ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.34 by root, Wed May 23 22:07:13 2007 UTC vs.
Revision 1.121 by root, Tue Oct 29 00:06:40 2013 UTC

1	#include "EXTERN.h"	1	#include "EXTERN.h"
2	#include "perl.h"	2	#include "perl.h"
3	#include "XSUB.h"	3	#include "XSUB.h"
4		4
5	#include "assert.h"	5	#include <assert.h>
6	#include "string.h"	6	#include <string.h>
7	#include "stdlib.h"	7	#include <stdlib.h>
8	#include "stdio.h"	8	#include <stdio.h>
		9	#include <limits.h>
		10	#include <float.h>
9		11
10	#if defined(__BORLANDC__) || defined(_MSC_VER)	12	#if defined(__BORLANDC__) || defined(_MSC_VER)
11	# define snprintf _snprintf // C compilers have this in stdio.h	13	# define snprintf _snprintf // C compilers have this in stdio.h
12	#endif	14	#endif
13		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	// three extra for rounding, sign, and end of string
		23	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
		24
14	#define F_ASCII 0x00000001UL	25	#define F_ASCII 0x00000001UL
15	#define F_LATIN1 0x00000002UL	26	#define F_LATIN1 0x00000002UL
16	#define F_UTF8 0x00000004UL	27	#define F_UTF8 0x00000004UL
17	#define F_INDENT 0x00000008UL	28	#define F_INDENT 0x00000008UL
18	#define F_CANONICAL 0x00000010UL	29	#define F_CANONICAL 0x00000010UL
19	#define F_SPACE_BEFORE 0x00000020UL	30	#define F_SPACE_BEFORE 0x00000020UL
20	#define F_SPACE_AFTER 0x00000040UL	31	#define F_SPACE_AFTER 0x00000040UL
21	#define F_ALLOW_NONREF 0x00000100UL	32	#define F_ALLOW_NONREF 0x00000100UL
22	#define F_SHRINK 0x00000200UL	33	#define F_SHRINK 0x00000200UL
		34	#define F_ALLOW_BLESSED 0x00000400UL
		35	#define F_CONV_BLESSED 0x00000800UL
23	#define F_MAXDEPTH 0xf8000000UL	36	#define F_RELAXED 0x00001000UL
24	#define S_MAXDEPTH 27	37	#define F_ALLOW_UNKNOWN 0x00002000UL
25		38	#define F_ALLOW_TAGS 0x00004000UL
26	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))	39	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
27
28	// F_SELFCONVERT? <=> to_json/toJson
29	// F_BLESSED? <=> { $__class__$ => }
30		40
31	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	41	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
32	#define F_DEFAULT (9UL << S_MAXDEPTH)
33		42
34	#define INIT_SIZE 32 // initial scalar size to be allocated	43	#define INIT_SIZE 32 // initial scalar size to be allocated
35	#define INDENT_STEP 3 // spaces per indentation level	44	#define INDENT_STEP 3 // spaces per indentation level
36		45
37	#define SHORT_STRING_LEN 512 // special-case strings of up to this size	46	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
		47
		48	#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
38		49
39	#define SB do {	50	#define SB do {
40	#define SE } while (0)	51	#define SE } while (0)
41		52
		53	#if __GNUC__ >= 3
		54	# define expect(expr,value) __builtin_expect ((expr), (value))
		55	# define INLINE static inline
		56	#else
		57	# define expect(expr,value) (expr)
		58	# define INLINE static
		59	#endif
		60
		61	#define expect_false(expr) expect ((expr) != 0, 0)
		62	#define expect_true(expr) expect ((expr) != 0, 1)
		63
		64	#define IN_RANGE_INC(type,val,beg,end) \
		65	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		66	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		67
		68	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
		69
		70	#ifdef USE_ITHREADS
		71	# define JSON_SLOW 1
		72	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		73	#else
		74	# define JSON_SLOW 0
		75	# define JSON_STASH json_stash
		76	#endif
		77
		78	// the amount of HEs to allocate on the stack, when sorting keys
		79	#define STACK_HES 64
		80
42	static HV *json_stash; // JSON::XS::	81	static HV *json_stash, *types_boolean_stash; // JSON::XS::
		82	static SV *types_true, *types_false, *sv_json;
		83
		84	enum {
		85	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		86	INCR_M_STR, // inside string
		87	INCR_M_BS, // inside backslash
		88	INCR_M_C0, // inside comment in initial whitespace sequence
		89	INCR_M_C1, // inside comment in other places
		90	INCR_M_JSON // outside anything, count nesting
		91	};
		92
		93	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		94
		95	typedef struct {
		96	U32 flags;
		97	U32 max_depth;
		98	STRLEN max_size;
		99
		100	SV *cb_object;
		101	HV *cb_sk_object;
		102
		103	// for the incremental parser
		104	SV *incr_text; // the source text so far
		105	STRLEN incr_pos; // the current offset into the text
		106	int incr_nest; // {[]}-nesting level
		107	unsigned char incr_mode;
		108	} JSON;
		109
		110	INLINE void
		111	json_init (JSON *json)
		112	{
		113	Zero (json, 1, JSON);
		114	json->max_depth = 512;
		115	}
43		116
44	/////////////////////////////////////////////////////////////////////////////	117	/////////////////////////////////////////////////////////////////////////////
45	// utility functions	118	// utility functions
46		119
47	static UV *	120	INLINE SV *
48	SvJSON (SV *sv)	121	get_bool (const char *name)
49	{	122	{
50	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))	123	SV *sv = get_sv (name, 1);
51	croak ("object is not of type JSON::XS");
52		124
53	return &SvUVX (SvRV (sv));	125	SvREADONLY_on (sv);
54	}	126	SvREADONLY_on (SvRV (sv));
55		127
56	static void	128	return sv;
		129	}
		130
		131	INLINE void
57	shrink (SV *sv)	132	shrink (SV *sv)
58	{	133	{
59	sv_utf8_downgrade (sv, 1);	134	sv_utf8_downgrade (sv, 1);
		135
60	if (SvLEN (sv) > SvCUR (sv) + 1)	136	if (SvLEN (sv) > SvCUR (sv) + 1)
61	{	137	{
62	#ifdef SvPV_shrink_to_cur	138	#ifdef SvPV_shrink_to_cur
63	SvPV_shrink_to_cur (sv);	139	SvPV_shrink_to_cur (sv);
64	#elif defined (SvPV_renew)	140	#elif defined (SvPV_renew)
…		…
70	// decode an utf-8 character and return it, or (UV)-1 in	146	// decode an utf-8 character and return it, or (UV)-1 in
71	// case of an error.	147	// case of an error.
72	// we special-case "safe" characters from U+80 .. U+7FF,	148	// we special-case "safe" characters from U+80 .. U+7FF,
73	// but use the very good perl function to parse anything else.	149	// but use the very good perl function to parse anything else.
74	// note that we never call this function for a ascii codepoints	150	// note that we never call this function for a ascii codepoints
75	static UV	151	INLINE UV
76	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)	152	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
77	{	153	{
78	if (s[0] > 0xdf || s[0] < 0xc2)	154	if (expect_true (len >= 2
79	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	155	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
80	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	156	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
81	{	157	{
82	*clen = 2;	158	*clen = 2;
83	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);	159	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
84	}	160	}
85	else	161	else
86	{	162	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
87	*clen = (STRLEN)-1;	163	}
88	return (UV)-1;	164
		165	// likewise for encoding, also never called for ascii codepoints
		166	// this function takes advantage of this fact, although current gccs
		167	// seem to optimise the check for >= 0x80 away anyways
		168	INLINE unsigned char *
		169	encode_utf8 (unsigned char *s, UV ch)
		170	{
		171	if (expect_false (ch < 0x000080))
		172	*s++ = ch;
		173	else if (expect_true (ch < 0x000800))
		174	*s++ = 0xc0 | ( ch >> 6),
		175	*s++ = 0x80 | ( ch & 0x3f);
		176	else if ( ch < 0x010000)
		177	*s++ = 0xe0 | ( ch >> 12),
		178	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		179	*s++ = 0x80 | ( ch & 0x3f);
		180	else if ( ch < 0x110000)
		181	*s++ = 0xf0 | ( ch >> 18),
		182	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		183	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		184	*s++ = 0x80 | ( ch & 0x3f);
		185
		186	return s;
		187	}
		188
		189	// convert offset pointer to character index, sv must be string
		190	static STRLEN
		191	ptr_to_index (SV *sv, char *offset)
		192	{
		193	return SvUTF8 (sv)
		194	? utf8_distance (offset, SvPVX (sv))
		195	: offset - SvPVX (sv);
		196	}
		197
		198	/////////////////////////////////////////////////////////////////////////////
		199	// fp hell
		200
		201	// scan a group of digits, and a trailing exponent
		202	static void
		203	json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
		204	{
		205	UV uaccum = 0;
		206	int eaccum = 0;
		207
		208	// if we recurse too deep, skip all remaining digits
		209	// to avoid a stack overflow attack
		210	if (expect_false (--maxdepth <= 0))
		211	while (((U8)*s - '0') < 10)
		212	++s;
		213
		214	for (;;)
89	}	215	{
90	}	216	U8 dig = (U8)*s - '0';
91		217
		218	if (expect_false (dig >= 10))
		219	{
		220	if (dig == (U8)((U8)'.' - (U8)'0'))
		221	{
		222	++s;
		223	json_atof_scan1 (s, accum, expo, 1, maxdepth);
		224	}
		225	else if ((dig | ' ') == 'e' - '0')
		226	{
		227	int exp2 = 0;
		228	int neg = 0;
		229
		230	++s;
		231
		232	if (*s == '-')
		233	{
		234	++s;
		235	neg = 1;
		236	}
		237	else if (*s == '+')
		238	++s;
		239
		240	while ((dig = (U8)*s - '0') < 10)
		241	exp2 = exp2 * 10 + *s++ - '0';
		242
		243	*expo += neg ? -exp2 : exp2;
		244	}
		245
		246	break;
		247	}
		248
		249	++s;
		250
		251	uaccum = uaccum * 10 + dig;
		252	++eaccum;
		253
		254	// if we have too many digits, then recurse for more
		255	// we actually do this for rather few digits
		256	if (uaccum >= (UV_MAX - 9) / 10)
		257	{
		258	if (postdp) *expo -= eaccum;
		259	json_atof_scan1 (s, accum, expo, postdp, maxdepth);
		260	if (postdp) *expo += eaccum;
		261
		262	break;
		263	}
		264	}
		265
		266	// this relies greatly on the quality of the pow ()
		267	// implementation of the platform, but a good
		268	// implementation is hard to beat.
		269	// (IEEE 754 conformant ones are required to be exact)
		270	if (postdp) *expo -= eaccum;
		271	*accum += uaccum * Perl_pow (10., *expo);
		272	*expo += eaccum;
		273	}
		274
		275	static NV
		276	json_atof (const char *s)
		277	{
		278	NV accum = 0.;
		279	int expo = 0;
		280	int neg = 0;
		281
		282	if (*s == '-')
		283	{
		284	++s;
		285	neg = 1;
		286	}
		287
		288	// a recursion depth of ten gives us >>500 bits
		289	json_atof_scan1 (s, &accum, &expo, 0, 10);
		290
		291	return neg ? -accum : accum;
		292	}
92	/////////////////////////////////////////////////////////////////////////////	293	/////////////////////////////////////////////////////////////////////////////
93	// encoder	294	// encoder
94		295
95	// structure used for encoding JSON	296	// structure used for encoding JSON
96	typedef struct	297	typedef struct
97	{	298	{
98	char *cur; // SvPVX (sv) + current output position	299	char *cur; // SvPVX (sv) + current output position
99	char *end; // SvEND (sv)	300	char *end; // SvEND (sv)
100	SV *sv; // result scalar	301	SV *sv; // result scalar
101	U32 flags; // F_*	302	JSON json;
102	U32 indent; // indentation level	303	U32 indent; // indentation level
103	U32 maxdepth; // max. indentation/recursion level	304	UV limit; // escape character values >= this value when encoding
104	} enc_t;	305	} enc_t;
105		306
106	static void	307	INLINE void
107	need (enc_t *enc, STRLEN len)	308	need (enc_t *enc, STRLEN len)
108	{	309	{
109	if (enc->cur + len >= enc->end)	310	if (expect_false (enc->cur + len >= enc->end))
110	{	311	{
111	STRLEN cur = enc->cur - SvPVX (enc->sv);	312	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
112	SvGROW (enc->sv, cur + len + 1);	313	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
113	enc->cur = SvPVX (enc->sv) + cur;	314	enc->cur = SvPVX (enc->sv) + cur;
114	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	315	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
115	}	316	}
116	}	317	}
117		318
118	static void	319	INLINE void
119	encode_ch (enc_t *enc, char ch)	320	encode_ch (enc_t *enc, char ch)
120	{	321	{
121	need (enc, 1);	322	need (enc, 1);
122	*enc->cur++ = ch;	323	*enc->cur++ = ch;
123	}	324	}
…		…
131		332
132	while (str < end)	333	while (str < end)
133	{	334	{
134	unsigned char ch = *(unsigned char *)str;	335	unsigned char ch = *(unsigned char *)str;
135		336
136	if (ch >= 0x20 && ch < 0x80) // most common case	337	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
137	{	338	{
138	if (ch == '"') // but with slow exceptions	339	if (expect_false (ch == '"')) // but with slow exceptions
139	{	340	{
140	need (enc, len += 1);	341	need (enc, len += 1);
141	*enc->cur++ = '\\';	342	*enc->cur++ = '\\';
142	*enc->cur++ = '"';	343	*enc->cur++ = '"';
143	}	344	}
144	else if (ch == '\\')	345	else if (expect_false (ch == '\\'))
145	{	346	{
146	need (enc, len += 1);	347	need (enc, len += 1);
147	*enc->cur++ = '\\';	348	*enc->cur++ = '\\';
148	*enc->cur++ = '\\';	349	*enc->cur++ = '\\';
149	}	350	}
…		…
167	STRLEN clen;	368	STRLEN clen;
168	UV uch;	369	UV uch;
169		370
170	if (is_utf8)	371	if (is_utf8)
171	{	372	{
172	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
173	uch = decode_utf8 (str, end - str, &clen);	373	uch = decode_utf8 (str, end - str, &clen);
174	if (clen == (STRLEN)-1)	374	if (clen == (STRLEN)-1)
175	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	375	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
176	}	376	}
177	else	377	else
178	{	378	{
179	uch = ch;	379	uch = ch;
180	clen = 1;	380	clen = 1;
181	}	381	}
182		382
183	if (uch > 0x10FFFFUL)	383	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
184	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
185
186	if (uch < 0x80 || enc->flags & F_ASCII || (enc->flags & F_LATIN1 && uch > 0xFF))
187	{	384	{
188	if (uch > 0xFFFFUL)	385	if (uch >= 0x10000UL)
189	{	386	{
		387	if (uch >= 0x110000UL)
		388	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		389
190	need (enc, len += 11);	390	need (enc, len += 11);
191	sprintf (enc->cur, "\\u%04x\\u%04x",	391	sprintf (enc->cur, "\\u%04x\\u%04x",
192	(int)((uch - 0x10000) / 0x400 + 0xD800),	392	(int)((uch - 0x10000) / 0x400 + 0xD800),
193	(int)((uch - 0x10000) % 0x400 + 0xDC00));	393	(int)((uch - 0x10000) % 0x400 + 0xDC00));
194	enc->cur += 12;	394	enc->cur += 12;
195	}	395	}
196	else	396	else
197	{	397	{
198	static char hexdigit [16] = "0123456789abcdef";
199	need (enc, len += 5);	398	need (enc, len += 5);
200	*enc->cur++ = '\\';	399	*enc->cur++ = '\\';
201	*enc->cur++ = 'u';	400	*enc->cur++ = 'u';
202	*enc->cur++ = hexdigit [ uch >> 12 ];	401	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
203	*enc->cur++ = hexdigit [(uch >> 8) & 15];	402	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
204	*enc->cur++ = hexdigit [(uch >> 4) & 15];	403	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
205	*enc->cur++ = hexdigit [(uch >> 0) & 15];	404	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
206	}	405	}
207		406
208	str += clen;	407	str += clen;
209	}	408	}
210	else if (enc->flags & F_LATIN1)	409	else if (enc->json.flags & F_LATIN1)
211	{	410	{
212	*enc->cur++ = uch;	411	*enc->cur++ = uch;
213	str += clen;	412	str += clen;
214	}	413	}
215	else if (is_utf8)	414	else if (is_utf8)
…		…
222	while (--clen);	421	while (--clen);
223	}	422	}
224	else	423	else
225	{	424	{
226	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	425	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
227	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	426	enc->cur = encode_utf8 (enc->cur, uch);
228	++str;	427	++str;
229	}	428	}
230	}	429	}
231	}	430	}
232	}	431	}
233		432
234	--len;	433	--len;
235	}	434	}
236	}	435	}
237		436
238	static void	437	INLINE void
239	encode_indent (enc_t *enc)	438	encode_indent (enc_t *enc)
240	{	439	{
241	if (enc->flags & F_INDENT)	440	if (enc->json.flags & F_INDENT)
242	{	441	{
243	int spaces = enc->indent * INDENT_STEP;	442	int spaces = enc->indent * INDENT_STEP;
244		443
245	need (enc, spaces);	444	need (enc, spaces);
246	memset (enc->cur, ' ', spaces);	445	memset (enc->cur, ' ', spaces);
247	enc->cur += spaces;	446	enc->cur += spaces;
248	}	447	}
249	}	448	}
250		449
251	static void	450	INLINE void
252	encode_space (enc_t *enc)	451	encode_space (enc_t *enc)
253	{	452	{
254	need (enc, 1);	453	need (enc, 1);
255	encode_ch (enc, ' ');	454	encode_ch (enc, ' ');
256	}	455	}
257		456
258	static void	457	INLINE void
259	encode_nl (enc_t *enc)	458	encode_nl (enc_t *enc)
260	{	459	{
261	if (enc->flags & F_INDENT)	460	if (enc->json.flags & F_INDENT)
262	{	461	{
263	need (enc, 1);	462	need (enc, 1);
264	encode_ch (enc, '\n');	463	encode_ch (enc, '\n');
265	}	464	}
266	}	465	}
267		466
268	static void	467	INLINE void
269	encode_comma (enc_t *enc)	468	encode_comma (enc_t *enc)
270	{	469	{
271	encode_ch (enc, ',');	470	encode_ch (enc, ',');
272		471
273	if (enc->flags & F_INDENT)	472	if (enc->json.flags & F_INDENT)
274	encode_nl (enc);	473	encode_nl (enc);
275	else if (enc->flags & F_SPACE_AFTER)	474	else if (enc->json.flags & F_SPACE_AFTER)
276	encode_space (enc);	475	encode_space (enc);
277	}	476	}
278		477
279	static void encode_sv (enc_t *enc, SV *sv);	478	static void encode_sv (enc_t *enc, SV *sv);
280		479
281	static void	480	static void
282	encode_av (enc_t *enc, AV *av)	481	encode_av (enc_t *enc, AV *av)
283	{	482	{
284	int i, len = av_len (av);	483	int i, len = av_len (av);
285		484
286	if (enc->indent >= enc->maxdepth)	485	if (enc->indent >= enc->json.max_depth)
287	croak ("data structure too deep (hit recursion limit)");	486	croak (ERR_NESTING_EXCEEDED);
288		487
289	encode_ch (enc, '['); encode_nl (enc);	488	encode_ch (enc, '[');
290	++enc->indent;
291		489
		490	if (len >= 0)
		491	{
		492	encode_nl (enc); ++enc->indent;
		493
292	for (i = 0; i <= len; ++i)	494	for (i = 0; i <= len; ++i)
293	{	495	{
		496	SV **svp = av_fetch (av, i, 0);
		497
294	encode_indent (enc);	498	encode_indent (enc);
295	encode_sv (enc, *av_fetch (av, i, 0));
296		499
		500	if (svp)
		501	encode_sv (enc, *svp);
		502	else
		503	encode_str (enc, "null", 4, 0);
		504
297	if (i < len)	505	if (i < len)
298	encode_comma (enc);	506	encode_comma (enc);
299	}	507	}
300		508
		509	encode_nl (enc); --enc->indent; encode_indent (enc);
		510	}
		511
301	encode_nl (enc);	512	encode_ch (enc, ']');
302
303	--enc->indent;
304	encode_indent (enc); encode_ch (enc, ']');
305	}	513	}
306		514
307	static void	515	static void
308	encode_he (enc_t *enc, HE *he)	516	encode_hk (enc_t *enc, HE *he)
309	{	517	{
310	encode_ch (enc, '"');	518	encode_ch (enc, '"');
311		519
312	if (HeKLEN (he) == HEf_SVKEY)	520	if (HeKLEN (he) == HEf_SVKEY)
313	{	521	{
314	SV *sv = HeSVKEY (he);	522	SV *sv = HeSVKEY (he);
315	STRLEN len;	523	STRLEN len;
316	char *str;	524	char *str;
317		525
318	SvGETMAGIC (sv);	526	SvGETMAGIC (sv);
319	str = SvPV (sv, len);	527	str = SvPV (sv, len);
320		528
321	encode_str (enc, str, len, SvUTF8 (sv));	529	encode_str (enc, str, len, SvUTF8 (sv));
322	}	530	}
323	else	531	else
324	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	532	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
325		533
326	encode_ch (enc, '"');	534	encode_ch (enc, '"');
327		535
328	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	536	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
329	encode_ch (enc, ':');	537	encode_ch (enc, ':');
330	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	538	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
331	encode_sv (enc, HeVAL (he));
332	}	539	}
333		540
334	// compare hash entries, used when all keys are bytestrings	541	// compare hash entries, used when all keys are bytestrings
335	static int	542	static int
336	he_cmp_fast (const void *a_, const void *b_)	543	he_cmp_fast (const void *a_, const void *b_)
…		…
341	HE *b = *(HE **)b_;	548	HE *b = *(HE **)b_;
342		549
343	STRLEN la = HeKLEN (a);	550	STRLEN la = HeKLEN (a);
344	STRLEN lb = HeKLEN (b);	551	STRLEN lb = HeKLEN (b);
345		552
346	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	553	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
347	cmp = la - lb;	554	cmp = lb - la;
348		555
349	return cmp;	556	return cmp;
350	}	557	}
351		558
352	// compare hash entries, used when some keys are sv's or utf-x	559	// compare hash entries, used when some keys are sv's or utf-x
353	static int	560	static int
354	he_cmp_slow (const void *a, const void *b)	561	he_cmp_slow (const void *a, const void *b)
355	{	562	{
356	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	563	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
357	}	564	}
358		565
359	static void	566	static void
360	encode_hv (enc_t *enc, HV *hv)	567	encode_hv (enc_t *enc, HV *hv)
361	{	568	{
362	int count, i;	569	HE *he;
363		570
364	if (enc->indent >= enc->maxdepth)	571	if (enc->indent >= enc->json.max_depth)
365	croak ("data structure too deep (hit recursion limit)");	572	croak (ERR_NESTING_EXCEEDED);
366		573
367	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	574	encode_ch (enc, '{');
368		575
369	if ((count = hv_iterinit (hv)))
370	{
371	// for canonical output we have to sort by keys first	576	// for canonical output we have to sort by keys first
372	// actually, this is mostly due to the stupid so-called	577	// actually, this is mostly due to the stupid so-called
373	// security workaround added somewhere in 5.8.x.	578	// security workaround added somewhere in 5.8.x
374	// that randomises hash orderings	579	// that randomises hash orderings
375	if (enc->flags & F_CANONICAL)	580	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
		581	{
		582	int count = hv_iterinit (hv);
		583
		584	if (SvMAGICAL (hv))
376	{	585	{
		586	// need to count by iterating. could improve by dynamically building the vector below
		587	// but I don't care for the speed of this special case.
		588	// note also that we will run into undefined behaviour when the two iterations
		589	// do not result in the same count, something I might care for in some later release.
		590
		591	count = 0;
		592	while (hv_iternext (hv))
		593	++count;
		594
		595	hv_iterinit (hv);
		596	}
		597
		598	if (count)
		599	{
377	int fast = 1;	600	int i, fast = 1;
378	HE *he;	601	HE *hes_stack [STACK_HES];
379	#if defined(__BORLANDC__) || defined(_MSC_VER)	602	HE **hes = hes_stack;
380	HE **hes = _alloca (count * sizeof (HE));	603
381	#else	604	// allocate larger arrays on the heap
382	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode	605	if (count > STACK_HES)
383	#endif	606	{
		607	SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
		608	hes = (HE **)SvPVX (sv);
		609	}
384		610
385	i = 0;	611	i = 0;
386	while ((he = hv_iternext (hv)))	612	while ((he = hv_iternext (hv)))
387	{	613	{
388	hes [i++] = he;	614	hes [i++] = he;
…		…
410		636
411	FREETMPS;	637	FREETMPS;
412	LEAVE;	638	LEAVE;
413	}	639	}
414		640
415	for (i = 0; i < count; ++i)	641	encode_nl (enc); ++enc->indent;
		642
		643	while (count--)
416	{	644	{
417	encode_indent (enc);	645	encode_indent (enc);
		646	he = hes [count];
418	encode_he (enc, hes [i]);	647	encode_hk (enc, he);
		648	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
419		649
420	if (i < count - 1)	650	if (count)
421	encode_comma (enc);	651	encode_comma (enc);
422	}	652	}
423		653
424	encode_nl (enc);	654	encode_nl (enc); --enc->indent; encode_indent (enc);
425	}	655	}
		656	}
426	else	657	else
		658	{
		659	if (hv_iterinit (hv) || SvMAGICAL (hv))
		660	if ((he = hv_iternext (hv)))
427	{	661	{
428	HE *he = hv_iternext (hv);	662	encode_nl (enc); ++enc->indent;
429		663
430	for (;;)	664	for (;;)
431	{	665	{
432	encode_indent (enc);	666	encode_indent (enc);
433	encode_he (enc, he);	667	encode_hk (enc, he);
		668	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
434		669
435	if (!(he = hv_iternext (hv)))	670	if (!(he = hv_iternext (hv)))
436	break;	671	break;
437		672
438	encode_comma (enc);	673	encode_comma (enc);
439	}	674	}
440		675
441	encode_nl (enc);	676	encode_nl (enc); --enc->indent; encode_indent (enc);
442	}	677	}
443	}	678	}
444		679
445	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	680	encode_ch (enc, '}');
446	}	681	}
447		682
448	// encode objects, arrays and special \0=false and \1=true values.	683	// encode objects, arrays and special \0=false and \1=true values.
449	static void	684	static void
450	encode_rv (enc_t *enc, SV *sv)	685	encode_rv (enc_t *enc, SV *sv)
451	{	686	{
452	svtype svt;	687	svtype svt;
		688	GV *method;
453		689
454	SvGETMAGIC (sv);	690	SvGETMAGIC (sv);
455	svt = SvTYPE (sv);	691	svt = SvTYPE (sv);
456		692
		693	if (expect_false (SvOBJECT (sv)))
		694	{
		695	HV *boolean_stash = !JSON_SLOW || types_boolean_stash
		696	? types_boolean_stash
		697	: gv_stashpv ("Types::Serialiser::Boolean", 1);
		698	HV *stash = SvSTASH (sv);
		699
		700	if (stash == boolean_stash)
		701	{
		702	if (SvIV (sv))
		703	encode_str (enc, "true", 4, 0);
		704	else
		705	encode_str (enc, "false", 5, 0);
		706	}
		707	else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
		708	{
		709	int count;
		710	dSP;
		711
		712	ENTER; SAVETMPS; PUSHMARK (SP);
		713	EXTEND (SP, 2);
		714	// we re-bless the reference to get overload and other niceties right
		715	PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		716	PUSHs (sv_json);
		717
		718	PUTBACK;
		719	count = call_sv ((SV *)GvCV (method), G_ARRAY);
		720	SPAGAIN;
		721
		722	// catch this surprisingly common error
		723	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		724	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		725
		726	encode_ch (enc, '(');
		727	encode_ch (enc, '"');
		728	encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
		729	encode_ch (enc, '"');
		730	encode_ch (enc, ')');
		731	encode_ch (enc, '[');
		732
		733	while (count)
		734	{
		735	encode_sv (enc, SP[1 - count--]);
		736
		737	if (count)
		738	encode_ch (enc, ',');
		739	}
		740
		741	encode_ch (enc, ']');
		742
		743	PUTBACK;
		744
		745	FREETMPS; LEAVE;
		746	}
		747	else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
		748	{
		749	dSP;
		750
		751	ENTER; SAVETMPS; PUSHMARK (SP);
		752	// we re-bless the reference to get overload and other niceties right
		753	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		754
		755	// calling with G_SCALAR ensures that we always get a 1 return value
		756	PUTBACK;
		757	call_sv ((SV *)GvCV (method), G_SCALAR);
		758	SPAGAIN;
		759
		760	// catch this surprisingly common error
		761	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		762	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		763
		764	sv = POPs;
		765	PUTBACK;
		766
		767	encode_sv (enc, sv);
		768
		769	FREETMPS; LEAVE;
		770	}
		771	else if (enc->json.flags & F_ALLOW_BLESSED)
		772	encode_str (enc, "null", 4, 0);
		773	else
		774	croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
		775	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		776	}
457	if (svt == SVt_PVHV)	777	else if (svt == SVt_PVHV)
458	encode_hv (enc, (HV *)sv);	778	encode_hv (enc, (HV *)sv);
459	else if (svt == SVt_PVAV)	779	else if (svt == SVt_PVAV)
460	encode_av (enc, (AV *)sv);	780	encode_av (enc, (AV *)sv);
461	else if (svt < SVt_PVAV)	781	else if (svt < SVt_PVAV)
462	{	782	{
463	if (SvNIOK (sv) && SvIV (sv) == 0)	783	STRLEN len = 0;
		784	char *pv = svt ? SvPV (sv, len) : 0;
		785
		786	if (len == 1 && *pv == '1')
		787	encode_str (enc, "true", 4, 0);
		788	else if (len == 1 && *pv == '0')
464	encode_str (enc, "false", 5, 0);	789	encode_str (enc, "false", 5, 0);
465	else if (SvNIOK (sv) && SvIV (sv) == 1)	790	else if (enc->json.flags & F_ALLOW_UNKNOWN)
466	encode_str (enc, "true", 4, 0);	791	encode_str (enc, "null", 4, 0);
467	else	792	else
468	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	793	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
469	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	794	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
470	}	795	}
		796	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		797	encode_str (enc, "null", 4, 0);
471	else	798	else
472	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	799	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
473	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	800	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
474	}	801	}
475		802
…		…
486	encode_str (enc, str, len, SvUTF8 (sv));	813	encode_str (enc, str, len, SvUTF8 (sv));
487	encode_ch (enc, '"');	814	encode_ch (enc, '"');
488	}	815	}
489	else if (SvNOKp (sv))	816	else if (SvNOKp (sv))
490	{	817	{
		818	// trust that perl will do the right thing w.r.t. JSON syntax.
491	need (enc, NV_DIG + 32);	819	need (enc, NV_DIG + 32);
492	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	820	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
493	enc->cur += strlen (enc->cur);	821	enc->cur += strlen (enc->cur);
494	}	822	}
495	else if (SvIOKp (sv))	823	else if (SvIOKp (sv))
496	{	824	{
		825	// we assume we can always read an IV as a UV and vice versa
		826	// we assume two's complement
		827	// we assume no aliasing issues in the union
		828	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		829	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		830	{
		831	// optimise the "small number case"
		832	// code will likely be branchless and use only a single multiplication
		833	// works for numbers up to 59074
		834	I32 i = SvIVX (sv);
		835	U32 u;
		836	char digit, nz = 0;
		837
497	need (enc, 64);	838	need (enc, 6);
		839
		840	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		841	u = i < 0 ? -i : i;
		842
		843	// convert to 4.28 fixed-point representation
		844	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		845
		846	// now output digit by digit, each time masking out the integer part
		847	// and multiplying by 5 while moving the decimal point one to the right,
		848	// resulting in a net multiplication by 10.
		849	// we always write the digit to memory but conditionally increment
		850	// the pointer, to enable the use of conditional move instructions.
		851	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		852	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		853	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		854	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		855	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		856	}
		857	else
		858	{
		859	// large integer, use the (rather slow) snprintf way.
		860	need (enc, IVUV_MAXCHARS);
498	enc->cur +=	861	enc->cur +=
499	SvIsUV(sv)	862	SvIsUV(sv)
500	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	863	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
501	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	864	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		865	}
502	}	866	}
503	else if (SvROK (sv))	867	else if (SvROK (sv))
504	encode_rv (enc, SvRV (sv));	868	encode_rv (enc, SvRV (sv));
505	else if (!SvOK (sv))	869	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
506	encode_str (enc, "null", 4, 0);	870	encode_str (enc, "null", 4, 0);
507	else	871	else
508	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	872	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
509	SvPV_nolen (sv), SvFLAGS (sv));	873	SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
510	}	874	}
511		875
512	static SV *	876	static SV *
513	encode_json (SV *scalar, U32 flags)	877	encode_json (SV *scalar, JSON *json)
514	{	878	{
515	enc_t enc;	879	enc_t enc;
516		880
517	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	881	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
518	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	882	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
519		883
520	enc.flags = flags;	884	enc.json = *json;
521	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	885	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
522	enc.cur = SvPVX (enc.sv);	886	enc.cur = SvPVX (enc.sv);
523	enc.end = SvEND (enc.sv);	887	enc.end = SvEND (enc.sv);
524	enc.indent = 0;	888	enc.indent = 0;
525	enc.maxdepth = DEC_DEPTH (flags);	889	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		890	: enc.json.flags & F_LATIN1 ? 0x000100UL
		891	: 0x110000UL;
526		892
527	SvPOK_only (enc.sv);	893	SvPOK_only (enc.sv);
528	encode_sv (&enc, scalar);	894	encode_sv (&enc, scalar);
		895	encode_nl (&enc);
529		896
530	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	897	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
531	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	898	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
532		899
533	if (!(flags & (F_ASCII | F_LATIN1 | F_UTF8)))	900	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
534	SvUTF8_on (enc.sv);	901	SvUTF8_on (enc.sv);
535		902
536	if (enc.flags & F_SHRINK)	903	if (enc.json.flags & F_SHRINK)
537	shrink (enc.sv);	904	shrink (enc.sv);
538		905
539	return enc.sv;	906	return enc.sv;
540	}	907	}
541		908
…		…
546	typedef struct	913	typedef struct
547	{	914	{
548	char *cur; // current parser pointer	915	char *cur; // current parser pointer
549	char *end; // end of input string	916	char *end; // end of input string
550	const char *err; // parse error, if != 0	917	const char *err; // parse error, if != 0
551	U32 flags; // F_*	918	JSON json;
552	U32 depth; // recursion depth	919	U32 depth; // recursion depth
553	U32 maxdepth; // recursion depth limit	920	U32 maxdepth; // recursion depth limit
554	} dec_t;	921	} dec_t;
555		922
556	static void	923	INLINE void
		924	decode_comment (dec_t *dec)
		925	{
		926	// only '#'-style comments allowed a.t.m.
		927
		928	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		929	++dec->cur;
		930	}
		931
		932	INLINE void
557	decode_ws (dec_t *dec)	933	decode_ws (dec_t *dec)
558	{	934	{
559	for (;;)	935	for (;;)
560	{	936	{
561	char ch = *dec->cur;	937	char ch = *dec->cur;
562		938
563	if (ch > 0x20	939	if (ch > 0x20)
		940	{
		941	if (expect_false (ch == '#'))
		942	{
		943	if (dec->json.flags & F_RELAXED)
		944	decode_comment (dec);
		945	else
		946	break;
		947	}
		948	else
		949	break;
		950	}
564	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	951	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
565	break;	952	break; // parse error, but let higher level handle it, gives better error messages
566		953
567	++dec->cur;	954	++dec->cur;
568	}	955	}
569	}	956	}
570		957
…		…
574	if (*dec->cur != ch) \	961	if (*dec->cur != ch) \
575	ERR (# ch " expected"); \	962	ERR (# ch " expected"); \
576	++dec->cur; \	963	++dec->cur; \
577	SE	964	SE
578		965
579	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	966	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
580	#define DEC_DEC_DEPTH --dec->depth	967	#define DEC_DEC_DEPTH --dec->depth
581		968
582	static SV *decode_sv (dec_t *dec);	969	static SV *decode_sv (dec_t *dec);
583		970
584	static signed char decode_hexdigit[256];	971	static signed char decode_hexdigit[256];
…		…
587	decode_4hex (dec_t *dec)	974	decode_4hex (dec_t *dec)
588	{	975	{
589	signed char d1, d2, d3, d4;	976	signed char d1, d2, d3, d4;
590	unsigned char *cur = (unsigned char *)dec->cur;	977	unsigned char *cur = (unsigned char *)dec->cur;
591		978
592	d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected");	979	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
593	d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected");	980	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
594	d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected");	981	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
595	d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected");	982	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
596		983
597	dec->cur += 4;	984	dec->cur += 4;
598		985
599	return ((UV)d1) << 12	986	return ((UV)d1) << 12
600	| ((UV)d2) << 8	987	| ((UV)d2) << 8
…		…
608	static SV *	995	static SV *
609	decode_str (dec_t *dec)	996	decode_str (dec_t *dec)
610	{	997	{
611	SV *sv = 0;	998	SV *sv = 0;
612	int utf8 = 0;	999	int utf8 = 0;
		1000	char *dec_cur = dec->cur;
613		1001
614	do	1002	do
615	{	1003	{
616	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];	1004	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
617	char *cur = buf;	1005	char *cur = buf;
618		1006
619	do	1007	do
620	{	1008	{
621	unsigned char ch = *(unsigned char *)dec->cur++;	1009	unsigned char ch = *(unsigned char *)dec_cur++;
622		1010
623	if (ch == '"')	1011	if (expect_false (ch == '"'))
624	{	1012	{
625	--dec->cur;	1013	--dec_cur;
626	break;	1014	break;
627	}	1015	}
628	else if (ch == '\\')	1016	else if (expect_false (ch == '\\'))
629	{	1017	{
630	switch (*dec->cur)	1018	switch (*dec_cur)
631	{	1019	{
632	case '\\':	1020	case '\\':
633	case '/':	1021	case '/':
634	case '"': *cur++ = *dec->cur++; break;	1022	case '"': *cur++ = *dec_cur++; break;
635		1023
636	case 'b': ++dec->cur; *cur++ = '\010'; break;	1024	case 'b': ++dec_cur; *cur++ = '\010'; break;
637	case 't': ++dec->cur; *cur++ = '\011'; break;	1025	case 't': ++dec_cur; *cur++ = '\011'; break;
638	case 'n': ++dec->cur; *cur++ = '\012'; break;	1026	case 'n': ++dec_cur; *cur++ = '\012'; break;
639	case 'f': ++dec->cur; *cur++ = '\014'; break;	1027	case 'f': ++dec_cur; *cur++ = '\014'; break;
640	case 'r': ++dec->cur; *cur++ = '\015'; break;	1028	case 'r': ++dec_cur; *cur++ = '\015'; break;
641		1029
642	case 'u':	1030	case 'u':
643	{	1031	{
644	UV lo, hi;	1032	UV lo, hi;
645	++dec->cur;	1033	++dec_cur;
646		1034
		1035	dec->cur = dec_cur;
647	hi = decode_4hex (dec);	1036	hi = decode_4hex (dec);
		1037	dec_cur = dec->cur;
648	if (hi == (UV)-1)	1038	if (hi == (UV)-1)
649	goto fail;	1039	goto fail;
650		1040
651	// possibly a surrogate pair	1041	// possibly a surrogate pair
652	if (hi >= 0xd800)	1042	if (hi >= 0xd800)
653	if (hi < 0xdc00)	1043	if (hi < 0xdc00)
654	{	1044	{
655	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	1045	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
656	ERR ("missing low surrogate character in surrogate pair");	1046	ERR ("missing low surrogate character in surrogate pair");
657		1047
658	dec->cur += 2;	1048	dec_cur += 2;
659		1049
		1050	dec->cur = dec_cur;
660	lo = decode_4hex (dec);	1051	lo = decode_4hex (dec);
		1052	dec_cur = dec->cur;
661	if (lo == (UV)-1)	1053	if (lo == (UV)-1)
662	goto fail;	1054	goto fail;
663		1055
664	if (lo < 0xdc00 || lo >= 0xe000)	1056	if (lo < 0xdc00 || lo >= 0xe000)
665	ERR ("surrogate pair expected");	1057	ERR ("surrogate pair expected");
…		…
671		1063
672	if (hi >= 0x80)	1064	if (hi >= 0x80)
673	{	1065	{
674	utf8 = 1;	1066	utf8 = 1;
675		1067
676	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	1068	cur = encode_utf8 (cur, hi);
677	}	1069	}
678	else	1070	else
679	*cur++ = hi;	1071	*cur++ = hi;
680	}	1072	}
681	break;	1073	break;
682		1074
683	default:	1075	default:
684	--dec->cur;	1076	--dec_cur;
685	ERR ("illegal backslash escape sequence in string");	1077	ERR ("illegal backslash escape sequence in string");
686	}	1078	}
687	}	1079	}
688	else if (ch >= 0x20 && ch <= 0x7f)	1080	else if (expect_true (ch >= 0x20 && ch < 0x80))
689	*cur++ = ch;	1081	*cur++ = ch;
690	else if (ch >= 0x80)	1082	else if (ch >= 0x80)
691	{	1083	{
692	STRLEN clen;	1084	STRLEN clen;
693	UV uch;
694		1085
695	--dec->cur;	1086	--dec_cur;
696		1087
697	uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	1088	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
698	if (clen == (STRLEN)-1)	1089	if (clen == (STRLEN)-1)
699	ERR ("malformed UTF-8 character in JSON string");	1090	ERR ("malformed UTF-8 character in JSON string");
700		1091
701	do	1092	do
702	*cur++ = *dec->cur++;	1093	*cur++ = *dec_cur++;
703	while (--clen);	1094	while (--clen);
704		1095
705	utf8 = 1;	1096	utf8 = 1;
706	}	1097	}
707	else	1098	else
708	{	1099	{
709	--dec->cur;	1100	--dec_cur;
710		1101
711	if (!ch)	1102	if (!ch)
712	ERR ("unexpected end of string while parsing JSON string");	1103	ERR ("unexpected end of string while parsing JSON string");
713	else	1104	else
714	ERR ("invalid character encountered while parsing JSON string");	1105	ERR ("invalid character encountered while parsing JSON string");
…		…
719	{	1110	{
720	STRLEN len = cur - buf;	1111	STRLEN len = cur - buf;
721		1112
722	if (sv)	1113	if (sv)
723	{	1114	{
724	SvGROW (sv, SvCUR (sv) + len + 1);	1115	STRLEN cur = SvCUR (sv);
		1116
		1117	if (SvLEN (sv) <= cur + len)
		1118	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1119
725	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	1120	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
726	SvCUR_set (sv, SvCUR (sv) + len);	1121	SvCUR_set (sv, SvCUR (sv) + len);
727	}	1122	}
728	else	1123	else
729	sv = newSVpvn (buf, len);	1124	sv = newSVpvn (buf, len);
730	}	1125	}
731	}	1126	}
732	while (*dec->cur != '"');	1127	while (*dec_cur != '"');
733		1128
734	++dec->cur;	1129	++dec_cur;
735		1130
736	if (sv)	1131	if (sv)
737	{	1132	{
738	SvPOK_only (sv);	1133	SvPOK_only (sv);
739	*SvEND (sv) = 0;	1134	*SvEND (sv) = 0;
…		…
742	SvUTF8_on (sv);	1137	SvUTF8_on (sv);
743	}	1138	}
744	else	1139	else
745	sv = newSVpvn ("", 0);	1140	sv = newSVpvn ("", 0);
746		1141
		1142	dec->cur = dec_cur;
747	return sv;	1143	return sv;
748		1144
749	fail:	1145	fail:
		1146	dec->cur = dec_cur;
750	return 0;	1147	return 0;
751	}	1148	}
752		1149
753	static SV *	1150	static SV *
754	decode_num (dec_t *dec)	1151	decode_num (dec_t *dec)
…		…
812	is_nv = 1;	1209	is_nv = 1;
813	}	1210	}
814		1211
815	if (!is_nv)	1212	if (!is_nv)
816	{	1213	{
817	UV uv;	1214	int len = dec->cur - start;
818	int numtype = grok_number (start, dec->cur - start, &uv);	1215
819	if (numtype & IS_NUMBER_IN_UV)	1216	// special case the rather common 1..5-digit-int case
820	if (numtype & IS_NUMBER_NEG)	1217	if (*start == '-')
		1218	switch (len)
821	{	1219	{
822	if (uv < (UV)IV_MIN)	1220	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
823	return newSViv (-(IV)uv);	1221	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
		1222	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1223	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1224	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
824	}	1225	}
		1226	else
		1227	switch (len)
		1228	{
		1229	case 1: return newSViv ( start [0] - '0' * 1);
		1230	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1231	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1232	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1233	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1234	}
		1235
		1236	{
		1237	UV uv;
		1238	int numtype = grok_number (start, len, &uv);
		1239	if (numtype & IS_NUMBER_IN_UV)
		1240	if (numtype & IS_NUMBER_NEG)
		1241	{
		1242	if (uv < (UV)IV_MIN)
		1243	return newSViv (-(IV)uv);
		1244	}
825	else	1245	else
826	return newSVuv (uv);	1246	return newSVuv (uv);
827	}	1247	}
828		1248
		1249	len -= *start == '-' ? 1 : 0;
		1250
		1251	// does not fit into IV or UV, try NV
		1252	if (len <= NV_DIG)
		1253	// fits into NV without loss of precision
		1254	return newSVnv (json_atof (start));
		1255
		1256	// everything else fails, convert it to a string
		1257	return newSVpvn (start, dec->cur - start);
		1258	}
		1259
		1260	// loss of precision here
829	return newSVnv (Atof (start));	1261	return newSVnv (json_atof (start));
830		1262
831	fail:	1263	fail:
832	return 0;	1264	return 0;
833	}	1265	}
834		1266
…		…
858	if (*dec->cur == ']')	1290	if (*dec->cur == ']')
859	{	1291	{
860	++dec->cur;	1292	++dec->cur;
861	break;	1293	break;
862	}	1294	}
863		1295
864	if (*dec->cur != ',')	1296	if (*dec->cur != ',')
865	ERR (", or ] expected while parsing array");	1297	ERR (", or ] expected while parsing array");
866		1298
867	++dec->cur;	1299	++dec->cur;
		1300
		1301	decode_ws (dec);
		1302
		1303	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1304	{
		1305	++dec->cur;
		1306	break;
		1307	}
868	}	1308	}
869		1309
870	DEC_DEC_DEPTH;	1310	DEC_DEC_DEPTH;
871	return newRV_noinc ((SV *)av);	1311	return newRV_noinc ((SV *)av);
872		1312
…		…
877	}	1317	}
878		1318
879	static SV *	1319	static SV *
880	decode_hv (dec_t *dec)	1320	decode_hv (dec_t *dec)
881	{	1321	{
		1322	SV *sv;
882	HV *hv = newHV ();	1323	HV *hv = newHV ();
883		1324
884	DEC_INC_DEPTH;	1325	DEC_INC_DEPTH;
885	decode_ws (dec);	1326	decode_ws (dec);
886		1327
887	if (*dec->cur == '}')	1328	if (*dec->cur == '}')
888	++dec->cur;	1329	++dec->cur;
889	else	1330	else
890	for (;;)	1331	for (;;)
891	{	1332	{
		1333	EXPECT_CH ('"');
		1334
		1335	// heuristic: assume that
		1336	// a) decode_str + hv_store_ent are abysmally slow.
		1337	// b) most hash keys are short, simple ascii text.
		1338	// => try to "fast-match" such strings to avoid
		1339	// the overhead of decode_str + hv_store_ent.
		1340	{
892	SV *key, *value;	1341	SV *value;
		1342	char *p = dec->cur;
		1343	char *e = p + 24; // only try up to 24 bytes
893		1344
894	decode_ws (dec); EXPECT_CH ('"');	1345	for (;;)
895
896	key = decode_str (dec);
897	if (!key)
898	goto fail;
899
900	decode_ws (dec); EXPECT_CH (':');
901
902	value = decode_sv (dec);
903	if (!value)
904	{	1346	{
		1347	// the >= 0x80 is false on most architectures
		1348	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1349	{
		1350	// slow path, back up and use decode_str
		1351	SV *key = decode_str (dec);
		1352	if (!key)
		1353	goto fail;
		1354
		1355	decode_ws (dec); EXPECT_CH (':');
		1356
		1357	decode_ws (dec);
		1358	value = decode_sv (dec);
		1359	if (!value)
		1360	{
		1361	SvREFCNT_dec (key);
		1362	goto fail;
		1363	}
		1364
		1365	hv_store_ent (hv, key, value, 0);
905	SvREFCNT_dec (key);	1366	SvREFCNT_dec (key);
		1367
		1368	break;
		1369	}
		1370	else if (*p == '"')
		1371	{
		1372	// fast path, got a simple key
		1373	char *key = dec->cur;
		1374	int len = p - key;
		1375	dec->cur = p + 1;
		1376
		1377	decode_ws (dec); EXPECT_CH (':');
		1378
		1379	decode_ws (dec);
		1380	value = decode_sv (dec);
		1381	if (!value)
906	goto fail;	1382	goto fail;
		1383
		1384	hv_store (hv, key, len, value, 0);
		1385
		1386	break;
		1387	}
		1388
		1389	++p;
907	}	1390	}
908		1391	}
909	hv_store_ent (hv, key, value, 0);
910	SvREFCNT_dec (key);
911		1392
912	decode_ws (dec);	1393	decode_ws (dec);
913		1394
914	if (*dec->cur == '}')	1395	if (*dec->cur == '}')
915	{	1396	{
…		…
919		1400
920	if (*dec->cur != ',')	1401	if (*dec->cur != ',')
921	ERR (", or } expected while parsing object/hash");	1402	ERR (", or } expected while parsing object/hash");
922		1403
923	++dec->cur;	1404	++dec->cur;
		1405
		1406	decode_ws (dec);
		1407
		1408	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1409	{
		1410	++dec->cur;
		1411	break;
		1412	}
924	}	1413	}
925		1414
926	DEC_DEC_DEPTH;	1415	DEC_DEC_DEPTH;
927	return newRV_noinc ((SV *)hv);	1416	sv = newRV_noinc ((SV *)hv);
		1417
		1418	// check filter callbacks
		1419	if (dec->json.flags & F_HOOK)
		1420	{
		1421	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1422	{
		1423	HE *cb, *he;
		1424
		1425	hv_iterinit (hv);
		1426	he = hv_iternext (hv);
		1427	hv_iterinit (hv);
		1428
		1429	// the next line creates a mortal sv each time its called.
		1430	// might want to optimise this for common cases.
		1431	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1432
		1433	if (cb)
		1434	{
		1435	dSP;
		1436	int count;
		1437
		1438	ENTER; SAVETMPS; PUSHMARK (SP);
		1439	XPUSHs (HeVAL (he));
		1440	sv_2mortal (sv);
		1441
		1442	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1443
		1444	if (count == 1)
		1445	{
		1446	sv = newSVsv (POPs);
		1447	FREETMPS; LEAVE;
		1448	return sv;
		1449	}
		1450
		1451	SvREFCNT_inc (sv);
		1452	FREETMPS; LEAVE;
		1453	}
		1454	}
		1455
		1456	if (dec->json.cb_object)
		1457	{
		1458	dSP;
		1459	int count;
		1460
		1461	ENTER; SAVETMPS; PUSHMARK (SP);
		1462	XPUSHs (sv_2mortal (sv));
		1463
		1464	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1465
		1466	if (count == 1)
		1467	{
		1468	sv = newSVsv (POPs);
		1469	FREETMPS; LEAVE;
		1470	return sv;
		1471	}
		1472
		1473	SvREFCNT_inc (sv);
		1474	FREETMPS; LEAVE;
		1475	}
		1476	}
		1477
		1478	return sv;
928		1479
929	fail:	1480	fail:
930	SvREFCNT_dec (hv);	1481	SvREFCNT_dec (hv);
931	DEC_DEC_DEPTH;	1482	DEC_DEC_DEPTH;
932	return 0;	1483	return 0;
933	}	1484	}
934		1485
935	static SV *	1486	static SV *
		1487	decode_tag (dec_t *dec)
		1488	{
		1489	SV *tag = 0;
		1490	SV *val = 0;
		1491
		1492	if (!(dec->json.flags & F_ALLOW_TAGS))
		1493	ERR ("malformed JSON string, neither array, object, number, string or atom");
		1494
		1495	++dec->cur;
		1496
		1497	decode_ws (dec);
		1498
		1499	tag = decode_sv (dec);
		1500	if (!tag)
		1501	goto fail;
		1502
		1503	if (!SvPOK (tag))
		1504	ERR ("malformed JSON string, (tag) must be a string");
		1505
		1506	decode_ws (dec);
		1507
		1508	if (*dec->cur != ')')
		1509	ERR (") expected after tag");
		1510
		1511	++dec->cur;
		1512
		1513	decode_ws (dec);
		1514
		1515	val = decode_sv (dec);
		1516	if (!val)
		1517	goto fail;
		1518
		1519	if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
		1520	ERR ("malformed JSON string, tag value must be an array");
		1521
		1522	{
		1523	AV *av = (AV *)SvRV (val);
		1524	int i, len = av_len (av) + 1;
		1525	HV *stash = gv_stashsv (tag, 0);
		1526	SV *sv;
		1527
		1528	if (!stash)
		1529	ERR ("cannot decode perl-object (package does not exist)");
		1530
		1531	GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
		1532
		1533	if (!method)
		1534	ERR ("cannot decode perl-object (package does not have a THAW method)");
		1535
		1536	dSP;
		1537
		1538	ENTER; SAVETMPS; PUSHMARK (SP);
		1539	EXTEND (SP, len + 2);
		1540	// we re-bless the reference to get overload and other niceties right
		1541	PUSHs (tag);
		1542	PUSHs (sv_json);
		1543
		1544	for (i = 0; i < len; ++i)
		1545	PUSHs (*av_fetch (av, i, 1));
		1546
		1547	PUTBACK;
		1548	call_sv ((SV *)GvCV (method), G_SCALAR);
		1549	SPAGAIN;
		1550
		1551	SvREFCNT_dec (tag);
		1552	SvREFCNT_dec (val);
		1553	sv = SvREFCNT_inc (POPs);
		1554
		1555	PUTBACK;
		1556
		1557	FREETMPS; LEAVE;
		1558
		1559	return sv;
		1560	}
		1561
		1562	fail:
		1563	SvREFCNT_dec (tag);
		1564	SvREFCNT_dec (val);
		1565	return 0;
		1566	}
		1567
		1568	static SV *
936	decode_sv (dec_t *dec)	1569	decode_sv (dec_t *dec)
937	{	1570	{
938	decode_ws (dec);	1571	// the beauty of JSON: you need exactly one character lookahead
		1572	// to parse everything.
939	switch (*dec->cur)	1573	switch (*dec->cur)
940	{	1574	{
941	case '"': ++dec->cur; return decode_str (dec);	1575	case '"': ++dec->cur; return decode_str (dec);
942	case '[': ++dec->cur; return decode_av (dec);	1576	case '[': ++dec->cur; return decode_av (dec);
943	case '{': ++dec->cur; return decode_hv (dec);	1577	case '{': ++dec->cur; return decode_hv (dec);
		1578	case '(': return decode_tag (dec);
944		1579
945	case '-':	1580	case '-':
946	case '0': case '1': case '2': case '3': case '4':	1581	case '0': case '1': case '2': case '3': case '4':
947	case '5': case '6': case '7': case '8': case '9':	1582	case '5': case '6': case '7': case '8': case '9':
948	return decode_num (dec);	1583	return decode_num (dec);
949		1584
950	case 't':	1585	case 't':
951	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1586	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
952	{	1587	{
953	dec->cur += 4;	1588	dec->cur += 4;
		1589	#if JSON_SLOW
		1590	types_true = get_bool ("Types::Serialiser::true");
		1591	#endif
954	return newSViv (1);	1592	return newSVsv (types_true);
955	}	1593	}
956	else	1594	else
957	ERR ("'true' expected");	1595	ERR ("'true' expected");
958		1596
959	break;	1597	break;
960		1598
961	case 'f':	1599	case 'f':
962	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1600	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
963	{	1601	{
964	dec->cur += 5;	1602	dec->cur += 5;
		1603	#if JSON_SLOW
		1604	types_false = get_bool ("Types::Serialiser::false");
		1605	#endif
965	return newSViv (0);	1606	return newSVsv (types_false);
966	}	1607	}
967	else	1608	else
968	ERR ("'false' expected");	1609	ERR ("'false' expected");
969		1610
970	break;	1611	break;
…		…
979	ERR ("'null' expected");	1620	ERR ("'null' expected");
980		1621
981	break;	1622	break;
982		1623
983	default:	1624	default:
984	ERR ("malformed JSON string, neither array, object, number, string or atom");	1625	ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
985	break;	1626	break;
986	}	1627	}
987		1628
988	fail:	1629	fail:
989	return 0;	1630	return 0;
990	}	1631	}
991		1632
992	static SV *	1633	static SV *
993	decode_json (SV *string, U32 flags, UV *offset_return)	1634	decode_json (SV *string, JSON *json, char **offset_return)
994	{	1635	{
995	dec_t dec;	1636	dec_t dec;
996	UV offset;
997	SV *sv;	1637	SV *sv;
998		1638
		1639	/* work around bugs in 5.10 where manipulating magic values
		1640	* makes perl ignore the magic in subsequent accesses.
		1641	* also make a copy of non-PV values, to get them into a clean
		1642	* state (SvPV should do that, but it's buggy, see below).
		1643	*/
999	SvGETMAGIC (string);	1644	/*SvGETMAGIC (string);*/
		1645	if (SvMAGICAL (string) || !SvPOK (string))
		1646	string = sv_2mortal (newSVsv (string));
		1647
1000	SvUPGRADE (string, SVt_PV);	1648	SvUPGRADE (string, SVt_PV);
1001		1649
1002	if (flags & F_UTF8)	1650	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1651	* assertion with -DDEBUGGING, although SvCUR is documented to
		1652	* return the xpv_cur field which certainly exists after upgrading.
		1653	* according to nicholas clark, calling SvPOK fixes this.
		1654	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1655	* and hope for the best.
		1656	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1657	* assertion business is seriously broken, try yet another workaround
		1658	* for the broken -DDEBUGGING.
		1659	*/
		1660	{
		1661	#ifdef DEBUGGING
		1662	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1663	#else
		1664	STRLEN offset = SvCUR (string);
		1665	#endif
		1666
		1667	if (offset > json->max_size && json->max_size)
		1668	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1669	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1670	}
		1671
		1672	if (DECODE_WANTS_OCTETS (json))
1003	sv_utf8_downgrade (string, 0);	1673	sv_utf8_downgrade (string, 0);
1004	else	1674	else
1005	sv_utf8_upgrade (string);	1675	sv_utf8_upgrade (string);
1006		1676
1007	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1677	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1008		1678
1009	dec.flags = flags;	1679	dec.json = *json;
1010	dec.cur = SvPVX (string);	1680	dec.cur = SvPVX (string);
1011	dec.end = SvEND (string);	1681	dec.end = SvEND (string);
1012	dec.err = 0;	1682	dec.err = 0;
1013	dec.depth = 0;	1683	dec.depth = 0;
1014	dec.maxdepth = DEC_DEPTH (dec.flags);	1684
		1685	if (dec.json.cb_object || dec.json.cb_sk_object)
		1686	dec.json.flags |= F_HOOK;
1015		1687
1016	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1688	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1689
		1690	decode_ws (&dec);
1017	sv = decode_sv (&dec);	1691	sv = decode_sv (&dec);
		1692
		1693	if (offset_return)
		1694	*offset_return = dec.cur;
1018		1695
1019	if (!(offset_return || !sv))	1696	if (!(offset_return || !sv))
1020	{	1697	{
1021	// check for trailing garbage	1698	// check for trailing garbage
1022	decode_ws (&dec);	1699	decode_ws (&dec);
…		…
1025	{	1702	{
1026	dec.err = "garbage after JSON object";	1703	dec.err = "garbage after JSON object";
1027	SvREFCNT_dec (sv);	1704	SvREFCNT_dec (sv);
1028	sv = 0;	1705	sv = 0;
1029	}	1706	}
1030	}
1031
1032	if (offset_return || !sv)
1033	{
1034	offset = dec.flags & F_UTF8
1035	? dec.cur - SvPVX (string)
1036	: utf8_distance (dec.cur, SvPVX (string));
1037
1038	if (offset_return)
1039	*offset_return = offset;
1040	}	1707	}
1041		1708
1042	if (!sv)	1709	if (!sv)
1043	{	1710	{
1044	SV *uni = sv_newmortal ();	1711	SV *uni = sv_newmortal ();
…		…
1050	SAVEVPTR (PL_curcop);	1717	SAVEVPTR (PL_curcop);
1051	PL_curcop = &cop;	1718	PL_curcop = &cop;
1052	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1719	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1053	LEAVE;	1720	LEAVE;
1054		1721
1055	croak ("%s, at character offset %d [\"%s\"]",	1722	croak ("%s, at character offset %d (before \"%s\")",
1056	dec.err,	1723	dec.err,
1057	(int)offset,	1724	(int)ptr_to_index (string, dec.cur),
1058	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1725	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1059	}	1726	}
1060		1727
1061	sv = sv_2mortal (sv);	1728	sv = sv_2mortal (sv);
1062		1729
1063	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1730	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1064	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1731	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1065		1732
1066	return sv;	1733	return sv;
		1734	}
		1735
		1736	/////////////////////////////////////////////////////////////////////////////
		1737	// incremental parser
		1738
		1739	static void
		1740	incr_parse (JSON *self)
		1741	{
		1742	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1743
		1744	// the state machine here is a bit convoluted and could be simplified a lot
		1745	// but this would make it slower, so...
		1746
		1747	for (;;)
		1748	{
		1749	//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
		1750	switch (self->incr_mode)
		1751	{
		1752	// only used for initial whitespace skipping
		1753	case INCR_M_WS:
		1754	for (;;)
		1755	{
		1756	if (*p > 0x20)
		1757	{
		1758	if (*p == '#')
		1759	{
		1760	self->incr_mode = INCR_M_C0;
		1761	goto incr_m_c;
		1762	}
		1763	else
		1764	{
		1765	self->incr_mode = INCR_M_JSON;
		1766	goto incr_m_json;
		1767	}
		1768	}
		1769	else if (!*p)
		1770	goto interrupt;
		1771
		1772	++p;
		1773	}
		1774
		1775	// skip a single char inside a string (for \\-processing)
		1776	case INCR_M_BS:
		1777	if (!*p)
		1778	goto interrupt;
		1779
		1780	++p;
		1781	self->incr_mode = INCR_M_STR;
		1782	goto incr_m_str;
		1783
		1784	// inside #-style comments
		1785	case INCR_M_C0:
		1786	case INCR_M_C1:
		1787	incr_m_c:
		1788	for (;;)
		1789	{
		1790	if (*p == '\n')
		1791	{
		1792	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1793	break;
		1794	}
		1795	else if (!*p)
		1796	goto interrupt;
		1797
		1798	++p;
		1799	}
		1800
		1801	break;
		1802
		1803	// inside a string
		1804	case INCR_M_STR:
		1805	incr_m_str:
		1806	for (;;)
		1807	{
		1808	if (*p == '"')
		1809	{
		1810	++p;
		1811	self->incr_mode = INCR_M_JSON;
		1812
		1813	if (!self->incr_nest)
		1814	goto interrupt;
		1815
		1816	goto incr_m_json;
		1817	}
		1818	else if (*p == '\\')
		1819	{
		1820	++p; // "virtually" consumes character after \
		1821
		1822	if (!*p) // if at end of string we have to switch modes
		1823	{
		1824	self->incr_mode = INCR_M_BS;
		1825	goto interrupt;
		1826	}
		1827	}
		1828	else if (!*p)
		1829	goto interrupt;
		1830
		1831	++p;
		1832	}
		1833
		1834	// after initial ws, outside string
		1835	case INCR_M_JSON:
		1836	incr_m_json:
		1837	for (;;)
		1838	{
		1839	switch (*p++)
		1840	{
		1841	case 0:
		1842	--p;
		1843	goto interrupt;
		1844
		1845	case 0x09:
		1846	case 0x0a:
		1847	case 0x0d:
		1848	case 0x20:
		1849	if (!self->incr_nest)
		1850	{
		1851	--p; // do not eat the whitespace, let the next round do it
		1852	goto interrupt;
		1853	}
		1854	break;
		1855
		1856	case '"':
		1857	self->incr_mode = INCR_M_STR;
		1858	goto incr_m_str;
		1859
		1860	case '[':
		1861	case '{':
		1862	case '(':
		1863	if (++self->incr_nest > self->max_depth)
		1864	croak (ERR_NESTING_EXCEEDED);
		1865	break;
		1866
		1867	case ']':
		1868	case '}':
		1869	if (--self->incr_nest <= 0)
		1870	goto interrupt;
		1871	break;
		1872
		1873	case ')':
		1874	--self->incr_nest;
		1875	break;
		1876
		1877	case '#':
		1878	self->incr_mode = INCR_M_C1;
		1879	goto incr_m_c;
		1880	}
		1881	}
		1882	}
		1883
		1884	modechange:
		1885	;
		1886	}
		1887
		1888	interrupt:
		1889	self->incr_pos = p - SvPVX (self->incr_text);
		1890	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1891	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1067	}	1892	}
1068		1893
1069	/////////////////////////////////////////////////////////////////////////////	1894	/////////////////////////////////////////////////////////////////////////////
1070	// XS interface functions	1895	// XS interface functions
1071		1896
1072	MODULE = JSON::XS PACKAGE = JSON::XS	1897	MODULE = JSON::XS PACKAGE = JSON::XS
1073		1898
1074	BOOT:	1899	BOOT:
1075	{	1900	{
1076	int i;	1901	int i;
1077
1078	memset (decode_hexdigit, 0xff, 256);
1079		1902
1080	for (i = 0; i < 256; ++i)	1903	for (i = 0; i < 256; ++i)
1081	decode_hexdigit [i] =	1904	decode_hexdigit [i] =
1082	i >= '0' && i <= '9' ? i - '0'	1905	i >= '0' && i <= '9' ? i - '0'
1083	: i >= 'a' && i <= 'f' ? i - 'a' + 10	1906	: i >= 'a' && i <= 'f' ? i - 'a' + 10
1084	: i >= 'A' && i <= 'F' ? i - 'A' + 10	1907	: i >= 'A' && i <= 'F' ? i - 'A' + 10
1085	: -1;	1908	: -1;
1086		1909
1087	json_stash = gv_stashpv ("JSON::XS", 1);	1910	json_stash = gv_stashpv ("JSON::XS" , 1);
		1911	types_boolean_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
		1912
		1913	types_true = get_bool ("Types::Serialiser::true");
		1914	types_false = get_bool ("Types::Serialiser::false");
		1915
		1916	sv_json = newSVpv ("JSON", 0);
		1917	SvREADONLY_on (sv_json);
		1918
		1919	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1088	}	1920	}
1089		1921
1090	PROTOTYPES: DISABLE	1922	PROTOTYPES: DISABLE
1091		1923
1092	SV *new (char *dummy)	1924	void CLONE (...)
1093	CODE:	1925	CODE:
1094	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1926	json_stash = 0;
		1927	types_boolean_stash = 0;
		1928
		1929	void new (char *klass)
		1930	PPCODE:
		1931	{
		1932	SV *pv = NEWSV (0, sizeof (JSON));
		1933	SvPOK_only (pv);
		1934	json_init ((JSON *)SvPVX (pv));
		1935	XPUSHs (sv_2mortal (sv_bless (
		1936	newRV_noinc (pv),
		1937	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1938	)));
		1939	}
		1940
		1941	void ascii (JSON *self, int enable = 1)
		1942	ALIAS:
		1943	ascii = F_ASCII
		1944	latin1 = F_LATIN1
		1945	utf8 = F_UTF8
		1946	indent = F_INDENT
		1947	canonical = F_CANONICAL
		1948	space_before = F_SPACE_BEFORE
		1949	space_after = F_SPACE_AFTER
		1950	pretty = F_PRETTY
		1951	allow_nonref = F_ALLOW_NONREF
		1952	shrink = F_SHRINK
		1953	allow_blessed = F_ALLOW_BLESSED
		1954	convert_blessed = F_CONV_BLESSED
		1955	relaxed = F_RELAXED
		1956	allow_unknown = F_ALLOW_UNKNOWN
		1957	allow_tags = F_ALLOW_TAGS
		1958	PPCODE:
		1959	{
		1960	if (enable)
		1961	self->flags |= ix;
		1962	else
		1963	self->flags &= ~ix;
		1964
		1965	XPUSHs (ST (0));
		1966	}
		1967
		1968	void get_ascii (JSON *self)
		1969	ALIAS:
		1970	get_ascii = F_ASCII
		1971	get_latin1 = F_LATIN1
		1972	get_utf8 = F_UTF8
		1973	get_indent = F_INDENT
		1974	get_canonical = F_CANONICAL
		1975	get_space_before = F_SPACE_BEFORE
		1976	get_space_after = F_SPACE_AFTER
		1977	get_allow_nonref = F_ALLOW_NONREF
		1978	get_shrink = F_SHRINK
		1979	get_allow_blessed = F_ALLOW_BLESSED
		1980	get_convert_blessed = F_CONV_BLESSED
		1981	get_relaxed = F_RELAXED
		1982	get_allow_unknown = F_ALLOW_UNKNOWN
		1983	get_allow_tags = F_ALLOW_TAGS
		1984	PPCODE:
		1985	XPUSHs (boolSV (self->flags & ix));
		1986
		1987	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		1988	PPCODE:
		1989	self->max_depth = max_depth;
		1990	XPUSHs (ST (0));
		1991
		1992	U32 get_max_depth (JSON *self)
		1993	CODE:
		1994	RETVAL = self->max_depth;
1095	OUTPUT:	1995	OUTPUT:
1096	RETVAL	1996	RETVAL
1097		1997
1098	SV *ascii (SV *self, int enable = 1)	1998	void max_size (JSON *self, U32 max_size = 0)
1099	ALIAS:	1999	PPCODE:
1100	ascii = F_ASCII	2000	self->max_size = max_size;
1101	latin1 = F_LATIN1	2001	XPUSHs (ST (0));
1102	utf8 = F_UTF8	2002
1103	indent = F_INDENT	2003	int get_max_size (JSON *self)
1104	canonical = F_CANONICAL
1105	space_before = F_SPACE_BEFORE
1106	space_after = F_SPACE_AFTER
1107	pretty = F_PRETTY
1108	allow_nonref = F_ALLOW_NONREF
1109	shrink = F_SHRINK
1110	CODE:	2004	CODE:
1111	{	2005	RETVAL = self->max_size;
1112	UV *uv = SvJSON (self);
1113	if (enable)
1114	*uv |= ix;
1115	else
1116	*uv &= ~ix;
1117
1118	RETVAL = newSVsv (self);
1119	}
1120	OUTPUT:	2006	OUTPUT:
1121	RETVAL	2007	RETVAL
1122		2008
1123	SV *max_depth (SV *self, UV max_depth = 0x80000000UL)	2009	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
		2010	PPCODE:
		2011	{
		2012	SvREFCNT_dec (self->cb_object);
		2013	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		2014
		2015	XPUSHs (ST (0));
		2016	}
		2017
		2018	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
		2019	PPCODE:
		2020	{
		2021	if (!self->cb_sk_object)
		2022	self->cb_sk_object = newHV ();
		2023
		2024	if (SvOK (cb))
		2025	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		2026	else
		2027	{
		2028	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		2029
		2030	if (!HvKEYS (self->cb_sk_object))
		2031	{
		2032	SvREFCNT_dec (self->cb_sk_object);
		2033	self->cb_sk_object = 0;
		2034	}
		2035	}
		2036
		2037	XPUSHs (ST (0));
		2038	}
		2039
		2040	void encode (JSON *self, SV *scalar)
		2041	PPCODE:
		2042	PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
		2043	XPUSHs (scalar);
		2044
		2045	void decode (JSON *self, SV *jsonstr)
		2046	PPCODE:
		2047	PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
		2048	XPUSHs (jsonstr);
		2049
		2050	void decode_prefix (JSON *self, SV *jsonstr)
		2051	PPCODE:
		2052	{
		2053	SV *sv;
		2054	char *offset;
		2055	PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
		2056	EXTEND (SP, 2);
		2057	PUSHs (sv);
		2058	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		2059	}
		2060
		2061	void incr_parse (JSON *self, SV *jsonstr = 0)
		2062	PPCODE:
		2063	{
		2064	if (!self->incr_text)
		2065	self->incr_text = newSVpvn ("", 0);
		2066
		2067	/* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
		2068	if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
		2069	if (DECODE_WANTS_OCTETS (self))
		2070	{
		2071	if (self->incr_pos)
		2072	self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
		2073	(U8 *)SvPVX (self->incr_text) + self->incr_pos);
		2074
		2075	sv_utf8_downgrade (self->incr_text, 0);
		2076	}
		2077	else
		2078	{
		2079	sv_utf8_upgrade (self->incr_text);
		2080
		2081	if (self->incr_pos)
		2082	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		2083	- (U8 *)SvPVX (self->incr_text);
		2084	}
		2085
		2086	// append data, if any
		2087	if (jsonstr)
		2088	{
		2089	/* make sure both strings have same encoding */
		2090	if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
		2091	if (SvUTF8 (jsonstr))
		2092	sv_utf8_downgrade (jsonstr, 0);
		2093	else
		2094	sv_utf8_upgrade (jsonstr);
		2095
		2096	/* and then just blindly append */
		2097	{
		2098	STRLEN len;
		2099	const char *str = SvPV (jsonstr, len);
		2100	STRLEN cur = SvCUR (self->incr_text);
		2101
		2102	if (SvLEN (self->incr_text) <= cur + len)
		2103	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		2104
		2105	Move (str, SvEND (self->incr_text), len, char);
		2106	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		2107	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		2108	}
		2109	}
		2110
		2111	if (GIMME_V != G_VOID)
		2112	do
		2113	{
		2114	SV *sv;
		2115	char *offset;
		2116
		2117	if (!INCR_DONE (self))
		2118	{
		2119	incr_parse (self);
		2120
		2121	if (self->incr_pos > self->max_size && self->max_size)
		2122	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		2123	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		2124
		2125	if (!INCR_DONE (self))
		2126	{
		2127	// as an optimisation, do not accumulate white space in the incr buffer
		2128	if (self->incr_mode == INCR_M_WS && self->incr_pos)
		2129	{
		2130	self->incr_pos = 0;
		2131	SvCUR_set (self->incr_text, 0);
		2132	}
		2133
		2134	break;
		2135	}
		2136	}
		2137
		2138	PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
		2139	XPUSHs (sv);
		2140
		2141	self->incr_pos -= offset - SvPVX (self->incr_text);
		2142	self->incr_nest = 0;
		2143	self->incr_mode = 0;
		2144
		2145	sv_chop (self->incr_text, offset);
		2146	}
		2147	while (GIMME_V == G_ARRAY);
		2148	}
		2149
		2150	SV *incr_text (JSON *self)
		2151	ATTRS: lvalue
1124	CODE:	2152	CODE:
1125	{	2153	{
1126	UV *uv = SvJSON (self);	2154	if (self->incr_pos)
1127	UV log2 = 0;	2155	croak ("incr_text can not be called when the incremental parser already started parsing");
1128		2156
1129	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	2157	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
1130
1131	while ((1UL << log2) < max_depth)
1132	++log2;
1133
1134	*uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1135
1136	RETVAL = newSVsv (self);
1137	}	2158	}
1138	OUTPUT:	2159	OUTPUT:
1139	RETVAL	2160	RETVAL
1140		2161
		2162	void incr_skip (JSON *self)
		2163	CODE:
		2164	{
		2165	if (self->incr_pos)
		2166	{
		2167	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2168	self->incr_pos = 0;
		2169	self->incr_nest = 0;
		2170	self->incr_mode = 0;
		2171	}
		2172	}
		2173
		2174	void incr_reset (JSON *self)
		2175	CODE:
		2176	{
		2177	SvREFCNT_dec (self->incr_text);
		2178	self->incr_text = 0;
		2179	self->incr_pos = 0;
		2180	self->incr_nest = 0;
		2181	self->incr_mode = 0;
		2182	}
		2183
		2184	void DESTROY (JSON *self)
		2185	CODE:
		2186	SvREFCNT_dec (self->cb_sk_object);
		2187	SvREFCNT_dec (self->cb_object);
		2188	SvREFCNT_dec (self->incr_text);
		2189
		2190	PROTOTYPES: ENABLE
		2191
1141	void encode (SV *self, SV *scalar)	2192	void encode_json (SV *scalar)
1142	PPCODE:	2193	PPCODE:
1143	XPUSHs (encode_json (scalar, *SvJSON (self)));	2194	{
		2195	JSON json;
		2196	json_init (&json);
		2197	json.flags |= F_UTF8;
		2198	PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
		2199	XPUSHs (scalar);
		2200	}
1144		2201
1145	void decode (SV *self, SV *jsonstr)	2202	void decode_json (SV *jsonstr)
1146	PPCODE:	2203	PPCODE:
1147	XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));
1148
1149	void decode_prefix (SV *self, SV *jsonstr)
1150	PPCODE:
1151	{	2204	{
1152	UV offset;	2205	JSON json;
1153	EXTEND (SP, 2);	2206	json_init (&json);
1154	PUSHs (decode_json (jsonstr, *SvJSON (self), &offset));	2207	json.flags |= F_UTF8;
1155	PUSHs (sv_2mortal (newSVuv (offset)));	2208	PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
		2209	XPUSHs (jsonstr);
1156	}	2210	}
1157		2211
1158	PROTOTYPES: ENABLE
1159
1160	void to_json (SV *scalar)
1161	ALIAS:
1162	objToJson = 0
1163	PPCODE:
1164	XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8));
1165
1166	void from_json (SV *jsonstr)
1167	ALIAS:
1168	jsonToObj = 0
1169	PPCODE:
1170	XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8, 0));
1171

Diff Legend

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