
		1465
		1466	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1467
		1468	if (count == 1)
		1469	{
		1470	sv = newSVsv (POPs);
		1471	FREETMPS; LEAVE;
		1472	return sv;
		1473	}
		1474
		1475	SvREFCNT_inc (sv);
		1476	FREETMPS; LEAVE;
		1477	}
		1478	}
		1479
		1480	if (dec->json.cb_object)
		1481	{
		1482	dSP;
		1483	int count;
		1484
		1485	ENTER; SAVETMPS;
		1486	SAVESTACK_POS ();
		1487	PUSHMARK (SP);
		1488	XPUSHs (sv_2mortal (sv));
		1489
		1490	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1491
		1492	if (count == 1)
		1493	{
		1494	sv = newSVsv (POPs);
		1495	FREETMPS; LEAVE;
		1496	return sv;
		1497	}
		1498
		1499	SvREFCNT_inc (sv);
		1500	FREETMPS; LEAVE;
		1501	}
		1502	}
		1503
		1504	return sv;
794		1505
795	fail:	1506	fail:
796	SvREFCNT_dec (hv);	1507	SvREFCNT_dec (hv);
		1508	DEC_DEC_DEPTH;
		1509	return 0;
		1510	}
		1511
		1512	static SV *
		1513	decode_tag (dec_t *dec)
		1514	{
		1515	SV *tag = 0;
		1516	SV *val = 0;
		1517
		1518	if (!(dec->json.flags & F_ALLOW_TAGS))
		1519	ERR ("malformed JSON string, neither array, object, number, string or atom");
		1520
		1521	++dec->cur;
		1522
		1523	decode_ws (dec);
		1524
		1525	tag = decode_sv (dec);
		1526	if (!tag)
		1527	goto fail;
		1528
		1529	if (!SvPOK (tag))
		1530	ERR ("malformed JSON string, (tag) must be a string");
		1531
		1532	decode_ws (dec);
		1533
		1534	if (*dec->cur != ')')
		1535	ERR (") expected after tag");
		1536
		1537	++dec->cur;
		1538
		1539	decode_ws (dec);
		1540
		1541	val = decode_sv (dec);
		1542	if (!val)
		1543	goto fail;
		1544
		1545	if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
		1546	ERR ("malformed JSON string, tag value must be an array");
		1547
		1548	{
		1549	AV *av = (AV *)SvRV (val);
		1550	int i, len = av_len (av) + 1;
		1551	HV *stash = gv_stashsv (tag, 0);
		1552	SV *sv;
		1553
		1554	if (!stash)
		1555	ERR ("cannot decode perl-object (package does not exist)");
		1556
		1557	GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
		1558
		1559	if (!method)
		1560	ERR ("cannot decode perl-object (package does not have a THAW method)");
		1561
		1562	dSP;
		1563
		1564	ENTER; SAVETMPS;
		1565	PUSHMARK (SP);
		1566	EXTEND (SP, len + 2);
		1567	// we re-bless the reference to get overload and other niceties right
		1568	PUSHs (tag);
		1569	PUSHs (sv_json);
		1570
		1571	for (i = 0; i < len; ++i)
		1572	PUSHs (*av_fetch (av, i, 1));
		1573
		1574	PUTBACK;
		1575	call_sv ((SV *)GvCV (method), G_SCALAR);
		1576	SPAGAIN;
		1577
		1578	SvREFCNT_dec (tag);
		1579	SvREFCNT_dec (val);
		1580	sv = SvREFCNT_inc (POPs);
		1581
		1582	PUTBACK;
		1583
		1584	FREETMPS; LEAVE;
		1585
		1586	return sv;
		1587	}
		1588
		1589	fail:
		1590	SvREFCNT_dec (tag);
		1591	SvREFCNT_dec (val);
797	return 0;	1592	return 0;
798	}	1593	}
799		1594
800	static SV *	1595	static SV *
801	decode_sv (dec_t *dec)	1596	decode_sv (dec_t *dec)
802	{	1597	{
803	WS;	1598	// the beauty of JSON: you need exactly one character lookahead
		1599	// to parse everything.
804	switch (*dec->cur)	1600	switch (*dec->cur)
805	{	1601	{
806	case '"': ++dec->cur; return decode_str (dec);	1602	case '"': ++dec->cur; return decode_str (dec);
807	case '[': ++dec->cur; return decode_av (dec);	1603	case '[': ++dec->cur; return decode_av (dec);
808	case '{': ++dec->cur; return decode_hv (dec);	1604	case '{': ++dec->cur; return decode_hv (dec);
		1605	case '(': return decode_tag (dec);
809		1606
810	case '-':	1607	case '-':
811	case '0': case '1': case '2': case '3': case '4':	1608	case '0': case '1': case '2': case '3': case '4':
812	case '5': case '6': case '7': case '8': case '9':	1609	case '5': case '6': case '7': case '8': case '9':
813	return decode_num (dec);	1610	return decode_num (dec);
814		1611
815	case 't':	1612	case 't':
816	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1613	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
817	{	1614	{
818	dec->cur += 4;	1615	dec->cur += 4;
		1616	#if JSON_SLOW
		1617	types_true = get_bool ("Types::Serialiser::true");
		1618	#endif
819	return newSViv (1);	1619	return newSVsv (types_true);
820	}	1620	}
821	else	1621	else
822	ERR ("'true' expected");	1622	ERR ("'true' expected");
823		1623
824	break;	1624	break;
825		1625
826	case 'f':	1626	case 'f':
827	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1627	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
828	{	1628	{
829	dec->cur += 5;	1629	dec->cur += 5;
		1630	#if JSON_SLOW
		1631	types_false = get_bool ("Types::Serialiser::false");
		1632	#endif
830	return newSViv (0);	1633	return newSVsv (types_false);
831	}	1634	}
832	else	1635	else
833	ERR ("'false' expected");	1636	ERR ("'false' expected");
834		1637
835	break;	1638	break;
…		…
844	ERR ("'null' expected");	1647	ERR ("'null' expected");
845		1648
846	break;	1649	break;
847		1650
848	default:	1651	default:
849	ERR ("malformed json string, neither array, object, number, string or atom");	1652	ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
850	break;	1653	break;
851	}	1654	}
852		1655
853	fail:	1656	fail:
854	return 0;	1657	return 0;
855	}	1658	}
856		1659
857	static SV *	1660	static SV *
858	decode_json (SV *string, UV flags)	1661	decode_json (SV *string, JSON *json, char **offset_return)
859	{	1662	{
		1663	dec_t dec;
860	SV *sv;	1664	SV *sv;
861		1665
862	if (flags & F_UTF8)	1666	/* work around bugs in 5.10 where manipulating magic values
		1667	* makes perl ignore the magic in subsequent accesses.
		1668	* also make a copy of non-PV values, to get them into a clean
		1669	* state (SvPV should do that, but it's buggy, see below).
		1670	*/
		1671	/*SvGETMAGIC (string);*/
		1672	if (SvMAGICAL (string) || !SvPOK (string))
		1673	string = sv_2mortal (newSVsv (string));
		1674
		1675	SvUPGRADE (string, SVt_PV);
		1676
		1677	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1678	* assertion with -DDEBUGGING, although SvCUR is documented to
		1679	* return the xpv_cur field which certainly exists after upgrading.
		1680	* according to nicholas clark, calling SvPOK fixes this.
		1681	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1682	* and hope for the best.
		1683	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1684	* assertion business is seriously broken, try yet another workaround
		1685	* for the broken -DDEBUGGING.
		1686	*/
		1687	{
		1688	#ifdef DEBUGGING
		1689	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1690	#else
		1691	STRLEN offset = SvCUR (string);
		1692	#endif
		1693
		1694	if (offset > json->max_size && json->max_size)
		1695	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1696	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1697	}
		1698
		1699	if (DECODE_WANTS_OCTETS (json))
863	sv_utf8_downgrade (string, 0);	1700	sv_utf8_downgrade (string, 0);
864	else	1701	else
865	sv_utf8_upgrade (string);	1702	sv_utf8_upgrade (string);
866		1703
867	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1704	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
868		1705
869	dec_t dec;	1706	dec.json = *json;
870	dec.flags = flags;
871	dec.cur = SvPVX (string);	1707	dec.cur = SvPVX (string);
872	dec.end = SvEND (string);	1708	dec.end = SvEND (string);
873	dec.err = 0;	1709	dec.err = 0;
		1710	dec.depth = 0;
874		1711
		1712	if (dec.json.cb_object || dec.json.cb_sk_object)
		1713	dec.json.flags |= F_HOOK;
		1714
		1715	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1716
		1717	decode_ws (&dec);
875	sv = decode_sv (&dec);	1718	sv = decode_sv (&dec);
876		1719
		1720	if (offset_return)
		1721	*offset_return = dec.cur;
		1722
		1723	if (!(offset_return || !sv))
		1724	{
		1725	// check for trailing garbage
		1726	decode_ws (&dec);
		1727
		1728	if (*dec.cur)
		1729	{
		1730	dec.err = "garbage after JSON object";
		1731	SvREFCNT_dec (sv);
		1732	sv = 0;
		1733	}
		1734	}
		1735
877	if (!sv)	1736	if (!sv)
878	{	1737	{
879	IV offset = dec.flags & F_UTF8
880	? dec.cur - SvPVX (string)
881	: utf8_distance (dec.cur, SvPVX (string));
882	SV *uni = sv_newmortal ();	1738	SV *uni = sv_newmortal ();
		1739
883	// horrible hack to silence warning inside pv_uni_display	1740	// horrible hack to silence warning inside pv_uni_display
884	COP cop;	1741	COP cop = *PL_curcop;
885	memset (&cop, 0, sizeof (cop));
886	cop.cop_warnings = pWARN_NONE;	1742	cop.cop_warnings = pWARN_NONE;
		1743	ENTER;
887	SAVEVPTR (PL_curcop);	1744	SAVEVPTR (PL_curcop);
888	PL_curcop = &cop;	1745	PL_curcop = &cop;
889
890	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1746	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
		1747	LEAVE;
		1748
891	croak ("%s, at character offset %d (%s)",	1749	croak ("%s, at character offset %d (before \"%s\")",
892	dec.err,	1750	dec.err,
893	(int)offset,	1751	(int)ptr_to_index (string, dec.cur),
894	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1752	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
895	}	1753	}
896		1754
897	sv = sv_2mortal (sv);	1755	sv = sv_2mortal (sv);
898		1756
899	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1757	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
900	croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)");	1758	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
901		1759
902	return sv;	1760	return sv;
903	}	1761	}
904		1762
		1763	/////////////////////////////////////////////////////////////////////////////
		1764	// incremental parser
		1765
		1766	static void
		1767	incr_parse (JSON *self)
		1768	{
		1769	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1770
		1771	// the state machine here is a bit convoluted and could be simplified a lot
		1772	// but this would make it slower, so...
		1773
		1774	for (;;)
		1775	{
		1776	//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
		1777	switch (self->incr_mode)
		1778	{
		1779	// only used for initial whitespace skipping
		1780	case INCR_M_WS:
		1781	for (;;)
		1782	{
		1783	if (*p > 0x20)
		1784	{
		1785	if (*p == '#')
		1786	{
		1787	self->incr_mode = INCR_M_C0;
		1788	goto incr_m_c;
		1789	}
		1790	else
		1791	{
		1792	self->incr_mode = INCR_M_JSON;
		1793	goto incr_m_json;
		1794	}
		1795	}
		1796	else if (!*p)
		1797	goto interrupt;
		1798
		1799	++p;
		1800	}
		1801
		1802	// skip a single char inside a string (for \\-processing)
		1803	case INCR_M_BS:
		1804	if (!*p)
		1805	goto interrupt;
		1806
		1807	++p;
		1808	self->incr_mode = INCR_M_STR;
		1809	goto incr_m_str;
		1810
		1811	// inside #-style comments
		1812	case INCR_M_C0:
		1813	case INCR_M_C1:
		1814	incr_m_c:
		1815	for (;;)
		1816	{
		1817	if (*p == '\n')
		1818	{
		1819	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1820	break;
		1821	}
		1822	else if (!*p)
		1823	goto interrupt;
		1824
		1825	++p;
		1826	}
		1827
		1828	break;
		1829
		1830	// inside a string
		1831	case INCR_M_STR:
		1832	incr_m_str:
		1833	for (;;)
		1834	{
		1835	if (*p == '"')
		1836	{
		1837	++p;
		1838	self->incr_mode = INCR_M_JSON;
		1839
		1840	if (!self->incr_nest)
		1841	goto interrupt;
		1842
		1843	goto incr_m_json;
		1844	}
		1845	else if (*p == '\\')
		1846	{
		1847	++p; // "virtually" consumes character after \
		1848
		1849	if (!*p) // if at end of string we have to switch modes
		1850	{
		1851	self->incr_mode = INCR_M_BS;
		1852	goto interrupt;
		1853	}
		1854	}
		1855	else if (!*p)
		1856	goto interrupt;
		1857
		1858	++p;
		1859	}
		1860
		1861	// after initial ws, outside string
		1862	case INCR_M_JSON:
		1863	incr_m_json:
		1864	for (;;)
		1865	{
		1866	switch (*p++)
		1867	{
		1868	case 0:
		1869	--p;
		1870	goto interrupt;
		1871
		1872	case 0x09:
		1873	case 0x0a:
		1874	case 0x0d:
		1875	case 0x20:
		1876	if (!self->incr_nest)
		1877	{
		1878	--p; // do not eat the whitespace, let the next round do it
		1879	goto interrupt;
		1880	}
		1881	break;
		1882
		1883	case '"':
		1884	self->incr_mode = INCR_M_STR;
		1885	goto incr_m_str;
		1886
		1887	case '[':
		1888	case '{':
		1889	case '(':
		1890	if (++self->incr_nest > self->max_depth)
		1891	croak (ERR_NESTING_EXCEEDED);
		1892	break;
		1893
		1894	case ']':
		1895	case '}':
		1896	if (--self->incr_nest <= 0)
		1897	goto interrupt;
		1898	break;
		1899
		1900	case ')':
		1901	--self->incr_nest;
		1902	break;
		1903
		1904	case '#':
		1905	self->incr_mode = INCR_M_C1;
		1906	goto incr_m_c;
		1907	}
		1908	}
		1909	}
		1910
		1911	modechange:
		1912	;
		1913	}
		1914
		1915	interrupt:
		1916	self->incr_pos = p - SvPVX (self->incr_text);
		1917	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1918	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
		1919	}
		1920
		1921	/////////////////////////////////////////////////////////////////////////////
		1922	// XS interface functions
		1923
905	MODULE = JSON::XS PACKAGE = JSON::XS	1924	MODULE = JSON::XS PACKAGE = JSON::XS
906		1925
907	BOOT:	1926	BOOT:
908	{	1927	{
909	int i;	1928	int i;
910		1929
911	memset (decode_hexdigit, 0xff, 256);
912	for (i = 10; i--; )	1930	for (i = 0; i < 256; ++i)
913	decode_hexdigit ['0' + i] = i;	1931	decode_hexdigit [i] =
		1932	i >= '0' && i <= '9' ? i - '0'
		1933	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1934	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1935	: -1;
914		1936
915	for (i = 7; i--; )
916	{
917	decode_hexdigit ['a' + i] = 10 + i;
918	decode_hexdigit ['A' + i] = 10 + i;
919	}
920
921	json_stash = gv_stashpv ("JSON::XS", 1);	1937	json_stash = gv_stashpv ("JSON::XS" , 1);
		1938	types_boolean_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
		1939
		1940	types_true = get_bool ("Types::Serialiser::true");
		1941	types_false = get_bool ("Types::Serialiser::false");
		1942
		1943	sv_json = newSVpv ("JSON", 0);
		1944	SvREADONLY_on (sv_json);
		1945
		1946	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
922	}	1947	}
923		1948
924	PROTOTYPES: DISABLE	1949	PROTOTYPES: DISABLE
925		1950
926	SV *new (char *dummy)	1951	void CLONE (...)
927	CODE:	1952	CODE:
928	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1953	json_stash = 0;
		1954	types_boolean_stash = 0;
		1955
		1956	void new (char *klass)
		1957	PPCODE:
		1958	{
		1959	SV *pv = NEWSV (0, sizeof (JSON));
		1960	SvPOK_only (pv);
		1961	json_init ((JSON *)SvPVX (pv));
		1962	XPUSHs (sv_2mortal (sv_bless (
		1963	newRV_noinc (pv),
		1964	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1965	)));
		1966	}
		1967
		1968	void ascii (JSON *self, int enable = 1)
		1969	ALIAS:
		1970	ascii = F_ASCII
		1971	latin1 = F_LATIN1
		1972	utf8 = F_UTF8
		1973	indent = F_INDENT
		1974	canonical = F_CANONICAL
		1975	space_before = F_SPACE_BEFORE
		1976	space_after = F_SPACE_AFTER
		1977	pretty = F_PRETTY
		1978	allow_nonref = F_ALLOW_NONREF
		1979	shrink = F_SHRINK
		1980	allow_blessed = F_ALLOW_BLESSED
		1981	convert_blessed = F_CONV_BLESSED
		1982	relaxed = F_RELAXED
		1983	allow_unknown = F_ALLOW_UNKNOWN
		1984	allow_tags = F_ALLOW_TAGS
		1985	PPCODE:
		1986	{
		1987	if (enable)
		1988	self->flags |= ix;
		1989	else
		1990	self->flags &= ~ix;
		1991
		1992	XPUSHs (ST (0));
		1993	}
		1994
		1995	void get_ascii (JSON *self)
		1996	ALIAS:
		1997	get_ascii = F_ASCII
		1998	get_latin1 = F_LATIN1
		1999	get_utf8 = F_UTF8
		2000	get_indent = F_INDENT
		2001	get_canonical = F_CANONICAL
		2002	get_space_before = F_SPACE_BEFORE
		2003	get_space_after = F_SPACE_AFTER
		2004	get_allow_nonref = F_ALLOW_NONREF
		2005	get_shrink = F_SHRINK
		2006	get_allow_blessed = F_ALLOW_BLESSED
		2007	get_convert_blessed = F_CONV_BLESSED
		2008	get_relaxed = F_RELAXED
		2009	get_allow_unknown = F_ALLOW_UNKNOWN
		2010	get_allow_tags = F_ALLOW_TAGS
		2011	PPCODE:
		2012	XPUSHs (boolSV (self->flags & ix));
		2013
		2014	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		2015	PPCODE:
		2016	self->max_depth = max_depth;
		2017	XPUSHs (ST (0));
		2018
		2019	U32 get_max_depth (JSON *self)
		2020	CODE:
		2021	RETVAL = self->max_depth;
929	OUTPUT:	2022	OUTPUT:
930	RETVAL	2023	RETVAL
931		2024
932	SV *ascii (SV *self, int enable = 1)	2025	void max_size (JSON *self, U32 max_size = 0)
933	ALIAS:	2026	PPCODE:
934	ascii = F_ASCII	2027	self->max_size = max_size;
935	utf8 = F_UTF8	2028	XPUSHs (ST (0));
936	indent = F_INDENT	2029
937	canonical = F_CANONICAL	2030	int get_max_size (JSON *self)
938	space_before = F_SPACE_BEFORE
939	space_after = F_SPACE_AFTER
940	json_rpc = F_JSON_RPC
941	pretty = F_PRETTY
942	allow_nonref = F_ALLOW_NONREF
943	shrink = F_SHRINK
944	CODE:	2031	CODE:
945	{	2032	RETVAL = self->max_size;
946	UV *uv = SvJSON (self);
947	if (enable)
948	*uv |= ix;
949	else
950	*uv &= ~ix;
951
952	RETVAL = newSVsv (self);
953	}
954	OUTPUT:	2033	OUTPUT:
955	RETVAL	2034	RETVAL
956		2035
957	void encode (SV *self, SV *scalar)	2036	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
958	PPCODE:	2037	PPCODE:
959	XPUSHs (encode_json (scalar, *SvJSON (self)));	2038	{
		2039	SvREFCNT_dec (self->cb_object);
		2040	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
960		2041
961	void decode (SV *self, SV *jsonstr)	2042	XPUSHs (ST (0));
		2043	}
		2044
		2045	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
962	PPCODE:	2046	PPCODE:
963	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	2047	{
		2048	if (!self->cb_sk_object)
		2049	self->cb_sk_object = newHV ();
		2050
		2051	if (SvOK (cb))
		2052	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		2053	else
		2054	{
		2055	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		2056
		2057	if (!HvKEYS (self->cb_sk_object))
		2058	{
		2059	SvREFCNT_dec (self->cb_sk_object);
		2060	self->cb_sk_object = 0;
		2061	}
		2062	}
		2063
		2064	XPUSHs (ST (0));
		2065	}
		2066
		2067	void encode (JSON *self, SV *scalar)
		2068	PPCODE:
		2069	PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
		2070	XPUSHs (scalar);
		2071
		2072	void decode (JSON *self, SV *jsonstr)
		2073	PPCODE:
		2074	PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
		2075	XPUSHs (jsonstr);
		2076
		2077	void decode_prefix (JSON *self, SV *jsonstr)
		2078	PPCODE:
		2079	{
		2080	SV *sv;
		2081	char *offset;
		2082	PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
		2083	EXTEND (SP, 2);
		2084	PUSHs (sv);
		2085	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		2086	}
		2087
		2088	void incr_parse (JSON *self, SV *jsonstr = 0)
		2089	PPCODE:
		2090	{
		2091	if (!self->incr_text)
		2092	self->incr_text = newSVpvn ("", 0);
		2093
		2094	/* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
		2095	if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
		2096	if (DECODE_WANTS_OCTETS (self))
		2097	{
		2098	if (self->incr_pos)
		2099	self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
		2100	(U8 *)SvPVX (self->incr_text) + self->incr_pos);
		2101
		2102	sv_utf8_downgrade (self->incr_text, 0);
		2103	}
		2104	else
		2105	{
		2106	sv_utf8_upgrade (self->incr_text);
		2107
		2108	if (self->incr_pos)
		2109	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		2110	- (U8 *)SvPVX (self->incr_text);
		2111	}
		2112
		2113	// append data, if any
		2114	if (jsonstr)
		2115	{
		2116	/* make sure both strings have same encoding */
		2117	if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
		2118	if (SvUTF8 (jsonstr))
		2119	sv_utf8_downgrade (jsonstr, 0);
		2120	else
		2121	sv_utf8_upgrade (jsonstr);
		2122
		2123	/* and then just blindly append */
		2124	{
		2125	STRLEN len;
		2126	const char *str = SvPV (jsonstr, len);
		2127	STRLEN cur = SvCUR (self->incr_text);
		2128
		2129	if (SvLEN (self->incr_text) <= cur + len)
		2130	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		2131
		2132	Move (str, SvEND (self->incr_text), len, char);
		2133	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		2134	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		2135	}
		2136	}
		2137
		2138	if (GIMME_V != G_VOID)
		2139	do
		2140	{
		2141	SV *sv;
		2142	char *offset;
		2143
		2144	if (!INCR_DONE (self))
		2145	{
		2146	incr_parse (self);
		2147
		2148	if (self->incr_pos > self->max_size && self->max_size)
		2149	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		2150	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		2151
		2152	if (!INCR_DONE (self))
		2153	{
		2154	// as an optimisation, do not accumulate white space in the incr buffer
		2155	if (self->incr_mode == INCR_M_WS && self->incr_pos)
		2156	{
		2157	self->incr_pos = 0;
		2158	SvCUR_set (self->incr_text, 0);
		2159	}
		2160
		2161	break;
		2162	}
		2163	}
		2164
		2165	PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
		2166	XPUSHs (sv);
		2167
		2168	self->incr_pos -= offset - SvPVX (self->incr_text);
		2169	self->incr_nest = 0;
		2170	self->incr_mode = 0;
		2171
		2172	sv_chop (self->incr_text, offset);
		2173	}
		2174	while (GIMME_V == G_ARRAY);
		2175	}
		2176
		2177	SV *incr_text (JSON *self)
		2178	ATTRS: lvalue
		2179	CODE:
		2180	{
		2181	if (self->incr_pos)
		2182	croak ("incr_text can not be called when the incremental parser already started parsing");
		2183
		2184	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		2185	}
		2186	OUTPUT:
		2187	RETVAL
		2188
		2189	void incr_skip (JSON *self)
		2190	CODE:
		2191	{
		2192	if (self->incr_pos)
		2193	{
		2194	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2195	self->incr_pos = 0;
		2196	self->incr_nest = 0;
		2197	self->incr_mode = 0;
		2198	}
		2199	}
		2200
		2201	void incr_reset (JSON *self)
		2202	CODE:
		2203	{
		2204	SvREFCNT_dec (self->incr_text);
		2205	self->incr_text = 0;
		2206	self->incr_pos = 0;
		2207	self->incr_nest = 0;
		2208	self->incr_mode = 0;
		2209	}
		2210
		2211	void DESTROY (JSON *self)
		2212	CODE:
		2213	SvREFCNT_dec (self->cb_sk_object);
		2214	SvREFCNT_dec (self->cb_object);
		2215	SvREFCNT_dec (self->incr_text);
964		2216
965	PROTOTYPES: ENABLE	2217	PROTOTYPES: ENABLE
966		2218
967	void to_json (SV *scalar)	2219	void encode_json (SV *scalar)
968	PPCODE:	2220	PPCODE:
969	XPUSHs (encode_json (scalar, F_UTF8));	2221	{
		2222	JSON json;
		2223	json_init (&json);
		2224	json.flags |= F_UTF8;
		2225	PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
		2226	XPUSHs (scalar);
		2227	}
970		2228
971	void from_json (SV *jsonstr)	2229	void decode_json (SV *jsonstr)
972	PPCODE:	2230	PPCODE:
973	XPUSHs (decode_json (jsonstr, F_UTF8));	2231	{
		2232	JSON json;
		2233	json_init (&json);
		2234	json.flags |= F_UTF8;
		2235	PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
		2236	XPUSHs (jsonstr);
		2237	}
974		2238

Diff Legend

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