ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.18 by root, Sun Mar 25 21:19:13 2007 UTC vs.
Revision 1.124 by root, Mon Nov 24 18:42:51 2014 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 0x00000001UL	36	#define F_ASCII 0x00000001UL
		37	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002UL	38	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004UL	39	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008UL	40	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010UL	41	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020UL	42	#define F_SPACE_AFTER 0x00000040UL
15	#define F_ALLOW_NONREF 0x00000080UL	43	#define F_ALLOW_NONREF 0x00000100UL
16	#define F_SHRINK 0x00000100UL	44	#define F_SHRINK 0x00000200UL
		45	#define F_ALLOW_BLESSED 0x00000400UL
		46	#define F_CONV_BLESSED 0x00000800UL
17	#define F_MAXDEPTH 0xf8000000UL	47	#define F_RELAXED 0x00001000UL
18	#define S_MAXDEPTH 27	48	#define F_ALLOW_UNKNOWN 0x00002000UL
19		49	#define F_ALLOW_TAGS 0x00004000UL
20	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))	50	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
21
22	// F_SELFCONVERT? <=> to_json/toJson
23	// F_BLESSED? <=> { $__class__$ => }
24		51
25	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	52	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
26	#define F_DEFAULT (13UL << S_MAXDEPTH)
27		53
28	#define INIT_SIZE 32 // initial scalar size to be allocated	54	#define INIT_SIZE 32 // initial scalar size to be allocated
29	#define INDENT_STEP 3 // spaces per indentation level	55	#define INDENT_STEP 3 // spaces per indentation level
30		56
31	#define UTF8_MAX_LEN 11 // for perls UTF-X: max. number of octets per character
32	#define SHORT_STRING_LEN 512 // special-case strings of up to this size	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)
33		60
34	#define SB do {	61	#define SB do {
35	#define SE } while (0)	62	#define SE } while (0)
36		63
		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
37	static HV *json_stash; // JSON::XS::	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	}
38		127
39	/////////////////////////////////////////////////////////////////////////////	128	/////////////////////////////////////////////////////////////////////////////
40	// utility functions	129	// utility functions
41		130
42	static UV *	131	INLINE SV *
43	SvJSON (SV *sv)	132	get_bool (const char *name)
44	{	133	{
45	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))	134	SV *sv = get_sv (name, 1);
46	croak ("object is not of type JSON::XS");
47		135
48	return &SvUVX (SvRV (sv));	136	SvREADONLY_on (sv);
49	}	137	SvREADONLY_on (SvRV (sv));
50		138
51	static void	139	return sv;
		140	}
		141
		142	INLINE void
52	shrink (SV *sv)	143	shrink (SV *sv)
53	{	144	{
54	sv_utf8_downgrade (sv, 1);	145	sv_utf8_downgrade (sv, 1);
		146
55	if (SvLEN (sv) > SvCUR (sv) + 1)	147	if (SvLEN (sv) > SvCUR (sv) + 1)
56	{	148	{
57	#ifdef SvPV_shrink_to_cur	149	#ifdef SvPV_shrink_to_cur
58	SvPV_shrink_to_cur (sv);	150	SvPV_shrink_to_cur (sv);
59	#elif defined (SvPV_renew)	151	#elif defined (SvPV_renew)
…		…
65	// decode an utf-8 character and return it, or (UV)-1 in	157	// decode an utf-8 character and return it, or (UV)-1 in
66	// case of an error.	158	// case of an error.
67	// we special-case "safe" characters from U+80 .. U+7FF,	159	// we special-case "safe" characters from U+80 .. U+7FF,
68	// but use the very good perl function to parse anything else.	160	// but use the very good perl function to parse anything else.
69	// note that we never call this function for a ascii codepoints	161	// note that we never call this function for a ascii codepoints
70	static UV	162	INLINE UV
71	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)	163	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
72	{	164	{
73	if (s[0] > 0xdf || s[0] < 0xc2)	165	if (expect_true (len >= 2
74	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	166	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
75	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	167	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
76	{	168	{
77	*clen = 2;	169	*clen = 2;
78	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);	170	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
79	}	171	}
80	else	172	else
81	return (UV)-1;	173	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
82	}	174	}
83		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	}
84	/////////////////////////////////////////////////////////////////////////////	304	/////////////////////////////////////////////////////////////////////////////
85	// encoder	305	// encoder
86		306
87	// structure used for encoding JSON	307	// structure used for encoding JSON
88	typedef struct	308	typedef struct
89	{	309	{
90	char *cur; // SvPVX (sv) + current output position	310	char *cur; // SvPVX (sv) + current output position
91	char *end; // SvEND (sv)	311	char *end; // SvEND (sv)
92	SV *sv; // result scalar	312	SV *sv; // result scalar
93	U32 flags; // F_*	313	JSON json;
94	U32 indent; // indentation level	314	U32 indent; // indentation level
95	U32 maxdepth; // max. indentation/recursion level	315	UV limit; // escape character values >= this value when encoding
96	} enc_t;	316	} enc_t;
97		317
98	static void	318	INLINE void
99	need (enc_t *enc, STRLEN len)	319	need (enc_t *enc, STRLEN len)
100	{	320	{
101	if (enc->cur + len >= enc->end)	321	if (expect_false (enc->cur + len >= enc->end))
102	{	322	{
103	STRLEN cur = enc->cur - SvPVX (enc->sv);	323	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
104	SvGROW (enc->sv, cur + len + 1);	324	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
105	enc->cur = SvPVX (enc->sv) + cur;	325	enc->cur = SvPVX (enc->sv) + cur;
106	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	326	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
107	}	327	}
108	}	328	}
109		329
110	static void	330	INLINE void
111	encode_ch (enc_t *enc, char ch)	331	encode_ch (enc_t *enc, char ch)
112	{	332	{
113	need (enc, 1);	333	need (enc, 1);
114	*enc->cur++ = ch;	334	*enc->cur++ = ch;
115	}	335	}
…		…
123		343
124	while (str < end)	344	while (str < end)
125	{	345	{
126	unsigned char ch = *(unsigned char *)str;	346	unsigned char ch = *(unsigned char *)str;
127		347
128	if (ch >= 0x20 && ch < 0x80) // most common case	348	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
129	{	349	{
130	if (ch == '"') // but with slow exceptions	350	if (expect_false (ch == '"')) // but with slow exceptions
131	{	351	{
132	need (enc, len += 1);	352	need (enc, len += 1);
133	*enc->cur++ = '\\';	353	*enc->cur++ = '\\';
134	*enc->cur++ = '"';	354	*enc->cur++ = '"';
135	}	355	}
136	else if (ch == '\\')	356	else if (expect_false (ch == '\\'))
137	{	357	{
138	need (enc, len += 1);	358	need (enc, len += 1);
139	*enc->cur++ = '\\';	359	*enc->cur++ = '\\';
140	*enc->cur++ = '\\';	360	*enc->cur++ = '\\';
141	}	361	}
…		…
159	STRLEN clen;	379	STRLEN clen;
160	UV uch;	380	UV uch;
161		381
162	if (is_utf8)	382	if (is_utf8)
163	{	383	{
164	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
165	uch = decode_utf8 (str, end - str, &clen);	384	uch = decode_utf8 (str, end - str, &clen);
166	if (clen == (STRLEN)-1)	385	if (clen == (STRLEN)-1)
167	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	386	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
168	}	387	}
169	else	388	else
170	{	389	{
171	uch = ch;	390	uch = ch;
172	clen = 1;	391	clen = 1;
173	}	392	}
174		393
175	if (uch > 0x10FFFFUL)	394	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
176	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
177
178	if (uch < 0x80 || enc->flags & F_ASCII)
179	{	395	{
180	if (uch > 0xFFFFUL)	396	if (uch >= 0x10000UL)
181	{	397	{
		398	if (uch >= 0x110000UL)
		399	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		400
182	need (enc, len += 11);	401	need (enc, len += 11);
183	sprintf (enc->cur, "\\u%04x\\u%04x",	402	sprintf (enc->cur, "\\u%04x\\u%04x",
184	(int)((uch - 0x10000) / 0x400 + 0xD800),	403	(int)((uch - 0x10000) / 0x400 + 0xD800),
185	(int)((uch - 0x10000) % 0x400 + 0xDC00));	404	(int)((uch - 0x10000) % 0x400 + 0xDC00));
186	enc->cur += 12;	405	enc->cur += 12;
187	}	406	}
188	else	407	else
189	{	408	{
190	static char hexdigit [16] = "0123456789abcdef";
191	need (enc, len += 5);	409	need (enc, len += 5);
192	*enc->cur++ = '\\';	410	*enc->cur++ = '\\';
193	*enc->cur++ = 'u';	411	*enc->cur++ = 'u';
194	*enc->cur++ = hexdigit [ uch >> 12 ];	412	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
195	*enc->cur++ = hexdigit [(uch >> 8) & 15];	413	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
196	*enc->cur++ = hexdigit [(uch >> 4) & 15];	414	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
197	*enc->cur++ = hexdigit [(uch >> 0) & 15];	415	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
198	}	416	}
199		417
		418	str += clen;
		419	}
		420	else if (enc->json.flags & F_LATIN1)
		421	{
		422	*enc->cur++ = uch;
200	str += clen;	423	str += clen;
201	}	424	}
202	else if (is_utf8)	425	else if (is_utf8)
203	{	426	{
204	need (enc, len += clen);	427	need (enc, len += clen);
…		…
208	}	431	}
209	while (--clen);	432	while (--clen);
210	}	433	}
211	else	434	else
212	{	435	{
213	need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed	436	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
214	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	437	enc->cur = encode_utf8 (enc->cur, uch);
215	++str;	438	++str;
216	}	439	}
217	}	440	}
218	}	441	}
219	}	442	}
220		443
221	--len;	444	--len;
222	}	445	}
223	}	446	}
224		447
225	static void	448	INLINE void
226	encode_indent (enc_t *enc)	449	encode_indent (enc_t *enc)
227	{	450	{
228	if (enc->flags & F_INDENT)	451	if (enc->json.flags & F_INDENT)
229	{	452	{
230	int spaces = enc->indent * INDENT_STEP;	453	int spaces = enc->indent * INDENT_STEP;
231		454
232	need (enc, spaces);	455	need (enc, spaces);
233	memset (enc->cur, ' ', spaces);	456	memset (enc->cur, ' ', spaces);
234	enc->cur += spaces;	457	enc->cur += spaces;
235	}	458	}
236	}	459	}
237		460
238	static void	461	INLINE void
239	encode_space (enc_t *enc)	462	encode_space (enc_t *enc)
240	{	463	{
241	need (enc, 1);	464	need (enc, 1);
242	encode_ch (enc, ' ');	465	encode_ch (enc, ' ');
243	}	466	}
244		467
245	static void	468	INLINE void
246	encode_nl (enc_t *enc)	469	encode_nl (enc_t *enc)
247	{	470	{
248	if (enc->flags & F_INDENT)	471	if (enc->json.flags & F_INDENT)
249	{	472	{
250	need (enc, 1);	473	need (enc, 1);
251	encode_ch (enc, '\n');	474	encode_ch (enc, '\n');
252	}	475	}
253	}	476	}
254		477
255	static void	478	INLINE void
256	encode_comma (enc_t *enc)	479	encode_comma (enc_t *enc)
257	{	480	{
258	encode_ch (enc, ',');	481	encode_ch (enc, ',');
259		482
260	if (enc->flags & F_INDENT)	483	if (enc->json.flags & F_INDENT)
261	encode_nl (enc);	484	encode_nl (enc);
262	else if (enc->flags & F_SPACE_AFTER)	485	else if (enc->json.flags & F_SPACE_AFTER)
263	encode_space (enc);	486	encode_space (enc);
264	}	487	}
265		488
266	static void encode_sv (enc_t *enc, SV *sv);	489	static void encode_sv (enc_t *enc, SV *sv);
267		490
268	static void	491	static void
269	encode_av (enc_t *enc, AV *av)	492	encode_av (enc_t *enc, AV *av)
270	{	493	{
271	int i, len = av_len (av);	494	int i, len = av_len (av);
272		495
273	encode_ch (enc, '['); encode_nl (enc);	496	if (enc->indent >= enc->json.max_depth)
274	++enc->indent;	497	croak (ERR_NESTING_EXCEEDED);
275		498
		499	encode_ch (enc, '[');
		500
		501	if (len >= 0)
		502	{
		503	encode_nl (enc); ++enc->indent;
		504
276	for (i = 0; i <= len; ++i)	505	for (i = 0; i <= len; ++i)
277	{	506	{
		507	SV **svp = av_fetch (av, i, 0);
		508
278	encode_indent (enc);	509	encode_indent (enc);
279	encode_sv (enc, *av_fetch (av, i, 0));
280		510
		511	if (svp)
		512	encode_sv (enc, *svp);
		513	else
		514	encode_str (enc, "null", 4, 0);
		515
281	if (i < len)	516	if (i < len)
282	encode_comma (enc);	517	encode_comma (enc);
283	}	518	}
284		519
		520	encode_nl (enc); --enc->indent; encode_indent (enc);
		521	}
		522
285	encode_nl (enc);	523	encode_ch (enc, ']');
286
287	--enc->indent;
288	encode_indent (enc); encode_ch (enc, ']');
289	}	524	}
290		525
291	static void	526	static void
292	encode_he (enc_t *enc, HE *he)	527	encode_hk (enc_t *enc, HE *he)
293	{	528	{
294	encode_ch (enc, '"');	529	encode_ch (enc, '"');
295		530
296	if (HeKLEN (he) == HEf_SVKEY)	531	if (HeKLEN (he) == HEf_SVKEY)
297	{	532	{
298	SV *sv = HeSVKEY (he);	533	SV *sv = HeSVKEY (he);
299	STRLEN len;	534	STRLEN len;
300	char *str;	535	char *str;
301		536
302	SvGETMAGIC (sv);	537	SvGETMAGIC (sv);
303	str = SvPV (sv, len);	538	str = SvPV (sv, len);
304		539
305	encode_str (enc, str, len, SvUTF8 (sv));	540	encode_str (enc, str, len, SvUTF8 (sv));
306	}	541	}
307	else	542	else
308	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	543	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
309		544
310	encode_ch (enc, '"');	545	encode_ch (enc, '"');
311		546
312	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	547	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
313	encode_ch (enc, ':');	548	encode_ch (enc, ':');
314	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	549	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
315	encode_sv (enc, HeVAL (he));
316	}	550	}
317		551
318	// compare hash entries, used when all keys are bytestrings	552	// compare hash entries, used when all keys are bytestrings
319	static int	553	static int
320	he_cmp_fast (const void *a_, const void *b_)	554	he_cmp_fast (const void *a_, const void *b_)
…		…
325	HE *b = *(HE **)b_;	559	HE *b = *(HE **)b_;
326		560
327	STRLEN la = HeKLEN (a);	561	STRLEN la = HeKLEN (a);
328	STRLEN lb = HeKLEN (b);	562	STRLEN lb = HeKLEN (b);
329		563
330	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	564	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
331	cmp = la - lb;	565	cmp = lb - la;
332		566
333	return cmp;	567	return cmp;
334	}	568	}
335		569
336	// 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
337	static int	571	static int
338	he_cmp_slow (const void *a, const void *b)	572	he_cmp_slow (const void *a, const void *b)
339	{	573	{
340	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	574	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
341	}	575	}
342		576
343	static void	577	static void
344	encode_hv (enc_t *enc, HV *hv)	578	encode_hv (enc_t *enc, HV *hv)
345	{	579	{
346	int count, i;	580	HE *he;
347		581
348	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	582	if (enc->indent >= enc->json.max_depth)
		583	croak (ERR_NESTING_EXCEEDED);
349		584
350	if ((count = hv_iterinit (hv)))	585	encode_ch (enc, '{');
351	{	586
352	// for canonical output we have to sort by keys first	587	// for canonical output we have to sort by keys first
353	// actually, this is mostly due to the stupid so-called	588	// actually, this is mostly due to the stupid so-called
354	// security workaround added somewhere in 5.8.x.	589	// security workaround added somewhere in 5.8.x
355	// that randomises hash orderings	590	// that randomises hash orderings
356	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))
357	{	596	{
358	HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode	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	{
359	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	}
360		621
361	i = 0;	622	i = 0;
362	while ((he = hv_iternext (hv)))	623	while ((he = hv_iternext (hv)))
363	{	624	{
364	hes [i++] = he;	625	hes [i++] = he;
…		…
386		647
387	FREETMPS;	648	FREETMPS;
388	LEAVE;	649	LEAVE;
389	}	650	}
390		651
391	for (i = 0; i < count; ++i)	652	encode_nl (enc); ++enc->indent;
		653
		654	while (count--)
392	{	655	{
393	encode_indent (enc);	656	encode_indent (enc);
		657	he = hes [count];
394	encode_he (enc, hes [i]);	658	encode_hk (enc, he);
		659	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
395		660
396	if (i < count - 1)	661	if (count)
397	encode_comma (enc);	662	encode_comma (enc);
398	}	663	}
399		664
400	encode_nl (enc);	665	encode_nl (enc); --enc->indent; encode_indent (enc);
401	}	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; PUSHMARK (SP);
		724	EXTEND (SP, 2);
		725	// we re-bless the reference to get overload and other niceties right
		726	PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		727	PUSHs (sv_json);
		728
		729	PUTBACK;
		730	count = call_sv ((SV *)GvCV (method), G_ARRAY);
		731	SPAGAIN;
		732
		733	// catch this surprisingly common error
		734	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		735	croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		736
		737	encode_ch (enc, '(');
		738	encode_ch (enc, '"');
		739	encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
		740	encode_ch (enc, '"');
		741	encode_ch (enc, ')');
		742	encode_ch (enc, '[');
		743
		744	while (count)
		745	{
		746	encode_sv (enc, SP[1 - count--]);
		747
		748	if (count)
		749	encode_ch (enc, ',');
		750	}
		751
		752	encode_ch (enc, ']');
		753
		754	PUTBACK;
		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; PUSHMARK (SP);
		763	// we re-bless the reference to get overload and other niceties right
		764	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		765
		766	// calling with G_SCALAR ensures that we always get a 1 return value
		767	PUTBACK;
		768	call_sv ((SV *)GvCV (method), G_SCALAR);
		769	SPAGAIN;
		770
		771	// catch this surprisingly common error
		772	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		773	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		774
		775	sv = POPs;
		776	PUTBACK;
		777
		778	encode_sv (enc, sv);
		779
		780	FREETMPS; LEAVE;
		781	}
		782	else if (enc->json.flags & F_ALLOW_BLESSED)
		783	encode_str (enc, "null", 4, 0);
402	else	784	else
403	{	785	croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
404	SV *sv;	786	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
405	HE *he = hv_iternext (hv);	787	}
406		788	else if (svt == SVt_PVHV)
407	for (;;)	789	encode_hv (enc, (HV *)sv);
408	{	790	else if (svt == SVt_PVAV)
409	encode_indent (enc);	791	encode_av (enc, (AV *)sv);
410	encode_he (enc, he);	792	else if (svt < SVt_PVAV)
411
412	if (!(he = hv_iternext (hv)))
413	break;
414
415	encode_comma (enc);
416	}
417
418	encode_nl (enc);
419	}
420	}	793	{
		794	STRLEN len = 0;
		795	char *pv = svt ? SvPV (sv, len) : 0;
421		796
422	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	797	if (len == 1 && *pv == '1')
		798	encode_str (enc, "true", 4, 0);
		799	else if (len == 1 && *pv == '0')
		800	encode_str (enc, "false", 5, 0);
		801	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		802	encode_str (enc, "null", 4, 0);
		803	else
		804	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		805	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		806	}
		807	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		808	encode_str (enc, "null", 4, 0);
		809	else
		810	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		811	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
423	}	812	}
424		813
425	static void	814	static void
426	encode_sv (enc_t *enc, SV *sv)	815	encode_sv (enc_t *enc, SV *sv)
427	{	816	{
…		…
435	encode_str (enc, str, len, SvUTF8 (sv));	824	encode_str (enc, str, len, SvUTF8 (sv));
436	encode_ch (enc, '"');	825	encode_ch (enc, '"');
437	}	826	}
438	else if (SvNOKp (sv))	827	else if (SvNOKp (sv))
439	{	828	{
		829	// trust that perl will do the right thing w.r.t. JSON syntax.
440	need (enc, NV_DIG + 32);	830	need (enc, NV_DIG + 32);
441	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	831	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
442	enc->cur += strlen (enc->cur);	832	enc->cur += strlen (enc->cur);
443	}	833	}
444	else if (SvIOKp (sv))	834	else if (SvIOKp (sv))
445	{	835	{
		836	// we assume we can always read an IV as a UV and vice versa
		837	// we assume two's complement
		838	// we assume no aliasing issues in the union
		839	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		840	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		841	{
		842	// optimise the "small number case"
		843	// code will likely be branchless and use only a single multiplication
		844	// works for numbers up to 59074
		845	I32 i = SvIVX (sv);
		846	U32 u;
		847	char digit, nz = 0;
		848
446	need (enc, 64);	849	need (enc, 6);
		850
		851	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		852	u = i < 0 ? -i : i;
		853
		854	// convert to 4.28 fixed-point representation
		855	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		856
		857	// now output digit by digit, each time masking out the integer part
		858	// and multiplying by 5 while moving the decimal point one to the right,
		859	// resulting in a net multiplication by 10.
		860	// we always write the digit to memory but conditionally increment
		861	// the pointer, to enable the use of conditional move instructions.
		862	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		863	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		864	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		865	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		866	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		867	}
		868	else
		869	{
		870	// large integer, use the (rather slow) snprintf way.
		871	need (enc, IVUV_MAXCHARS);
447	enc->cur +=	872	enc->cur +=
448	SvIsUV(sv)	873	SvIsUV(sv)
449	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	874	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
450	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	875	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		876	}
451	}	877	}
452	else if (SvROK (sv))	878	else if (SvROK (sv))
453	{	879	encode_rv (enc, SvRV (sv));
454	SV *rv = SvRV (sv);	880	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
455
456	if (enc->indent >= enc->maxdepth)
457	croak ("data structure too deep (hit recursion limit)");
458
459	switch (SvTYPE (rv))
460	{
461	case SVt_PVAV: encode_av (enc, (AV *)rv); break;
462	case SVt_PVHV: encode_hv (enc, (HV *)rv); break;
463
464	default:
465	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
466	SvPV_nolen (sv));
467	}
468	}
469	else if (!SvOK (sv))
470	encode_str (enc, "null", 4, 0);	881	encode_str (enc, "null", 4, 0);
471	else	882	else
472	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	883	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
473	SvPV_nolen (sv), SvFLAGS (sv));	884	SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
474	}	885	}
475		886
476	static SV *	887	static SV *
477	encode_json (SV *scalar, U32 flags)	888	encode_json (SV *scalar, JSON *json)
478	{	889	{
		890	enc_t enc;
		891
479	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	892	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
480	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	893	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
481		894
482	enc_t enc;	895	enc.json = *json;
483	enc.flags = flags;
484	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	896	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
485	enc.cur = SvPVX (enc.sv);	897	enc.cur = SvPVX (enc.sv);
486	enc.end = SvEND (enc.sv);	898	enc.end = SvEND (enc.sv);
487	enc.indent = 0;	899	enc.indent = 0;
488	enc.maxdepth = DEC_DEPTH (flags);	900	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		901	: enc.json.flags & F_LATIN1 ? 0x000100UL
		902	: 0x110000UL;
489		903
490	SvPOK_only (enc.sv);	904	SvPOK_only (enc.sv);
491	encode_sv (&enc, scalar);	905	encode_sv (&enc, scalar);
		906	encode_nl (&enc);
492		907
		908	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		909	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		910
493	if (!(flags & (F_ASCII | F_UTF8)))	911	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
494	SvUTF8_on (enc.sv);	912	SvUTF8_on (enc.sv);
495		913
496	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
497
498	if (enc.flags & F_SHRINK)	914	if (enc.json.flags & F_SHRINK)
499	shrink (enc.sv);	915	shrink (enc.sv);
500		916
501	return enc.sv;	917	return enc.sv;
502	}	918	}
503		919
…		…
508	typedef struct	924	typedef struct
509	{	925	{
510	char *cur; // current parser pointer	926	char *cur; // current parser pointer
511	char *end; // end of input string	927	char *end; // end of input string
512	const char *err; // parse error, if != 0	928	const char *err; // parse error, if != 0
513	U32 flags; // F_*	929	JSON json;
514	U32 depth; // recursion depth	930	U32 depth; // recursion depth
515	U32 maxdepth; // recursion depth limit	931	U32 maxdepth; // recursion depth limit
516	} dec_t;	932	} dec_t;
517		933
518	static void	934	INLINE void
		935	decode_comment (dec_t *dec)
		936	{
		937	// only '#'-style comments allowed a.t.m.
		938
		939	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		940	++dec->cur;
		941	}
		942
		943	INLINE void
519	decode_ws (dec_t *dec)	944	decode_ws (dec_t *dec)
520	{	945	{
521	for (;;)	946	for (;;)
522	{	947	{
523	char ch = *dec->cur;	948	char ch = *dec->cur;
524		949
525	if (ch > 0x20	950	if (ch > 0x20)
		951	{
		952	if (expect_false (ch == '#'))
		953	{
		954	if (dec->json.flags & F_RELAXED)
		955	decode_comment (dec);
		956	else
		957	break;
		958	}
		959	else
		960	break;
		961	}
526	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	962	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
527	break;	963	break; // parse error, but let higher level handle it, gives better error messages
528		964
529	++dec->cur;	965	++dec->cur;
530	}	966	}
531	}	967	}
532		968
…		…
536	if (*dec->cur != ch) \	972	if (*dec->cur != ch) \
537	ERR (# ch " expected"); \	973	ERR (# ch " expected"); \
538	++dec->cur; \	974	++dec->cur; \
539	SE	975	SE
540		976
541	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	977	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
542	#define DEC_DEC_DEPTH --dec->depth	978	#define DEC_DEC_DEPTH --dec->depth
543		979
544	static SV *decode_sv (dec_t *dec);	980	static SV *decode_sv (dec_t *dec);
545		981
546	static signed char decode_hexdigit[256];	982	static signed char decode_hexdigit[256];
…		…
549	decode_4hex (dec_t *dec)	985	decode_4hex (dec_t *dec)
550	{	986	{
551	signed char d1, d2, d3, d4;	987	signed char d1, d2, d3, d4;
552	unsigned char *cur = (unsigned char *)dec->cur;	988	unsigned char *cur = (unsigned char *)dec->cur;
553		989
554	d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected");	990	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
555	d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected");	991	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
556	d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected");	992	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
557	d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected");	993	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
558		994
559	dec->cur += 4;	995	dec->cur += 4;
560		996
561	return ((UV)d1) << 12	997	return ((UV)d1) << 12
562	| ((UV)d2) << 8	998	| ((UV)d2) << 8
…		…
570	static SV *	1006	static SV *
571	decode_str (dec_t *dec)	1007	decode_str (dec_t *dec)
572	{	1008	{
573	SV *sv = 0;	1009	SV *sv = 0;
574	int utf8 = 0;	1010	int utf8 = 0;
		1011	char *dec_cur = dec->cur;
575		1012
576	do	1013	do
577	{	1014	{
578	char buf [SHORT_STRING_LEN + UTF8_MAX_LEN];	1015	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
579	char *cur = buf;	1016	char *cur = buf;
580		1017
581	do	1018	do
582	{	1019	{
583	unsigned char ch = *(unsigned char *)dec->cur++;	1020	unsigned char ch = *(unsigned char *)dec_cur++;
584		1021
585	if (ch == '"')	1022	if (expect_false (ch == '"'))
586	{	1023	{
587	--dec->cur;	1024	--dec_cur;
588	break;	1025	break;
589	}	1026	}
590	else if (ch == '\\')	1027	else if (expect_false (ch == '\\'))
591	{	1028	{
592	switch (*dec->cur)	1029	switch (*dec_cur)
593	{	1030	{
594	case '\\':	1031	case '\\':
595	case '/':	1032	case '/':
596	case '"': *cur++ = *dec->cur++; break;	1033	case '"': *cur++ = *dec_cur++; break;
597		1034
598	case 'b': ++dec->cur; *cur++ = '\010'; break;	1035	case 'b': ++dec_cur; *cur++ = '\010'; break;
599	case 't': ++dec->cur; *cur++ = '\011'; break;	1036	case 't': ++dec_cur; *cur++ = '\011'; break;
600	case 'n': ++dec->cur; *cur++ = '\012'; break;	1037	case 'n': ++dec_cur; *cur++ = '\012'; break;
601	case 'f': ++dec->cur; *cur++ = '\014'; break;	1038	case 'f': ++dec_cur; *cur++ = '\014'; break;
602	case 'r': ++dec->cur; *cur++ = '\015'; break;	1039	case 'r': ++dec_cur; *cur++ = '\015'; break;
603		1040
604	case 'u':	1041	case 'u':
605	{	1042	{
606	UV lo, hi;	1043	UV lo, hi;
607	++dec->cur;	1044	++dec_cur;
608		1045
		1046	dec->cur = dec_cur;
609	hi = decode_4hex (dec);	1047	hi = decode_4hex (dec);
		1048	dec_cur = dec->cur;
610	if (hi == (UV)-1)	1049	if (hi == (UV)-1)
611	goto fail;	1050	goto fail;
612		1051
613	// possibly a surrogate pair	1052	// possibly a surrogate pair
614	if (hi >= 0xd800)	1053	if (hi >= 0xd800)
615	if (hi < 0xdc00)	1054	if (hi < 0xdc00)
616	{	1055	{
617	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	1056	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
618	ERR ("missing low surrogate character in surrogate pair");	1057	ERR ("missing low surrogate character in surrogate pair");
619		1058
620	dec->cur += 2;	1059	dec_cur += 2;
621		1060
		1061	dec->cur = dec_cur;
622	lo = decode_4hex (dec);	1062	lo = decode_4hex (dec);
		1063	dec_cur = dec->cur;
623	if (lo == (UV)-1)	1064	if (lo == (UV)-1)
624	goto fail;	1065	goto fail;
625		1066
626	if (lo < 0xdc00 || lo >= 0xe000)	1067	if (lo < 0xdc00 || lo >= 0xe000)
627	ERR ("surrogate pair expected");	1068	ERR ("surrogate pair expected");
…		…
633		1074
634	if (hi >= 0x80)	1075	if (hi >= 0x80)
635	{	1076	{
636	utf8 = 1;	1077	utf8 = 1;
637		1078
638	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	1079	cur = encode_utf8 (cur, hi);
639	}	1080	}
640	else	1081	else
641	*cur++ = hi;	1082	*cur++ = hi;
642	}	1083	}
643	break;	1084	break;
644		1085
645	default:	1086	default:
646	--dec->cur;	1087	--dec_cur;
647	ERR ("illegal backslash escape sequence in string");	1088	ERR ("illegal backslash escape sequence in string");
648	}	1089	}
649	}	1090	}
650	else if (ch >= 0x20 && ch <= 0x7f)	1091	else if (expect_true (ch >= 0x20 && ch < 0x80))
651	*cur++ = ch;	1092	*cur++ = ch;
652	else if (ch >= 0x80)	1093	else if (ch >= 0x80)
653	{	1094	{
654	--dec->cur;
655
656	STRLEN clen;	1095	STRLEN clen;
		1096
		1097	--dec_cur;
		1098
657	UV uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	1099	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
658	if (clen == (STRLEN)-1)	1100	if (clen == (STRLEN)-1)
659	ERR ("malformed UTF-8 character in JSON string");	1101	ERR ("malformed UTF-8 character in JSON string");
660		1102
661	do	1103	do
662	{
663	*cur++ = *dec->cur++;	1104	*cur++ = *dec_cur++;
664	}
665	while (--clen);	1105	while (--clen);
666		1106
667	utf8 = 1;	1107	utf8 = 1;
668	}	1108	}
669	else if (!ch)	1109	else if (ch == '\t' && dec->json.flags & F_RELAXED)
670	ERR ("unexpected end of string while parsing json string");	1110	*cur++ = ch;
671	else	1111	else
		1112	{
		1113	--dec_cur;
		1114
		1115	if (!ch)
		1116	ERR ("unexpected end of string while parsing JSON string");
		1117	else
672	ERR ("invalid character encountered");	1118	ERR ("invalid character encountered while parsing JSON string");
673		1119	}
674	}	1120	}
675	while (cur < buf + SHORT_STRING_LEN);	1121	while (cur < buf + SHORT_STRING_LEN);
676		1122
		1123	{
677	STRLEN len = cur - buf;	1124	STRLEN len = cur - buf;
678		1125
679	if (sv)	1126	if (sv)
680	{	1127	{
681	SvGROW (sv, SvCUR (sv) + len + 1);	1128	STRLEN cur = SvCUR (sv);
		1129
		1130	if (SvLEN (sv) <= cur + len)
		1131	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1132
682	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	1133	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
683	SvCUR_set (sv, SvCUR (sv) + len);	1134	SvCUR_set (sv, SvCUR (sv) + len);
684	}	1135	}
685	else	1136	else
686	sv = newSVpvn (buf, len);	1137	sv = newSVpvn (buf, len);
687	}	1138	}
		1139	}
688	while (*dec->cur != '"');	1140	while (*dec_cur != '"');
689		1141
690	++dec->cur;	1142	++dec_cur;
691		1143
692	if (sv)	1144	if (sv)
693	{	1145	{
694	SvPOK_only (sv);	1146	SvPOK_only (sv);
695	*SvEND (sv) = 0;	1147	*SvEND (sv) = 0;
…		…
698	SvUTF8_on (sv);	1150	SvUTF8_on (sv);
699	}	1151	}
700	else	1152	else
701	sv = newSVpvn ("", 0);	1153	sv = newSVpvn ("", 0);
702		1154
		1155	dec->cur = dec_cur;
703	return sv;	1156	return sv;
704		1157
705	fail:	1158	fail:
		1159	dec->cur = dec_cur;
706	return 0;	1160	return 0;
707	}	1161	}
708		1162
709	static SV *	1163	static SV *
710	decode_num (dec_t *dec)	1164	decode_num (dec_t *dec)
…		…
768	is_nv = 1;	1222	is_nv = 1;
769	}	1223	}
770		1224
771	if (!is_nv)	1225	if (!is_nv)
772	{	1226	{
773	UV uv;	1227	int len = dec->cur - start;
774	int numtype = grok_number (start, dec->cur - start, &uv);	1228
775	if (numtype & IS_NUMBER_IN_UV)	1229	// special case the rather common 1..5-digit-int case
776	if (numtype & IS_NUMBER_NEG)	1230	if (*start == '-')
		1231	switch (len)
777	{	1232	{
778	if (uv < (UV)IV_MIN)	1233	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
779	return newSViv (-(IV)uv);	1234	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
		1235	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1236	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1237	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
780	}	1238	}
		1239	else
		1240	switch (len)
		1241	{
		1242	case 1: return newSViv ( start [0] - '0' * 1);
		1243	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1244	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1245	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1246	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1247	}
		1248
		1249	{
		1250	UV uv;
		1251	int numtype = grok_number (start, len, &uv);
		1252	if (numtype & IS_NUMBER_IN_UV)
		1253	if (numtype & IS_NUMBER_NEG)
		1254	{
		1255	if (uv < (UV)IV_MIN)
		1256	return newSViv (-(IV)uv);
		1257	}
781	else	1258	else
782	return newSVuv (uv);	1259	return newSVuv (uv);
783	}	1260	}
784		1261
		1262	len -= *start == '-' ? 1 : 0;
		1263
		1264	// does not fit into IV or UV, try NV
		1265	if (len <= NV_DIG)
		1266	// fits into NV without loss of precision
		1267	return newSVnv (json_atof (start));
		1268
		1269	// everything else fails, convert it to a string
		1270	return newSVpvn (start, dec->cur - start);
		1271	}
		1272
		1273	// loss of precision here
785	return newSVnv (Atof (start));	1274	return newSVnv (json_atof (start));
786		1275
787	fail:	1276	fail:
788	return 0;	1277	return 0;
789	}	1278	}
790		1279
…		…
814	if (*dec->cur == ']')	1303	if (*dec->cur == ']')
815	{	1304	{
816	++dec->cur;	1305	++dec->cur;
817	break;	1306	break;
818	}	1307	}
819		1308
820	if (*dec->cur != ',')	1309	if (*dec->cur != ',')
821	ERR (", or ] expected while parsing array");	1310	ERR (", or ] expected while parsing array");
822		1311
823	++dec->cur;	1312	++dec->cur;
		1313
		1314	decode_ws (dec);
		1315
		1316	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1317	{
		1318	++dec->cur;
		1319	break;
		1320	}
824	}	1321	}
825		1322
826	DEC_DEC_DEPTH;	1323	DEC_DEC_DEPTH;
827	return newRV_noinc ((SV *)av);	1324	return newRV_noinc ((SV *)av);
828		1325
…		…
833	}	1330	}
834		1331
835	static SV *	1332	static SV *
836	decode_hv (dec_t *dec)	1333	decode_hv (dec_t *dec)
837	{	1334	{
		1335	SV *sv;
838	HV *hv = newHV ();	1336	HV *hv = newHV ();
839		1337
840	DEC_INC_DEPTH;	1338	DEC_INC_DEPTH;
841	decode_ws (dec);	1339	decode_ws (dec);
842		1340
843	if (*dec->cur == '}')	1341	if (*dec->cur == '}')
844	++dec->cur;	1342	++dec->cur;
845	else	1343	else
846	for (;;)	1344	for (;;)
847	{	1345	{
		1346	EXPECT_CH ('"');
		1347
		1348	// heuristic: assume that
		1349	// a) decode_str + hv_store_ent are abysmally slow.
		1350	// b) most hash keys are short, simple ascii text.
		1351	// => try to "fast-match" such strings to avoid
		1352	// the overhead of decode_str + hv_store_ent.
		1353	{
848	SV *key, *value;	1354	SV *value;
		1355	char *p = dec->cur;
		1356	char *e = p + 24; // only try up to 24 bytes
849		1357
850	decode_ws (dec); EXPECT_CH ('"');	1358	for (;;)
851
852	key = decode_str (dec);
853	if (!key)
854	goto fail;
855
856	decode_ws (dec); EXPECT_CH (':');
857
858	value = decode_sv (dec);
859	if (!value)
860	{	1359	{
		1360	// the >= 0x80 is false on most architectures
		1361	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1362	{
		1363	// slow path, back up and use decode_str
		1364	SV *key = decode_str (dec);
		1365	if (!key)
		1366	goto fail;
		1367
		1368	decode_ws (dec); EXPECT_CH (':');
		1369
		1370	decode_ws (dec);
		1371	value = decode_sv (dec);
		1372	if (!value)
		1373	{
		1374	SvREFCNT_dec (key);
		1375	goto fail;
		1376	}
		1377
		1378	hv_store_ent (hv, key, value, 0);
861	SvREFCNT_dec (key);	1379	SvREFCNT_dec (key);
		1380
		1381	break;
		1382	}
		1383	else if (*p == '"')
		1384	{
		1385	// fast path, got a simple key
		1386	char *key = dec->cur;
		1387	int len = p - key;
		1388	dec->cur = p + 1;
		1389
		1390	decode_ws (dec); EXPECT_CH (':');
		1391
		1392	decode_ws (dec);
		1393	value = decode_sv (dec);
		1394	if (!value)
862	goto fail;	1395	goto fail;
		1396
		1397	hv_store (hv, key, len, value, 0);
		1398
		1399	break;
		1400	}
		1401
		1402	++p;
863	}	1403	}
864		1404	}
865	hv_store_ent (hv, key, value, 0);
866	SvREFCNT_dec (key);
867		1405
868	decode_ws (dec);	1406	decode_ws (dec);
869		1407
870	if (*dec->cur == '}')	1408	if (*dec->cur == '}')
871	{	1409	{
…		…
875		1413
876	if (*dec->cur != ',')	1414	if (*dec->cur != ',')
877	ERR (", or } expected while parsing object/hash");	1415	ERR (", or } expected while parsing object/hash");
878		1416
879	++dec->cur;	1417	++dec->cur;
		1418
		1419	decode_ws (dec);
		1420
		1421	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1422	{
		1423	++dec->cur;
		1424	break;
		1425	}
880	}	1426	}
881		1427
882	DEC_DEC_DEPTH;	1428	DEC_DEC_DEPTH;
883	return newRV_noinc ((SV *)hv);	1429	sv = newRV_noinc ((SV *)hv);
		1430
		1431	// check filter callbacks
		1432	if (dec->json.flags & F_HOOK)
		1433	{
		1434	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1435	{
		1436	HE *cb, *he;
		1437
		1438	hv_iterinit (hv);
		1439	he = hv_iternext (hv);
		1440	hv_iterinit (hv);
		1441
		1442	// the next line creates a mortal sv each time its called.
		1443	// might want to optimise this for common cases.
		1444	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1445
		1446	if (cb)
		1447	{
		1448	dSP;
		1449	int count;
		1450
		1451	ENTER; SAVETMPS; PUSHMARK (SP);
		1452	XPUSHs (HeVAL (he));
		1453	sv_2mortal (sv);
		1454
		1455	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1456
		1457	if (count == 1)
		1458	{
		1459	sv = newSVsv (POPs);
		1460	FREETMPS; LEAVE;
		1461	return sv;
		1462	}
		1463
		1464	SvREFCNT_inc (sv);
		1465	FREETMPS; LEAVE;
		1466	}
		1467	}
		1468
		1469	if (dec->json.cb_object)
		1470	{
		1471	dSP;
		1472	int count;
		1473
		1474	ENTER; SAVETMPS; PUSHMARK (SP);
		1475	XPUSHs (sv_2mortal (sv));
		1476
		1477	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1478
		1479	if (count == 1)
		1480	{
		1481	sv = newSVsv (POPs);
		1482	FREETMPS; LEAVE;
		1483	return sv;
		1484	}
		1485
		1486	SvREFCNT_inc (sv);
		1487	FREETMPS; LEAVE;
		1488	}
		1489	}
		1490
		1491	return sv;
884		1492
885	fail:	1493	fail:
886	SvREFCNT_dec (hv);	1494	SvREFCNT_dec (hv);
887	DEC_DEC_DEPTH;	1495	DEC_DEC_DEPTH;
888	return 0;	1496	return 0;
889	}	1497	}
890		1498
891	static SV *	1499	static SV *
		1500	decode_tag (dec_t *dec)
		1501	{
		1502	SV *tag = 0;
		1503	SV *val = 0;
		1504
		1505	if (!(dec->json.flags & F_ALLOW_TAGS))
		1506	ERR ("malformed JSON string, neither array, object, number, string or atom");
		1507
		1508	++dec->cur;
		1509
		1510	decode_ws (dec);
		1511
		1512	tag = decode_sv (dec);
		1513	if (!tag)
		1514	goto fail;
		1515
		1516	if (!SvPOK (tag))
		1517	ERR ("malformed JSON string, (tag) must be a string");
		1518
		1519	decode_ws (dec);
		1520
		1521	if (*dec->cur != ')')
		1522	ERR (") expected after tag");
		1523
		1524	++dec->cur;
		1525
		1526	decode_ws (dec);
		1527
		1528	val = decode_sv (dec);
		1529	if (!val)
		1530	goto fail;
		1531
		1532	if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
		1533	ERR ("malformed JSON string, tag value must be an array");
		1534
		1535	{
		1536	AV *av = (AV *)SvRV (val);
		1537	int i, len = av_len (av) + 1;
		1538	HV *stash = gv_stashsv (tag, 0);
		1539	SV *sv;
		1540
		1541	if (!stash)
		1542	ERR ("cannot decode perl-object (package does not exist)");
		1543
		1544	GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
		1545
		1546	if (!method)
		1547	ERR ("cannot decode perl-object (package does not have a THAW method)");
		1548
		1549	dSP;
		1550
		1551	ENTER; SAVETMPS; PUSHMARK (SP);
		1552	EXTEND (SP, len + 2);
		1553	// we re-bless the reference to get overload and other niceties right
		1554	PUSHs (tag);
		1555	PUSHs (sv_json);
		1556
		1557	for (i = 0; i < len; ++i)
		1558	PUSHs (*av_fetch (av, i, 1));
		1559
		1560	PUTBACK;
		1561	call_sv ((SV *)GvCV (method), G_SCALAR);
		1562	SPAGAIN;
		1563
		1564	SvREFCNT_dec (tag);
		1565	SvREFCNT_dec (val);
		1566	sv = SvREFCNT_inc (POPs);
		1567
		1568	PUTBACK;
		1569
		1570	FREETMPS; LEAVE;
		1571
		1572	return sv;
		1573	}
		1574
		1575	fail:
		1576	SvREFCNT_dec (tag);
		1577	SvREFCNT_dec (val);
		1578	return 0;
		1579	}
		1580
		1581	static SV *
892	decode_sv (dec_t *dec)	1582	decode_sv (dec_t *dec)
893	{	1583	{
894	decode_ws (dec);	1584	// the beauty of JSON: you need exactly one character lookahead
		1585	// to parse everything.
895	switch (*dec->cur)	1586	switch (*dec->cur)
896	{	1587	{
897	case '"': ++dec->cur; return decode_str (dec);	1588	case '"': ++dec->cur; return decode_str (dec);
898	case '[': ++dec->cur; return decode_av (dec);	1589	case '[': ++dec->cur; return decode_av (dec);
899	case '{': ++dec->cur; return decode_hv (dec);	1590	case '{': ++dec->cur; return decode_hv (dec);
		1591	case '(': return decode_tag (dec);
900		1592
901	case '-':	1593	case '-':
902	case '0': case '1': case '2': case '3': case '4':	1594	case '0': case '1': case '2': case '3': case '4':
903	case '5': case '6': case '7': case '8': case '9':	1595	case '5': case '6': case '7': case '8': case '9':
904	return decode_num (dec);	1596	return decode_num (dec);
905		1597
906	case 't':	1598	case 't':
907	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1599	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
908	{	1600	{
909	dec->cur += 4;	1601	dec->cur += 4;
		1602	#if JSON_SLOW
		1603	types_true = get_bool ("Types::Serialiser::true");
		1604	#endif
910	return newSViv (1);	1605	return newSVsv (types_true);
911	}	1606	}
912	else	1607	else
913	ERR ("'true' expected");	1608	ERR ("'true' expected");
914		1609
915	break;	1610	break;
916		1611
917	case 'f':	1612	case 'f':
918	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1613	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
919	{	1614	{
920	dec->cur += 5;	1615	dec->cur += 5;
		1616	#if JSON_SLOW
		1617	types_false = get_bool ("Types::Serialiser::false");
		1618	#endif
921	return newSViv (0);	1619	return newSVsv (types_false);
922	}	1620	}
923	else	1621	else
924	ERR ("'false' expected");	1622	ERR ("'false' expected");
925		1623
926	break;	1624	break;
…		…
935	ERR ("'null' expected");	1633	ERR ("'null' expected");
936		1634
937	break;	1635	break;
938		1636
939	default:	1637	default:
940	ERR ("malformed json string, neither array, object, number, string or atom");	1638	ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
941	break;	1639	break;
942	}	1640	}
943		1641
944	fail:	1642	fail:
945	return 0;	1643	return 0;
946	}	1644	}
947		1645
948	static SV *	1646	static SV *
949	decode_json (SV *string, U32 flags)	1647	decode_json (SV *string, JSON *json, char **offset_return)
950	{	1648	{
		1649	dec_t dec;
951	SV *sv;	1650	SV *sv;
952		1651
953	if (flags & F_UTF8)	1652	/* work around bugs in 5.10 where manipulating magic values
		1653	* makes perl ignore the magic in subsequent accesses.
		1654	* also make a copy of non-PV values, to get them into a clean
		1655	* state (SvPV should do that, but it's buggy, see below).
		1656	*/
		1657	/*SvGETMAGIC (string);*/
		1658	if (SvMAGICAL (string) || !SvPOK (string))
		1659	string = sv_2mortal (newSVsv (string));
		1660
		1661	SvUPGRADE (string, SVt_PV);
		1662
		1663	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1664	* assertion with -DDEBUGGING, although SvCUR is documented to
		1665	* return the xpv_cur field which certainly exists after upgrading.
		1666	* according to nicholas clark, calling SvPOK fixes this.
		1667	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1668	* and hope for the best.
		1669	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1670	* assertion business is seriously broken, try yet another workaround
		1671	* for the broken -DDEBUGGING.
		1672	*/
		1673	{
		1674	#ifdef DEBUGGING
		1675	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1676	#else
		1677	STRLEN offset = SvCUR (string);
		1678	#endif
		1679
		1680	if (offset > json->max_size && json->max_size)
		1681	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1682	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1683	}
		1684
		1685	if (DECODE_WANTS_OCTETS (json))
954	sv_utf8_downgrade (string, 0);	1686	sv_utf8_downgrade (string, 0);
955	else	1687	else
956	sv_utf8_upgrade (string);	1688	sv_utf8_upgrade (string);
957		1689
958	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1690	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
959		1691
960	dec_t dec;	1692	dec.json = *json;
961	dec.flags = flags;
962	dec.cur = SvPVX (string);	1693	dec.cur = SvPVX (string);
963	dec.end = SvEND (string);	1694	dec.end = SvEND (string);
964	dec.err = 0;	1695	dec.err = 0;
965	dec.depth = 0;	1696	dec.depth = 0;
966	dec.maxdepth = DEC_DEPTH (dec.flags);
967		1697
968	*SvEND (sv) = 0; // this shou[ld basically be a nop, too	1698	if (dec.json.cb_object || dec.json.cb_sk_object)
		1699	dec.json.flags |= F_HOOK;
		1700
		1701	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1702
		1703	decode_ws (&dec);
969	sv = decode_sv (&dec);	1704	sv = decode_sv (&dec);
970		1705
		1706	if (offset_return)
		1707	*offset_return = dec.cur;
		1708
		1709	if (!(offset_return || !sv))
		1710	{
		1711	// check for trailing garbage
		1712	decode_ws (&dec);
		1713
		1714	if (*dec.cur)
		1715	{
		1716	dec.err = "garbage after JSON object";
		1717	SvREFCNT_dec (sv);
		1718	sv = 0;
		1719	}
		1720	}
		1721
971	if (!sv)	1722	if (!sv)
972	{	1723	{
973	IV offset = dec.flags & F_UTF8
974	? dec.cur - SvPVX (string)
975	: utf8_distance (dec.cur, SvPVX (string));
976	SV *uni = sv_newmortal ();	1724	SV *uni = sv_newmortal ();
977		1725
978	// horrible hack to silence warning inside pv_uni_display	1726	// horrible hack to silence warning inside pv_uni_display
979	COP cop = *PL_curcop;	1727	COP cop = *PL_curcop;
980	cop.cop_warnings = pWARN_NONE;	1728	cop.cop_warnings = pWARN_NONE;
…		…
982	SAVEVPTR (PL_curcop);	1730	SAVEVPTR (PL_curcop);
983	PL_curcop = &cop;	1731	PL_curcop = &cop;
984	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1732	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
985	LEAVE;	1733	LEAVE;
986		1734
987	croak ("%s, at character offset %d (%s)",	1735	croak ("%s, at character offset %d (before \"%s\")",
988	dec.err,	1736	dec.err,
989	(int)offset,	1737	(int)ptr_to_index (string, dec.cur),
990	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1738	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
991	}	1739	}
992		1740
993	sv = sv_2mortal (sv);	1741	sv = sv_2mortal (sv);
994		1742
995	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1743	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
996	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1744	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
997		1745
998	return sv;	1746	return sv;
		1747	}
		1748
		1749	/////////////////////////////////////////////////////////////////////////////
		1750	// incremental parser
		1751
		1752	static void
		1753	incr_parse (JSON *self)
		1754	{
		1755	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1756
		1757	// the state machine here is a bit convoluted and could be simplified a lot
		1758	// but this would make it slower, so...
		1759
		1760	for (;;)
		1761	{
		1762	//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
		1763	switch (self->incr_mode)
		1764	{
		1765	// only used for initial whitespace skipping
		1766	case INCR_M_WS:
		1767	for (;;)
		1768	{
		1769	if (*p > 0x20)
		1770	{
		1771	if (*p == '#')
		1772	{
		1773	self->incr_mode = INCR_M_C0;
		1774	goto incr_m_c;
		1775	}
		1776	else
		1777	{
		1778	self->incr_mode = INCR_M_JSON;
		1779	goto incr_m_json;
		1780	}
		1781	}
		1782	else if (!*p)
		1783	goto interrupt;
		1784
		1785	++p;
		1786	}
		1787
		1788	// skip a single char inside a string (for \\-processing)
		1789	case INCR_M_BS:
		1790	if (!*p)
		1791	goto interrupt;
		1792
		1793	++p;
		1794	self->incr_mode = INCR_M_STR;
		1795	goto incr_m_str;
		1796
		1797	// inside #-style comments
		1798	case INCR_M_C0:
		1799	case INCR_M_C1:
		1800	incr_m_c:
		1801	for (;;)
		1802	{
		1803	if (*p == '\n')
		1804	{
		1805	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1806	break;
		1807	}
		1808	else if (!*p)
		1809	goto interrupt;
		1810
		1811	++p;
		1812	}
		1813
		1814	break;
		1815
		1816	// inside a string
		1817	case INCR_M_STR:
		1818	incr_m_str:
		1819	for (;;)
		1820	{
		1821	if (*p == '"')
		1822	{
		1823	++p;
		1824	self->incr_mode = INCR_M_JSON;
		1825
		1826	if (!self->incr_nest)
		1827	goto interrupt;
		1828
		1829	goto incr_m_json;
		1830	}
		1831	else if (*p == '\\')
		1832	{
		1833	++p; // "virtually" consumes character after \
		1834
		1835	if (!*p) // if at end of string we have to switch modes
		1836	{
		1837	self->incr_mode = INCR_M_BS;
		1838	goto interrupt;
		1839	}
		1840	}
		1841	else if (!*p)
		1842	goto interrupt;
		1843
		1844	++p;
		1845	}
		1846
		1847	// after initial ws, outside string
		1848	case INCR_M_JSON:
		1849	incr_m_json:
		1850	for (;;)
		1851	{
		1852	switch (*p++)
		1853	{
		1854	case 0:
		1855	--p;
		1856	goto interrupt;
		1857
		1858	case 0x09:
		1859	case 0x0a:
		1860	case 0x0d:
		1861	case 0x20:
		1862	if (!self->incr_nest)
		1863	{
		1864	--p; // do not eat the whitespace, let the next round do it
		1865	goto interrupt;
		1866	}
		1867	break;
		1868
		1869	case '"':
		1870	self->incr_mode = INCR_M_STR;
		1871	goto incr_m_str;
		1872
		1873	case '[':
		1874	case '{':
		1875	case '(':
		1876	if (++self->incr_nest > self->max_depth)
		1877	croak (ERR_NESTING_EXCEEDED);
		1878	break;
		1879
		1880	case ']':
		1881	case '}':
		1882	if (--self->incr_nest <= 0)
		1883	goto interrupt;
		1884	break;
		1885
		1886	case ')':
		1887	--self->incr_nest;
		1888	break;
		1889
		1890	case '#':
		1891	self->incr_mode = INCR_M_C1;
		1892	goto incr_m_c;
		1893	}
		1894	}
		1895	}
		1896
		1897	modechange:
		1898	;
		1899	}
		1900
		1901	interrupt:
		1902	self->incr_pos = p - SvPVX (self->incr_text);
		1903	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1904	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
999	}	1905	}
1000		1906
1001	/////////////////////////////////////////////////////////////////////////////	1907	/////////////////////////////////////////////////////////////////////////////
1002	// XS interface functions	1908	// XS interface functions
1003		1909
1004	MODULE = JSON::XS PACKAGE = JSON::XS	1910	MODULE = JSON::XS PACKAGE = JSON::XS
1005		1911
1006	BOOT:	1912	BOOT:
1007	{	1913	{
1008	int i;	1914	int i;
1009
1010	memset (decode_hexdigit, 0xff, 256);
1011		1915
1012	for (i = 0; i < 256; ++i)	1916	for (i = 0; i < 256; ++i)
1013	decode_hexdigit [i] =	1917	decode_hexdigit [i] =
1014	i >= '0' && i <= '9' ? i - '0'	1918	i >= '0' && i <= '9' ? i - '0'
1015	: i >= 'a' && i <= 'f' ? i - 'a' + 10	1919	: i >= 'a' && i <= 'f' ? i - 'a' + 10
1016	: i >= 'A' && i <= 'F' ? i - 'A' + 10	1920	: i >= 'A' && i <= 'F' ? i - 'A' + 10
1017	: -1;	1921	: -1;
1018		1922
1019	json_stash = gv_stashpv ("JSON::XS", 1);	1923	json_stash = gv_stashpv ("JSON::XS" , 1);
		1924	types_boolean_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
		1925
		1926	types_true = get_bool ("Types::Serialiser::true");
		1927	types_false = get_bool ("Types::Serialiser::false");
		1928
		1929	sv_json = newSVpv ("JSON", 0);
		1930	SvREADONLY_on (sv_json);
		1931
		1932	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1020	}	1933	}
1021		1934
1022	PROTOTYPES: DISABLE	1935	PROTOTYPES: DISABLE
1023		1936
1024	SV *new (char *dummy)	1937	void CLONE (...)
1025	CODE:	1938	CODE:
1026	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1939	json_stash = 0;
		1940	types_boolean_stash = 0;
		1941
		1942	void new (char *klass)
		1943	PPCODE:
		1944	{
		1945	SV *pv = NEWSV (0, sizeof (JSON));
		1946	SvPOK_only (pv);
		1947	json_init ((JSON *)SvPVX (pv));
		1948	XPUSHs (sv_2mortal (sv_bless (
		1949	newRV_noinc (pv),
		1950	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1951	)));
		1952	}
		1953
		1954	void ascii (JSON *self, int enable = 1)
		1955	ALIAS:
		1956	ascii = F_ASCII
		1957	latin1 = F_LATIN1
		1958	utf8 = F_UTF8
		1959	indent = F_INDENT
		1960	canonical = F_CANONICAL
		1961	space_before = F_SPACE_BEFORE
		1962	space_after = F_SPACE_AFTER
		1963	pretty = F_PRETTY
		1964	allow_nonref = F_ALLOW_NONREF
		1965	shrink = F_SHRINK
		1966	allow_blessed = F_ALLOW_BLESSED
		1967	convert_blessed = F_CONV_BLESSED
		1968	relaxed = F_RELAXED
		1969	allow_unknown = F_ALLOW_UNKNOWN
		1970	allow_tags = F_ALLOW_TAGS
		1971	PPCODE:
		1972	{
		1973	if (enable)
		1974	self->flags |= ix;
		1975	else
		1976	self->flags &= ~ix;
		1977
		1978	XPUSHs (ST (0));
		1979	}
		1980
		1981	void get_ascii (JSON *self)
		1982	ALIAS:
		1983	get_ascii = F_ASCII
		1984	get_latin1 = F_LATIN1
		1985	get_utf8 = F_UTF8
		1986	get_indent = F_INDENT
		1987	get_canonical = F_CANONICAL
		1988	get_space_before = F_SPACE_BEFORE
		1989	get_space_after = F_SPACE_AFTER
		1990	get_allow_nonref = F_ALLOW_NONREF
		1991	get_shrink = F_SHRINK
		1992	get_allow_blessed = F_ALLOW_BLESSED
		1993	get_convert_blessed = F_CONV_BLESSED
		1994	get_relaxed = F_RELAXED
		1995	get_allow_unknown = F_ALLOW_UNKNOWN
		1996	get_allow_tags = F_ALLOW_TAGS
		1997	PPCODE:
		1998	XPUSHs (boolSV (self->flags & ix));
		1999
		2000	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		2001	PPCODE:
		2002	self->max_depth = max_depth;
		2003	XPUSHs (ST (0));
		2004
		2005	U32 get_max_depth (JSON *self)
		2006	CODE:
		2007	RETVAL = self->max_depth;
1027	OUTPUT:	2008	OUTPUT:
1028	RETVAL	2009	RETVAL
1029		2010
1030	SV *ascii (SV *self, int enable = 1)	2011	void max_size (JSON *self, U32 max_size = 0)
1031	ALIAS:	2012	PPCODE:
1032	ascii = F_ASCII	2013	self->max_size = max_size;
1033	utf8 = F_UTF8	2014	XPUSHs (ST (0));
1034	indent = F_INDENT	2015
1035	canonical = F_CANONICAL	2016	int get_max_size (JSON *self)
1036	space_before = F_SPACE_BEFORE
1037	space_after = F_SPACE_AFTER
1038	pretty = F_PRETTY
1039	allow_nonref = F_ALLOW_NONREF
1040	shrink = F_SHRINK
1041	CODE:	2017	CODE:
1042	{	2018	RETVAL = self->max_size;
1043	UV *uv = SvJSON (self);
1044	if (enable)
1045	*uv |= ix;
1046	else
1047	*uv &= ~ix;
1048
1049	RETVAL = newSVsv (self);
1050	}
1051	OUTPUT:	2019	OUTPUT:
1052	RETVAL	2020	RETVAL
1053		2021
1054	SV *max_depth (SV *self, int max_depth = 0x80000000UL)	2022	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
		2023	PPCODE:
		2024	{
		2025	SvREFCNT_dec (self->cb_object);
		2026	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		2027
		2028	XPUSHs (ST (0));
		2029	}
		2030
		2031	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
		2032	PPCODE:
		2033	{
		2034	if (!self->cb_sk_object)
		2035	self->cb_sk_object = newHV ();
		2036
		2037	if (SvOK (cb))
		2038	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		2039	else
		2040	{
		2041	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		2042
		2043	if (!HvKEYS (self->cb_sk_object))
		2044	{
		2045	SvREFCNT_dec (self->cb_sk_object);
		2046	self->cb_sk_object = 0;
		2047	}
		2048	}
		2049
		2050	XPUSHs (ST (0));
		2051	}
		2052
		2053	void encode (JSON *self, SV *scalar)
		2054	PPCODE:
		2055	PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
		2056	XPUSHs (scalar);
		2057
		2058	void decode (JSON *self, SV *jsonstr)
		2059	PPCODE:
		2060	PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
		2061	XPUSHs (jsonstr);
		2062
		2063	void decode_prefix (JSON *self, SV *jsonstr)
		2064	PPCODE:
		2065	{
		2066	SV *sv;
		2067	char *offset;
		2068	PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
		2069	EXTEND (SP, 2);
		2070	PUSHs (sv);
		2071	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		2072	}
		2073
		2074	void incr_parse (JSON *self, SV *jsonstr = 0)
		2075	PPCODE:
		2076	{
		2077	if (!self->incr_text)
		2078	self->incr_text = newSVpvn ("", 0);
		2079
		2080	/* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
		2081	if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
		2082	if (DECODE_WANTS_OCTETS (self))
		2083	{
		2084	if (self->incr_pos)
		2085	self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
		2086	(U8 *)SvPVX (self->incr_text) + self->incr_pos);
		2087
		2088	sv_utf8_downgrade (self->incr_text, 0);
		2089	}
		2090	else
		2091	{
		2092	sv_utf8_upgrade (self->incr_text);
		2093
		2094	if (self->incr_pos)
		2095	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		2096	- (U8 *)SvPVX (self->incr_text);
		2097	}
		2098
		2099	// append data, if any
		2100	if (jsonstr)
		2101	{
		2102	/* make sure both strings have same encoding */
		2103	if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
		2104	if (SvUTF8 (jsonstr))
		2105	sv_utf8_downgrade (jsonstr, 0);
		2106	else
		2107	sv_utf8_upgrade (jsonstr);
		2108
		2109	/* and then just blindly append */
		2110	{
		2111	STRLEN len;
		2112	const char *str = SvPV (jsonstr, len);
		2113	STRLEN cur = SvCUR (self->incr_text);
		2114
		2115	if (SvLEN (self->incr_text) <= cur + len)
		2116	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		2117
		2118	Move (str, SvEND (self->incr_text), len, char);
		2119	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		2120	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		2121	}
		2122	}
		2123
		2124	if (GIMME_V != G_VOID)
		2125	do
		2126	{
		2127	SV *sv;
		2128	char *offset;
		2129
		2130	if (!INCR_DONE (self))
		2131	{
		2132	incr_parse (self);
		2133
		2134	if (self->incr_pos > self->max_size && self->max_size)
		2135	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		2136	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		2137
		2138	if (!INCR_DONE (self))
		2139	{
		2140	// as an optimisation, do not accumulate white space in the incr buffer
		2141	if (self->incr_mode == INCR_M_WS && self->incr_pos)
		2142	{
		2143	self->incr_pos = 0;
		2144	SvCUR_set (self->incr_text, 0);
		2145	}
		2146
		2147	break;
		2148	}
		2149	}
		2150
		2151	PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
		2152	XPUSHs (sv);
		2153
		2154	self->incr_pos -= offset - SvPVX (self->incr_text);
		2155	self->incr_nest = 0;
		2156	self->incr_mode = 0;
		2157
		2158	sv_chop (self->incr_text, offset);
		2159	}
		2160	while (GIMME_V == G_ARRAY);
		2161	}
		2162
		2163	SV *incr_text (JSON *self)
		2164	ATTRS: lvalue
1055	CODE:	2165	CODE:
1056	{	2166	{
1057	UV *uv = SvJSON (self);	2167	if (self->incr_pos)
1058	UV log2 = 0;	2168	croak ("incr_text can not be called when the incremental parser already started parsing");
1059		2169
1060	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	2170	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
1061
1062	while ((1UL << log2) < max_depth)
1063	++log2;
1064
1065	*uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1066
1067	RETVAL = newSVsv (self);
1068	}	2171	}
1069	OUTPUT:	2172	OUTPUT:
1070	RETVAL	2173	RETVAL
1071		2174
		2175	void incr_skip (JSON *self)
		2176	CODE:
		2177	{
		2178	if (self->incr_pos)
		2179	{
		2180	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2181	self->incr_pos = 0;
		2182	self->incr_nest = 0;
		2183	self->incr_mode = 0;
		2184	}
		2185	}
		2186
		2187	void incr_reset (JSON *self)
		2188	CODE:
		2189	{
		2190	SvREFCNT_dec (self->incr_text);
		2191	self->incr_text = 0;
		2192	self->incr_pos = 0;
		2193	self->incr_nest = 0;
		2194	self->incr_mode = 0;
		2195	}
		2196
		2197	void DESTROY (JSON *self)
		2198	CODE:
		2199	SvREFCNT_dec (self->cb_sk_object);
		2200	SvREFCNT_dec (self->cb_object);
		2201	SvREFCNT_dec (self->incr_text);
		2202
		2203	PROTOTYPES: ENABLE
		2204
1072	void encode (SV *self, SV *scalar)	2205	void encode_json (SV *scalar)
1073	PPCODE:	2206	PPCODE:
1074	XPUSHs (encode_json (scalar, *SvJSON (self)));	2207	{
		2208	JSON json;
		2209	json_init (&json);
		2210	json.flags |= F_UTF8;
		2211	PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
		2212	XPUSHs (scalar);
		2213	}
1075		2214
1076	void decode (SV *self, SV *jsonstr)	2215	void decode_json (SV *jsonstr)
1077	PPCODE:	2216	PPCODE:
1078	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	2217	{
		2218	JSON json;
		2219	json_init (&json);
		2220	json.flags |= F_UTF8;
		2221	PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
		2222	XPUSHs (jsonstr);
		2223	}
1079		2224
1080	PROTOTYPES: ENABLE
1081
1082	void to_json (SV *scalar)
1083	ALIAS:
1084	objToJson = 0
1085	PPCODE:
1086	XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8));
1087
1088	void from_json (SV *jsonstr)
1089	ALIAS:
1090	jsonToObj = 0
1091	PPCODE:
1092	XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8));
1093

Diff Legend

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