ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.35 by root, Wed Jun 6 14:52:49 2007 UTC vs.
Revision 1.107 by root, Wed Mar 17 01:45:43 2010 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_HOOK 0x00080000UL // some hooks exist, so slow-path processing
26	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
27
28	// F_SELFCONVERT? <=> to_json/toJson
29	// F_BLESSED? <=> { $__class__$ => }
30		39
31	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	40	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
32	#define F_DEFAULT (9UL << S_MAXDEPTH)
33		41
34	#define INIT_SIZE 32 // initial scalar size to be allocated	42	#define INIT_SIZE 32 // initial scalar size to be allocated
35	#define INDENT_STEP 3 // spaces per indentation level	43	#define INDENT_STEP 3 // spaces per indentation level
36		44
37	#define SHORT_STRING_LEN 512 // special-case strings of up to this size	45	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
38		46
39	#define SB do {	47	#define SB do {
40	#define SE } while (0)	48	#define SE } while (0)
41		49
42	#if __GNUC__ >= 3	50	#if __GNUC__ >= 3
43	# define expect(expr,value) __builtin_expect ((expr),(value))	51	# define expect(expr,value) __builtin_expect ((expr), (value))
44	# define inline inline	52	# define INLINE static inline
45	#else	53	#else
46	# define expect(expr,value) (expr)	54	# define expect(expr,value) (expr)
47	# define inline static	55	# define INLINE static
48	#endif	56	#endif
49		57
50	#define expect_false(expr) expect ((expr) != 0, 0)	58	#define expect_false(expr) expect ((expr) != 0, 0)
51	#define expect_true(expr) expect ((expr) != 0, 1)	59	#define expect_true(expr) expect ((expr) != 0, 1)
52		60
		61	#define IN_RANGE_INC(type,val,beg,end) \
		62	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		63	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		64
		65	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
		66
		67	#ifdef USE_ITHREADS
		68	# define JSON_SLOW 1
		69	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		70	#else
		71	# define JSON_SLOW 0
		72	# define JSON_STASH json_stash
		73	#endif
		74
53	static HV *json_stash; // JSON::XS::	75	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		76	static SV *json_true, *json_false;
		77
		78	enum {
		79	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		80	INCR_M_STR, // inside string
		81	INCR_M_BS, // inside backslash
		82	INCR_M_C0, // inside comment in initial whitespace sequence
		83	INCR_M_C1, // inside comment in other places
		84	INCR_M_JSON // outside anything, count nesting
		85	};
		86
		87	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		88
		89	typedef struct {
		90	U32 flags;
		91	U32 max_depth;
		92	STRLEN max_size;
		93
		94	SV *cb_object;
		95	HV *cb_sk_object;
		96
		97	// for the incremental parser
		98	SV *incr_text; // the source text so far
		99	STRLEN incr_pos; // the current offset into the text
		100	int incr_nest; // {[]}-nesting level
		101	unsigned char incr_mode;
		102	} JSON;
		103
		104	INLINE void
		105	json_init (JSON *json)
		106	{
		107	Zero (json, 1, JSON);
		108	json->max_depth = 512;
		109	}
54		110
55	/////////////////////////////////////////////////////////////////////////////	111	/////////////////////////////////////////////////////////////////////////////
56	// utility functions	112	// utility functions
57		113
58	static UV *	114	INLINE SV *
59	SvJSON (SV *sv)	115	get_bool (const char *name)
60	{	116	{
61	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))	117	SV *sv = get_sv (name, 1);
62	croak ("object is not of type JSON::XS");
63		118
64	return &SvUVX (SvRV (sv));	119	SvREADONLY_on (sv);
65	}	120	SvREADONLY_on (SvRV (sv));
66		121
67	static void	122	return sv;
		123	}
		124
		125	INLINE void
68	shrink (SV *sv)	126	shrink (SV *sv)
69	{	127	{
70	sv_utf8_downgrade (sv, 1);	128	sv_utf8_downgrade (sv, 1);
		129
71	if (SvLEN (sv) > SvCUR (sv) + 1)	130	if (SvLEN (sv) > SvCUR (sv) + 1)
72	{	131	{
73	#ifdef SvPV_shrink_to_cur	132	#ifdef SvPV_shrink_to_cur
74	SvPV_shrink_to_cur (sv);	133	SvPV_shrink_to_cur (sv);
75	#elif defined (SvPV_renew)	134	#elif defined (SvPV_renew)
…		…
81	// decode an utf-8 character and return it, or (UV)-1 in	140	// decode an utf-8 character and return it, or (UV)-1 in
82	// case of an error.	141	// case of an error.
83	// we special-case "safe" characters from U+80 .. U+7FF,	142	// we special-case "safe" characters from U+80 .. U+7FF,
84	// but use the very good perl function to parse anything else.	143	// but use the very good perl function to parse anything else.
85	// note that we never call this function for a ascii codepoints	144	// note that we never call this function for a ascii codepoints
86	inline UV	145	INLINE UV
87	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)	146	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
88	{	147	{
89	if (expect_false (s[0] > 0xdf || s[0] < 0xc2))	148	if (expect_true (len >= 2
90	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	149	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
91	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	150	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
92	{	151	{
93	*clen = 2;	152	*clen = 2;
94	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);	153	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
95	}	154	}
96	else	155	else
97	{	156	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
98	*clen = (STRLEN)-1;	157	}
99	return (UV)-1;	158
		159	// likewise for encoding, also never called for ascii codepoints
		160	// this function takes advantage of this fact, although current gccs
		161	// seem to optimise the check for >= 0x80 away anyways
		162	INLINE unsigned char *
		163	encode_utf8 (unsigned char *s, UV ch)
		164	{
		165	if (expect_false (ch < 0x000080))
		166	*s++ = ch;
		167	else if (expect_true (ch < 0x000800))
		168	*s++ = 0xc0 | ( ch >> 6),
		169	*s++ = 0x80 | ( ch & 0x3f);
		170	else if ( ch < 0x010000)
		171	*s++ = 0xe0 | ( ch >> 12),
		172	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		173	*s++ = 0x80 | ( ch & 0x3f);
		174	else if ( ch < 0x110000)
		175	*s++ = 0xf0 | ( ch >> 18),
		176	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		177	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		178	*s++ = 0x80 | ( ch & 0x3f);
		179
		180	return s;
		181	}
		182
		183	// convert offset pointer to character index, sv must be string
		184	static STRLEN
		185	ptr_to_index (SV *sv, char *offset)
		186	{
		187	return SvUTF8 (sv)
		188	? utf8_distance (offset, SvPVX (sv))
		189	: offset - SvPVX (sv);
		190	}
		191
		192	/////////////////////////////////////////////////////////////////////////////
		193	// fp hell
		194
		195	// scan a group of digits, and a trailing exponent
		196	static void
		197	json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
		198	{
		199	UV uaccum = 0;
		200	int eaccum = 0;
		201
		202	// if we recurse too deep, skip all remaining digits
		203	// to avoid a stack overflow attack
		204	if (expect_false (--maxdepth <= 0))
		205	while (((U8)*s - '0') < 10)
		206	++s;
		207
		208	for (;;)
100	}	209	{
101	}	210	U8 dig = (U8)*s - '0';
102		211
		212	if (expect_false (dig >= 10))
		213	{
		214	if (dig == (U8)((U8)'.' - (U8)'0'))
		215	{
		216	++s;
		217	json_atof_scan1 (s, accum, expo, 1, maxdepth);
		218	}
		219	else if ((dig | ' ') == 'e' - '0')
		220	{
		221	int exp2 = 0;
		222	int neg = 0;
		223
		224	++s;
		225
		226	if (*s == '-')
		227	{
		228	++s;
		229	neg = 1;
		230	}
		231	else if (*s == '+')
		232	++s;
		233
		234	while ((dig = (U8)*s - '0') < 10)
		235	exp2 = exp2 * 10 + *s++ - '0';
		236
		237	*expo += neg ? -exp2 : exp2;
		238	}
		239
		240	break;
		241	}
		242
		243	++s;
		244
		245	uaccum = uaccum * 10 + dig;
		246	++eaccum;
		247
		248	// if we have too many digits, then recurse for more
		249	// we actually do this for rather few digits
		250	if (uaccum >= (UV_MAX - 9) / 10)
		251	{
		252	if (postdp) *expo -= eaccum;
		253	json_atof_scan1 (s, accum, expo, postdp, maxdepth);
		254	if (postdp) *expo += eaccum;
		255
		256	break;
		257	}
		258	}
		259
		260	// this relies greatly on the quality of the pow ()
		261	// implementation of the platform, but a good
		262	// implementation is hard to beat.
		263	if (postdp) *expo -= eaccum;
		264	*accum += uaccum * Perl_pow (10., *expo);
		265	*expo += eaccum;
		266	}
		267
		268	static NV
		269	json_atof (const char *s)
		270	{
		271	NV accum = 0.;
		272	int expo = 0;
		273	int neg = 0;
		274
		275	if (*s == '-')
		276	{
		277	++s;
		278	neg = 1;
		279	}
		280
		281	// a recursion depth of ten gives us >>500 bits
		282	json_atof_scan1 (s, &accum, &expo, 0, 10);
		283
		284	return neg ? -accum : accum;
		285	}
103	/////////////////////////////////////////////////////////////////////////////	286	/////////////////////////////////////////////////////////////////////////////
104	// encoder	287	// encoder
105		288
106	// structure used for encoding JSON	289	// structure used for encoding JSON
107	typedef struct	290	typedef struct
108	{	291	{
109	char *cur; // SvPVX (sv) + current output position	292	char *cur; // SvPVX (sv) + current output position
110	char *end; // SvEND (sv)	293	char *end; // SvEND (sv)
111	SV *sv; // result scalar	294	SV *sv; // result scalar
112	U32 flags; // F_*	295	JSON json;
113	U32 indent; // indentation level	296	U32 indent; // indentation level
114	U32 maxdepth; // max. indentation/recursion level	297	UV limit; // escape character values >= this value when encoding
115	} enc_t;	298	} enc_t;
116		299
117	inline void	300	INLINE void
118	need (enc_t *enc, STRLEN len)	301	need (enc_t *enc, STRLEN len)
119	{	302	{
120	if (expect_false (enc->cur + len >= enc->end))	303	if (expect_false (enc->cur + len >= enc->end))
121	{	304	{
122	STRLEN cur = enc->cur - SvPVX (enc->sv);	305	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
123	SvGROW (enc->sv, cur + len + 1);	306	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
124	enc->cur = SvPVX (enc->sv) + cur;	307	enc->cur = SvPVX (enc->sv) + cur;
125	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	308	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
126	}	309	}
127	}	310	}
128		311
129	inline void	312	INLINE void
130	encode_ch (enc_t *enc, char ch)	313	encode_ch (enc_t *enc, char ch)
131	{	314	{
132	need (enc, 1);	315	need (enc, 1);
133	*enc->cur++ = ch;	316	*enc->cur++ = ch;
134	}	317	}
…		…
178	STRLEN clen;	361	STRLEN clen;
179	UV uch;	362	UV uch;
180		363
181	if (is_utf8)	364	if (is_utf8)
182	{	365	{
183	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
184	uch = decode_utf8 (str, end - str, &clen);	366	uch = decode_utf8 (str, end - str, &clen);
185	if (clen == (STRLEN)-1)	367	if (clen == (STRLEN)-1)
186	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	368	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
187	}	369	}
188	else	370	else
189	{	371	{
190	uch = ch;	372	uch = ch;
191	clen = 1;	373	clen = 1;
192	}	374	}
193		375
194	if (uch > 0x10FFFFUL)	376	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
195	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
196
197	if (uch < 0x80 || enc->flags & F_ASCII || (enc->flags & F_LATIN1 && uch > 0xFF))
198	{	377	{
199	if (uch > 0xFFFFUL)	378	if (uch >= 0x10000UL)
200	{	379	{
		380	if (uch >= 0x110000UL)
		381	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		382
201	need (enc, len += 11);	383	need (enc, len += 11);
202	sprintf (enc->cur, "\\u%04x\\u%04x",	384	sprintf (enc->cur, "\\u%04x\\u%04x",
203	(int)((uch - 0x10000) / 0x400 + 0xD800),	385	(int)((uch - 0x10000) / 0x400 + 0xD800),
204	(int)((uch - 0x10000) % 0x400 + 0xDC00));	386	(int)((uch - 0x10000) % 0x400 + 0xDC00));
205	enc->cur += 12;	387	enc->cur += 12;
206	}	388	}
207	else	389	else
208	{	390	{
209	static char hexdigit [16] = "0123456789abcdef";
210	need (enc, len += 5);	391	need (enc, len += 5);
211	*enc->cur++ = '\\';	392	*enc->cur++ = '\\';
212	*enc->cur++ = 'u';	393	*enc->cur++ = 'u';
213	*enc->cur++ = hexdigit [ uch >> 12 ];	394	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
214	*enc->cur++ = hexdigit [(uch >> 8) & 15];	395	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
215	*enc->cur++ = hexdigit [(uch >> 4) & 15];	396	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
216	*enc->cur++ = hexdigit [(uch >> 0) & 15];	397	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
217	}	398	}
218		399
219	str += clen;	400	str += clen;
220	}	401	}
221	else if (enc->flags & F_LATIN1)	402	else if (enc->json.flags & F_LATIN1)
222	{	403	{
223	*enc->cur++ = uch;	404	*enc->cur++ = uch;
224	str += clen;	405	str += clen;
225	}	406	}
226	else if (is_utf8)	407	else if (is_utf8)
…		…
233	while (--clen);	414	while (--clen);
234	}	415	}
235	else	416	else
236	{	417	{
237	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	418	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
238	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	419	enc->cur = encode_utf8 (enc->cur, uch);
239	++str;	420	++str;
240	}	421	}
241	}	422	}
242	}	423	}
243	}	424	}
244		425
245	--len;	426	--len;
246	}	427	}
247	}	428	}
248		429
249	inline void	430	INLINE void
250	encode_indent (enc_t *enc)	431	encode_indent (enc_t *enc)
251	{	432	{
252	if (enc->flags & F_INDENT)	433	if (enc->json.flags & F_INDENT)
253	{	434	{
254	int spaces = enc->indent * INDENT_STEP;	435	int spaces = enc->indent * INDENT_STEP;
255		436
256	need (enc, spaces);	437	need (enc, spaces);
257	memset (enc->cur, ' ', spaces);	438	memset (enc->cur, ' ', spaces);
258	enc->cur += spaces;	439	enc->cur += spaces;
259	}	440	}
260	}	441	}
261		442
262	inline void	443	INLINE void
263	encode_space (enc_t *enc)	444	encode_space (enc_t *enc)
264	{	445	{
265	need (enc, 1);	446	need (enc, 1);
266	encode_ch (enc, ' ');	447	encode_ch (enc, ' ');
267	}	448	}
268		449
269	inline void	450	INLINE void
270	encode_nl (enc_t *enc)	451	encode_nl (enc_t *enc)
271	{	452	{
272	if (enc->flags & F_INDENT)	453	if (enc->json.flags & F_INDENT)
273	{	454	{
274	need (enc, 1);	455	need (enc, 1);
275	encode_ch (enc, '\n');	456	encode_ch (enc, '\n');
276	}	457	}
277	}	458	}
278		459
279	inline void	460	INLINE void
280	encode_comma (enc_t *enc)	461	encode_comma (enc_t *enc)
281	{	462	{
282	encode_ch (enc, ',');	463	encode_ch (enc, ',');
283		464
284	if (enc->flags & F_INDENT)	465	if (enc->json.flags & F_INDENT)
285	encode_nl (enc);	466	encode_nl (enc);
286	else if (enc->flags & F_SPACE_AFTER)	467	else if (enc->json.flags & F_SPACE_AFTER)
287	encode_space (enc);	468	encode_space (enc);
288	}	469	}
289		470
290	static void encode_sv (enc_t *enc, SV *sv);	471	static void encode_sv (enc_t *enc, SV *sv);
291		472
292	static void	473	static void
293	encode_av (enc_t *enc, AV *av)	474	encode_av (enc_t *enc, AV *av)
294	{	475	{
295	int i, len = av_len (av);	476	int i, len = av_len (av);
296		477
297	if (enc->indent >= enc->maxdepth)	478	if (enc->indent >= enc->json.max_depth)
298	croak ("data structure too deep (hit recursion limit)");	479	croak (ERR_NESTING_EXCEEDED);
299		480
300	encode_ch (enc, '['); encode_nl (enc);	481	encode_ch (enc, '[');
301	++enc->indent;	482
		483	if (len >= 0)
		484	{
		485	encode_nl (enc); ++enc->indent;
302		486
303	for (i = 0; i <= len; ++i)	487	for (i = 0; i <= len; ++i)
304	{	488	{
		489	SV **svp = av_fetch (av, i, 0);
		490
305	encode_indent (enc);	491	encode_indent (enc);
306	encode_sv (enc, *av_fetch (av, i, 0));
307		492
		493	if (svp)
		494	encode_sv (enc, *svp);
		495	else
		496	encode_str (enc, "null", 4, 0);
		497
308	if (i < len)	498	if (i < len)
309	encode_comma (enc);	499	encode_comma (enc);
310	}	500	}
311		501
		502	encode_nl (enc); --enc->indent; encode_indent (enc);
		503	}
		504
312	encode_nl (enc);	505	encode_ch (enc, ']');
313
314	--enc->indent;
315	encode_indent (enc); encode_ch (enc, ']');
316	}	506	}
317		507
318	static void	508	static void
319	encode_he (enc_t *enc, HE *he)	509	encode_hk (enc_t *enc, HE *he)
320	{	510	{
321	encode_ch (enc, '"');	511	encode_ch (enc, '"');
322		512
323	if (HeKLEN (he) == HEf_SVKEY)	513	if (HeKLEN (he) == HEf_SVKEY)
324	{	514	{
…		…
334	else	524	else
335	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	525	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
336		526
337	encode_ch (enc, '"');	527	encode_ch (enc, '"');
338		528
339	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	529	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
340	encode_ch (enc, ':');	530	encode_ch (enc, ':');
341	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	531	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
342	encode_sv (enc, HeVAL (he));
343	}	532	}
344		533
345	// compare hash entries, used when all keys are bytestrings	534	// compare hash entries, used when all keys are bytestrings
346	static int	535	static int
347	he_cmp_fast (const void *a_, const void *b_)	536	he_cmp_fast (const void *a_, const void *b_)
…		…
352	HE *b = *(HE **)b_;	541	HE *b = *(HE **)b_;
353		542
354	STRLEN la = HeKLEN (a);	543	STRLEN la = HeKLEN (a);
355	STRLEN lb = HeKLEN (b);	544	STRLEN lb = HeKLEN (b);
356		545
357	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	546	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
358	cmp = la - lb;	547	cmp = lb - la;
359		548
360	return cmp;	549	return cmp;
361	}	550	}
362		551
363	// compare hash entries, used when some keys are sv's or utf-x	552	// compare hash entries, used when some keys are sv's or utf-x
364	static int	553	static int
365	he_cmp_slow (const void *a, const void *b)	554	he_cmp_slow (const void *a, const void *b)
366	{	555	{
367	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	556	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
368	}	557	}
369		558
370	static void	559	static void
371	encode_hv (enc_t *enc, HV *hv)	560	encode_hv (enc_t *enc, HV *hv)
372	{	561	{
373	int count, i;	562	HE *he;
374		563
375	if (enc->indent >= enc->maxdepth)	564	if (enc->indent >= enc->json.max_depth)
376	croak ("data structure too deep (hit recursion limit)");	565	croak (ERR_NESTING_EXCEEDED);
377		566
378	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	567	encode_ch (enc, '{');
379		568
380	if ((count = hv_iterinit (hv)))
381	{
382	// for canonical output we have to sort by keys first	569	// for canonical output we have to sort by keys first
383	// actually, this is mostly due to the stupid so-called	570	// actually, this is mostly due to the stupid so-called
384	// security workaround added somewhere in 5.8.x.	571	// security workaround added somewhere in 5.8.x
385	// that randomises hash orderings	572	// that randomises hash orderings
386	if (enc->flags & F_CANONICAL)	573	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
		574	{
		575	int count = hv_iterinit (hv);
		576
		577	if (SvMAGICAL (hv))
387	{	578	{
		579	// need to count by iterating. could improve by dynamically building the vector below
		580	// but I don't care for the speed of this special case.
		581	// note also that we will run into undefined behaviour when the two iterations
		582	// do not result in the same count, something I might care for in some later release.
		583
		584	count = 0;
		585	while (hv_iternext (hv))
		586	++count;
		587
		588	hv_iterinit (hv);
		589	}
		590
		591	if (count)
		592	{
388	int fast = 1;	593	int i, fast = 1;
389	HE *he;
390	#if defined(__BORLANDC__) || defined(_MSC_VER)	594	#if defined(__BORLANDC__) || defined(_MSC_VER)
391	HE **hes = _alloca (count * sizeof (HE));	595	HE **hes = _alloca (count * sizeof (HE));
392	#else	596	#else
393	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode	597	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
394	#endif	598	#endif
…		…
421		625
422	FREETMPS;	626	FREETMPS;
423	LEAVE;	627	LEAVE;
424	}	628	}
425		629
426	for (i = 0; i < count; ++i)	630	encode_nl (enc); ++enc->indent;
		631
		632	while (count--)
427	{	633	{
428	encode_indent (enc);	634	encode_indent (enc);
		635	he = hes [count];
429	encode_he (enc, hes [i]);	636	encode_hk (enc, he);
		637	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
430		638
431	if (i < count - 1)	639	if (count)
432	encode_comma (enc);	640	encode_comma (enc);
433	}	641	}
434		642
		643	encode_nl (enc); --enc->indent; encode_indent (enc);
		644	}
		645	}
		646	else
		647	{
		648	if (hv_iterinit (hv) || SvMAGICAL (hv))
		649	if ((he = hv_iternext (hv)))
		650	{
		651	encode_nl (enc); ++enc->indent;
		652
		653	for (;;)
		654	{
435	encode_nl (enc);	655	encode_indent (enc);
		656	encode_hk (enc, he);
		657	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		658
		659	if (!(he = hv_iternext (hv)))
		660	break;
		661
		662	encode_comma (enc);
		663	}
		664
		665	encode_nl (enc); --enc->indent; encode_indent (enc);
		666	}
		667	}
		668
		669	encode_ch (enc, '}');
		670	}
		671
		672	// encode objects, arrays and special \0=false and \1=true values.
		673	static void
		674	encode_rv (enc_t *enc, SV *sv)
		675	{
		676	svtype svt;
		677
		678	SvGETMAGIC (sv);
		679	svt = SvTYPE (sv);
		680
		681	if (expect_false (SvOBJECT (sv)))
		682	{
		683	HV *stash = !JSON_SLOW || json_boolean_stash
		684	? json_boolean_stash
		685	: gv_stashpv ("JSON::XS::Boolean", 1);
		686
		687	if (SvSTASH (sv) == stash)
		688	{
		689	if (SvIV (sv))
		690	encode_str (enc, "true", 4, 0);
		691	else
		692	encode_str (enc, "false", 5, 0);
436	}	693	}
437	else	694	else
438	{	695	{
439	HE *he = hv_iternext (hv);	696	#if 0
440		697	if (0 && sv_derived_from (rv, "JSON::Literal"))
441	for (;;)
442	{	698	{
443	encode_indent (enc);	699	// not yet
444	encode_he (enc, he);
445
446	if (!(he = hv_iternext (hv)))
447	break;
448
449	encode_comma (enc);
450	}	700	}
		701	#endif
		702	if (enc->json.flags & F_CONV_BLESSED)
		703	{
		704	// we re-bless the reference to get overload and other niceties right
		705	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
451		706
		707	if (to_json)
		708	{
		709	dSP;
		710
		711	ENTER; SAVETMPS; PUSHMARK (SP);
		712	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		713
		714	// calling with G_SCALAR ensures that we always get a 1 return value
		715	PUTBACK;
		716	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		717	SPAGAIN;
		718
		719	// catch this surprisingly common error
		720	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		721	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		722
		723	sv = POPs;
		724	PUTBACK;
		725
452	encode_nl (enc);	726	encode_sv (enc, sv);
		727
		728	FREETMPS; LEAVE;
		729	}
		730	else if (enc->json.flags & F_ALLOW_BLESSED)
		731	encode_str (enc, "null", 4, 0);
		732	else
		733	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		734	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		735	}
		736	else if (enc->json.flags & F_ALLOW_BLESSED)
		737	encode_str (enc, "null", 4, 0);
		738	else
		739	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		740	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
453	}	741	}
454	}	742	}
455
456	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
457	}
458
459	// encode objects, arrays and special \0=false and \1=true values.
460	static void
461	encode_rv (enc_t *enc, SV *sv)
462	{
463	svtype svt;
464
465	SvGETMAGIC (sv);
466	svt = SvTYPE (sv);
467
468	if (svt == SVt_PVHV)	743	else if (svt == SVt_PVHV)
469	encode_hv (enc, (HV *)sv);	744	encode_hv (enc, (HV *)sv);
470	else if (svt == SVt_PVAV)	745	else if (svt == SVt_PVAV)
471	encode_av (enc, (AV *)sv);	746	encode_av (enc, (AV *)sv);
472	else if (svt < SVt_PVAV)	747	else if (svt < SVt_PVAV)
473	{	748	{
474	if (SvNIOK (sv) && SvIV (sv) == 0)	749	STRLEN len = 0;
		750	char *pv = svt ? SvPV (sv, len) : 0;
		751
		752	if (len == 1 && *pv == '1')
		753	encode_str (enc, "true", 4, 0);
		754	else if (len == 1 && *pv == '0')
475	encode_str (enc, "false", 5, 0);	755	encode_str (enc, "false", 5, 0);
476	else if (SvNIOK (sv) && SvIV (sv) == 1)	756	else if (enc->json.flags & F_ALLOW_UNKNOWN)
477	encode_str (enc, "true", 4, 0);	757	encode_str (enc, "null", 4, 0);
478	else	758	else
479	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	759	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
480	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	760	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
481	}	761	}
		762	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		763	encode_str (enc, "null", 4, 0);
482	else	764	else
483	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	765	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
484	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	766	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
485	}	767	}
486		768
…		…
497	encode_str (enc, str, len, SvUTF8 (sv));	779	encode_str (enc, str, len, SvUTF8 (sv));
498	encode_ch (enc, '"');	780	encode_ch (enc, '"');
499	}	781	}
500	else if (SvNOKp (sv))	782	else if (SvNOKp (sv))
501	{	783	{
		784	// trust that perl will do the right thing w.r.t. JSON syntax.
502	need (enc, NV_DIG + 32);	785	need (enc, NV_DIG + 32);
503	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	786	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
504	enc->cur += strlen (enc->cur);	787	enc->cur += strlen (enc->cur);
505	}	788	}
506	else if (SvIOKp (sv))	789	else if (SvIOKp (sv))
507	{	790	{
508	// we assume we can always read an IV as a UV	791	// we assume we can always read an IV as a UV and vice versa
509	if (SvUV (sv) & ~(UV)0x7fff)	792	// we assume two's complement
		793	// we assume no aliasing issues in the union
		794	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		795	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
510	{	796	{
		797	// optimise the "small number case"
		798	// code will likely be branchless and use only a single multiplication
		799	// works for numbers up to 59074
		800	I32 i = SvIVX (sv);
		801	U32 u;
		802	char digit, nz = 0;
		803
511	need (enc, 32);	804	need (enc, 6);
512	enc->cur +=	805
513	SvIsUV(sv)	806	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
514	? snprintf (enc->cur, 32, "%"UVuf, (UV)SvUVX (sv))	807	u = i < 0 ? -i : i;
515	: snprintf (enc->cur, 32, "%"IVdf, (IV)SvIVX (sv));	808
		809	// convert to 4.28 fixed-point representation
		810	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		811
		812	// now output digit by digit, each time masking out the integer part
		813	// and multiplying by 5 while moving the decimal point one to the right,
		814	// resulting in a net multiplication by 10.
		815	// we always write the digit to memory but conditionally increment
		816	// the pointer, to enable the use of conditional move instructions.
		817	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		818	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		819	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		820	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		821	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
516	}	822	}
517	else	823	else
518	{	824	{
519	// optimise the "small number case"	825	// large integer, use the (rather slow) snprintf way.
520	// code will likely be branchless and use only a single multiplication
521	I32 i = SvIV (sv);
522	U32 u;
523
524	need (enc, 6);	826	need (enc, IVUV_MAXCHARS);
525		827	enc->cur +=
526	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;	828	SvIsUV(sv)
527	u = i < 0 ? -i : i;	829	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
528		830	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
529	// convert to 4.28 fixed-point representation
530	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
531
532	char digit, nz = 0;
533
534	digit = u >> 28; *enc->cur = digit + '0'; nz |= digit; enc->cur += nz ? 1 : 0; u = (u & 0xfffffff) * 5;
535	digit = u >> 27; *enc->cur = digit + '0'; nz |= digit; enc->cur += nz ? 1 : 0; u = (u & 0x7ffffff) * 5;
536	digit = u >> 26; *enc->cur = digit + '0'; nz |= digit; enc->cur += nz ? 1 : 0; u = (u & 0x3ffffff) * 5;
537	digit = u >> 25; *enc->cur = digit + '0'; nz |= digit; enc->cur += nz ? 1 : 0; u = (u & 0x1ffffff) * 5;
538	digit = u >> 24; *enc->cur = digit + '0'; nz |= digit; enc->cur += 1;
539	}	831	}
540	}	832	}
541	else if (SvROK (sv))	833	else if (SvROK (sv))
542	encode_rv (enc, SvRV (sv));	834	encode_rv (enc, SvRV (sv));
543	else if (!SvOK (sv))	835	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
544	encode_str (enc, "null", 4, 0);	836	encode_str (enc, "null", 4, 0);
545	else	837	else
546	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	838	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
547	SvPV_nolen (sv), SvFLAGS (sv));	839	SvPV_nolen (sv), SvFLAGS (sv));
548	}	840	}
549		841
550	static SV *	842	static SV *
551	encode_json (SV *scalar, U32 flags)	843	encode_json (SV *scalar, JSON *json)
552	{	844	{
553	enc_t enc;	845	enc_t enc;
554		846
555	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	847	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
556	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	848	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
557		849
558	enc.flags = flags;	850	enc.json = *json;
559	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	851	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
560	enc.cur = SvPVX (enc.sv);	852	enc.cur = SvPVX (enc.sv);
561	enc.end = SvEND (enc.sv);	853	enc.end = SvEND (enc.sv);
562	enc.indent = 0;	854	enc.indent = 0;
563	enc.maxdepth = DEC_DEPTH (flags);	855	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		856	: enc.json.flags & F_LATIN1 ? 0x000100UL
		857	: 0x110000UL;
564		858
565	SvPOK_only (enc.sv);	859	SvPOK_only (enc.sv);
566	encode_sv (&enc, scalar);	860	encode_sv (&enc, scalar);
		861	encode_nl (&enc);
567		862
568	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	863	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
569	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	864	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
570		865
571	if (!(flags & (F_ASCII | F_LATIN1 | F_UTF8)))	866	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
572	SvUTF8_on (enc.sv);	867	SvUTF8_on (enc.sv);
573		868
574	if (enc.flags & F_SHRINK)	869	if (enc.json.flags & F_SHRINK)
575	shrink (enc.sv);	870	shrink (enc.sv);
576		871
577	return enc.sv;	872	return enc.sv;
578	}	873	}
579		874
…		…
584	typedef struct	879	typedef struct
585	{	880	{
586	char *cur; // current parser pointer	881	char *cur; // current parser pointer
587	char *end; // end of input string	882	char *end; // end of input string
588	const char *err; // parse error, if != 0	883	const char *err; // parse error, if != 0
589	U32 flags; // F_*	884	JSON json;
590	U32 depth; // recursion depth	885	U32 depth; // recursion depth
591	U32 maxdepth; // recursion depth limit	886	U32 maxdepth; // recursion depth limit
592	} dec_t;	887	} dec_t;
593		888
594	inline void	889	INLINE void
		890	decode_comment (dec_t *dec)
		891	{
		892	// only '#'-style comments allowed a.t.m.
		893
		894	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		895	++dec->cur;
		896	}
		897
		898	INLINE void
595	decode_ws (dec_t *dec)	899	decode_ws (dec_t *dec)
596	{	900	{
597	for (;;)	901	for (;;)
598	{	902	{
599	char ch = *dec->cur;	903	char ch = *dec->cur;
600		904
601	if (ch > 0x20	905	if (ch > 0x20)
		906	{
		907	if (expect_false (ch == '#'))
		908	{
		909	if (dec->json.flags & F_RELAXED)
		910	decode_comment (dec);
		911	else
		912	break;
		913	}
		914	else
		915	break;
		916	}
602	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	917	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
603	break;	918	break; // parse error, but let higher level handle it, gives better error messages
604		919
605	++dec->cur;	920	++dec->cur;
606	}	921	}
607	}	922	}
608		923
…		…
612	if (*dec->cur != ch) \	927	if (*dec->cur != ch) \
613	ERR (# ch " expected"); \	928	ERR (# ch " expected"); \
614	++dec->cur; \	929	++dec->cur; \
615	SE	930	SE
616		931
617	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	932	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
618	#define DEC_DEC_DEPTH --dec->depth	933	#define DEC_DEC_DEPTH --dec->depth
619		934
620	static SV *decode_sv (dec_t *dec);	935	static SV *decode_sv (dec_t *dec);
621		936
622	static signed char decode_hexdigit[256];	937	static signed char decode_hexdigit[256];
…		…
625	decode_4hex (dec_t *dec)	940	decode_4hex (dec_t *dec)
626	{	941	{
627	signed char d1, d2, d3, d4;	942	signed char d1, d2, d3, d4;
628	unsigned char *cur = (unsigned char *)dec->cur;	943	unsigned char *cur = (unsigned char *)dec->cur;
629		944
630	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("four hexadecimal digits expected");	945	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
631	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("four hexadecimal digits expected");	946	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
632	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("four hexadecimal digits expected");	947	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
633	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("four hexadecimal digits expected");	948	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
634		949
635	dec->cur += 4;	950	dec->cur += 4;
636		951
637	return ((UV)d1) << 12	952	return ((UV)d1) << 12
638	| ((UV)d2) << 8	953	| ((UV)d2) << 8
…		…
646	static SV *	961	static SV *
647	decode_str (dec_t *dec)	962	decode_str (dec_t *dec)
648	{	963	{
649	SV *sv = 0;	964	SV *sv = 0;
650	int utf8 = 0;	965	int utf8 = 0;
		966	char *dec_cur = dec->cur;
651		967
652	do	968	do
653	{	969	{
654	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];	970	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
655	char *cur = buf;	971	char *cur = buf;
656		972
657	do	973	do
658	{	974	{
659	unsigned char ch = *(unsigned char *)dec->cur++;	975	unsigned char ch = *(unsigned char *)dec_cur++;
660		976
661	if (expect_false (ch == '"'))	977	if (expect_false (ch == '"'))
662	{	978	{
663	--dec->cur;	979	--dec_cur;
664	break;	980	break;
665	}	981	}
666	else if (expect_false (ch == '\\'))	982	else if (expect_false (ch == '\\'))
667	{	983	{
668	switch (*dec->cur)	984	switch (*dec_cur)
669	{	985	{
670	case '\\':	986	case '\\':
671	case '/':	987	case '/':
672	case '"': *cur++ = *dec->cur++; break;	988	case '"': *cur++ = *dec_cur++; break;
673		989
674	case 'b': ++dec->cur; *cur++ = '\010'; break;	990	case 'b': ++dec_cur; *cur++ = '\010'; break;
675	case 't': ++dec->cur; *cur++ = '\011'; break;	991	case 't': ++dec_cur; *cur++ = '\011'; break;
676	case 'n': ++dec->cur; *cur++ = '\012'; break;	992	case 'n': ++dec_cur; *cur++ = '\012'; break;
677	case 'f': ++dec->cur; *cur++ = '\014'; break;	993	case 'f': ++dec_cur; *cur++ = '\014'; break;
678	case 'r': ++dec->cur; *cur++ = '\015'; break;	994	case 'r': ++dec_cur; *cur++ = '\015'; break;
679		995
680	case 'u':	996	case 'u':
681	{	997	{
682	UV lo, hi;	998	UV lo, hi;
683	++dec->cur;	999	++dec_cur;
684		1000
		1001	dec->cur = dec_cur;
685	hi = decode_4hex (dec);	1002	hi = decode_4hex (dec);
		1003	dec_cur = dec->cur;
686	if (hi == (UV)-1)	1004	if (hi == (UV)-1)
687	goto fail;	1005	goto fail;
688		1006
689	// possibly a surrogate pair	1007	// possibly a surrogate pair
690	if (hi >= 0xd800)	1008	if (hi >= 0xd800)
691	if (hi < 0xdc00)	1009	if (hi < 0xdc00)
692	{	1010	{
693	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	1011	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
694	ERR ("missing low surrogate character in surrogate pair");	1012	ERR ("missing low surrogate character in surrogate pair");
695		1013
696	dec->cur += 2;	1014	dec_cur += 2;
697		1015
		1016	dec->cur = dec_cur;
698	lo = decode_4hex (dec);	1017	lo = decode_4hex (dec);
		1018	dec_cur = dec->cur;
699	if (lo == (UV)-1)	1019	if (lo == (UV)-1)
700	goto fail;	1020	goto fail;
701		1021
702	if (lo < 0xdc00 || lo >= 0xe000)	1022	if (lo < 0xdc00 || lo >= 0xe000)
703	ERR ("surrogate pair expected");	1023	ERR ("surrogate pair expected");
…		…
709		1029
710	if (hi >= 0x80)	1030	if (hi >= 0x80)
711	{	1031	{
712	utf8 = 1;	1032	utf8 = 1;
713		1033
714	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	1034	cur = encode_utf8 (cur, hi);
715	}	1035	}
716	else	1036	else
717	*cur++ = hi;	1037	*cur++ = hi;
718	}	1038	}
719	break;	1039	break;
720		1040
721	default:	1041	default:
722	--dec->cur;	1042	--dec_cur;
723	ERR ("illegal backslash escape sequence in string");	1043	ERR ("illegal backslash escape sequence in string");
724	}	1044	}
725	}	1045	}
726	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	1046	else if (expect_true (ch >= 0x20 && ch < 0x80))
727	*cur++ = ch;	1047	*cur++ = ch;
728	else if (ch >= 0x80)	1048	else if (ch >= 0x80)
729	{	1049	{
730	STRLEN clen;	1050	STRLEN clen;
731	UV uch;
732		1051
733	--dec->cur;	1052	--dec_cur;
734		1053
735	uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	1054	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
736	if (clen == (STRLEN)-1)	1055	if (clen == (STRLEN)-1)
737	ERR ("malformed UTF-8 character in JSON string");	1056	ERR ("malformed UTF-8 character in JSON string");
738		1057
739	do	1058	do
740	*cur++ = *dec->cur++;	1059	*cur++ = *dec_cur++;
741	while (--clen);	1060	while (--clen);
742		1061
743	utf8 = 1;	1062	utf8 = 1;
744	}	1063	}
745	else	1064	else
746	{	1065	{
747	--dec->cur;	1066	--dec_cur;
748		1067
749	if (!ch)	1068	if (!ch)
750	ERR ("unexpected end of string while parsing JSON string");	1069	ERR ("unexpected end of string while parsing JSON string");
751	else	1070	else
752	ERR ("invalid character encountered while parsing JSON string");	1071	ERR ("invalid character encountered while parsing JSON string");
…		…
757	{	1076	{
758	STRLEN len = cur - buf;	1077	STRLEN len = cur - buf;
759		1078
760	if (sv)	1079	if (sv)
761	{	1080	{
762	SvGROW (sv, SvCUR (sv) + len + 1);	1081	STRLEN cur = SvCUR (sv);
		1082
		1083	if (SvLEN (sv) <= cur + len)
		1084	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1085
763	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	1086	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
764	SvCUR_set (sv, SvCUR (sv) + len);	1087	SvCUR_set (sv, SvCUR (sv) + len);
765	}	1088	}
766	else	1089	else
767	sv = newSVpvn (buf, len);	1090	sv = newSVpvn (buf, len);
768	}	1091	}
769	}	1092	}
770	while (*dec->cur != '"');	1093	while (*dec_cur != '"');
771		1094
772	++dec->cur;	1095	++dec_cur;
773		1096
774	if (sv)	1097	if (sv)
775	{	1098	{
776	SvPOK_only (sv);	1099	SvPOK_only (sv);
777	*SvEND (sv) = 0;	1100	*SvEND (sv) = 0;
…		…
780	SvUTF8_on (sv);	1103	SvUTF8_on (sv);
781	}	1104	}
782	else	1105	else
783	sv = newSVpvn ("", 0);	1106	sv = newSVpvn ("", 0);
784		1107
		1108	dec->cur = dec_cur;
785	return sv;	1109	return sv;
786		1110
787	fail:	1111	fail:
		1112	dec->cur = dec_cur;
788	return 0;	1113	return 0;
789	}	1114	}
790		1115
791	static SV *	1116	static SV *
792	decode_num (dec_t *dec)	1117	decode_num (dec_t *dec)
…		…
850	is_nv = 1;	1175	is_nv = 1;
851	}	1176	}
852		1177
853	if (!is_nv)	1178	if (!is_nv)
854	{	1179	{
		1180	int len = dec->cur - start;
		1181
855	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1182	// special case the rather common 1..5-digit-int case
856	if (*start == '-')	1183	if (*start == '-')
857	switch (dec->cur - start)	1184	switch (len)
858	{	1185	{
859	case 2: return newSViv (-( start [1] - '0' ));	1186	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
860	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1187	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
861	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));	1188	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
862	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));	1189	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1190	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
863	}	1191	}
864	else	1192	else
865	switch (dec->cur - start)	1193	switch (len)
866	{	1194	{
867	case 1: return newSViv ( start [0] - '0' );	1195	case 1: return newSViv ( start [0] - '0' * 1);
868	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1196	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
869	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);	1197	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
870	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);	1198	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1199	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
871	}	1200	}
872		1201
873	{	1202	{
874	UV uv;	1203	UV uv;
875	int numtype = grok_number (start, dec->cur - start, &uv);	1204	int numtype = grok_number (start, len, &uv);
876	if (numtype & IS_NUMBER_IN_UV)	1205	if (numtype & IS_NUMBER_IN_UV)
877	if (numtype & IS_NUMBER_NEG)	1206	if (numtype & IS_NUMBER_NEG)
878	{	1207	{
879	if (uv < (UV)IV_MIN)	1208	if (uv < (UV)IV_MIN)
880	return newSViv (-(IV)uv);	1209	return newSViv (-(IV)uv);
881	}	1210	}
882	else	1211	else
883	return newSVuv (uv);	1212	return newSVuv (uv);
884	}	1213	}
885	}
886		1214
		1215	len -= *start == '-' ? 1 : 0;
		1216
		1217	// does not fit into IV or UV, try NV
		1218	if (len <= NV_DIG)
		1219	// fits into NV without loss of precision
		1220	return newSVnv (json_atof (start));
		1221
		1222	// everything else fails, convert it to a string
		1223	return newSVpvn (start, dec->cur - start);
		1224	}
		1225
		1226	// loss of precision here
887	return newSVnv (Atof (start));	1227	return newSVnv (json_atof (start));
888		1228
889	fail:	1229	fail:
890	return 0;	1230	return 0;
891	}	1231	}
892		1232
…		…
921		1261
922	if (*dec->cur != ',')	1262	if (*dec->cur != ',')
923	ERR (", or ] expected while parsing array");	1263	ERR (", or ] expected while parsing array");
924		1264
925	++dec->cur;	1265	++dec->cur;
		1266
		1267	decode_ws (dec);
		1268
		1269	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1270	{
		1271	++dec->cur;
		1272	break;
		1273	}
926	}	1274	}
927		1275
928	DEC_DEC_DEPTH;	1276	DEC_DEC_DEPTH;
929	return newRV_noinc ((SV *)av);	1277	return newRV_noinc ((SV *)av);
930		1278
…		…
935	}	1283	}
936		1284
937	static SV *	1285	static SV *
938	decode_hv (dec_t *dec)	1286	decode_hv (dec_t *dec)
939	{	1287	{
		1288	SV *sv;
940	HV *hv = newHV ();	1289	HV *hv = newHV ();
941		1290
942	DEC_INC_DEPTH;	1291	DEC_INC_DEPTH;
943	decode_ws (dec);	1292	decode_ws (dec);
944		1293
945	if (*dec->cur == '}')	1294	if (*dec->cur == '}')
946	++dec->cur;	1295	++dec->cur;
947	else	1296	else
948	for (;;)	1297	for (;;)
949	{	1298	{
		1299	EXPECT_CH ('"');
		1300
		1301	// heuristic: assume that
		1302	// a) decode_str + hv_store_ent are abysmally slow.
		1303	// b) most hash keys are short, simple ascii text.
		1304	// => try to "fast-match" such strings to avoid
		1305	// the overhead of decode_str + hv_store_ent.
		1306	{
950	SV *key, *value;	1307	SV *value;
		1308	char *p = dec->cur;
		1309	char *e = p + 24; // only try up to 24 bytes
951		1310
952	decode_ws (dec); EXPECT_CH ('"');	1311	for (;;)
953
954	key = decode_str (dec);
955	if (!key)
956	goto fail;
957
958	decode_ws (dec); EXPECT_CH (':');
959
960	value = decode_sv (dec);
961	if (!value)
962	{	1312	{
		1313	// the >= 0x80 is false on most architectures
		1314	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1315	{
		1316	// slow path, back up and use decode_str
		1317	SV *key = decode_str (dec);
		1318	if (!key)
		1319	goto fail;
		1320
		1321	decode_ws (dec); EXPECT_CH (':');
		1322
		1323	decode_ws (dec);
		1324	value = decode_sv (dec);
		1325	if (!value)
		1326	{
		1327	SvREFCNT_dec (key);
		1328	goto fail;
		1329	}
		1330
		1331	hv_store_ent (hv, key, value, 0);
963	SvREFCNT_dec (key);	1332	SvREFCNT_dec (key);
		1333
		1334	break;
		1335	}
		1336	else if (*p == '"')
		1337	{
		1338	// fast path, got a simple key
		1339	char *key = dec->cur;
		1340	int len = p - key;
		1341	dec->cur = p + 1;
		1342
		1343	decode_ws (dec); EXPECT_CH (':');
		1344
		1345	decode_ws (dec);
		1346	value = decode_sv (dec);
		1347	if (!value)
964	goto fail;	1348	goto fail;
		1349
		1350	hv_store (hv, key, len, value, 0);
		1351
		1352	break;
		1353	}
		1354
		1355	++p;
965	}	1356	}
966		1357	}
967	hv_store_ent (hv, key, value, 0);
968	SvREFCNT_dec (key);
969		1358
970	decode_ws (dec);	1359	decode_ws (dec);
971		1360
972	if (*dec->cur == '}')	1361	if (*dec->cur == '}')
973	{	1362	{
…		…
977		1366
978	if (*dec->cur != ',')	1367	if (*dec->cur != ',')
979	ERR (", or } expected while parsing object/hash");	1368	ERR (", or } expected while parsing object/hash");
980		1369
981	++dec->cur;	1370	++dec->cur;
		1371
		1372	decode_ws (dec);
		1373
		1374	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1375	{
		1376	++dec->cur;
		1377	break;
		1378	}
982	}	1379	}
983		1380
984	DEC_DEC_DEPTH;	1381	DEC_DEC_DEPTH;
985	return newRV_noinc ((SV *)hv);	1382	sv = newRV_noinc ((SV *)hv);
		1383
		1384	// check filter callbacks
		1385	if (dec->json.flags & F_HOOK)
		1386	{
		1387	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1388	{
		1389	HE *cb, *he;
		1390
		1391	hv_iterinit (hv);
		1392	he = hv_iternext (hv);
		1393	hv_iterinit (hv);
		1394
		1395	// the next line creates a mortal sv each time its called.
		1396	// might want to optimise this for common cases.
		1397	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1398
		1399	if (cb)
		1400	{
		1401	dSP;
		1402	int count;
		1403
		1404	ENTER; SAVETMPS; PUSHMARK (SP);
		1405	XPUSHs (HeVAL (he));
		1406	sv_2mortal (sv);
		1407
		1408	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1409
		1410	if (count == 1)
		1411	{
		1412	sv = newSVsv (POPs);
		1413	FREETMPS; LEAVE;
		1414	return sv;
		1415	}
		1416
		1417	SvREFCNT_inc (sv);
		1418	FREETMPS; LEAVE;
		1419	}
		1420	}
		1421
		1422	if (dec->json.cb_object)
		1423	{
		1424	dSP;
		1425	int count;
		1426
		1427	ENTER; SAVETMPS; PUSHMARK (SP);
		1428	XPUSHs (sv_2mortal (sv));
		1429
		1430	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1431
		1432	if (count == 1)
		1433	{
		1434	sv = newSVsv (POPs);
		1435	FREETMPS; LEAVE;
		1436	return sv;
		1437	}
		1438
		1439	SvREFCNT_inc (sv);
		1440	FREETMPS; LEAVE;
		1441	}
		1442	}
		1443
		1444	return sv;
986		1445
987	fail:	1446	fail:
988	SvREFCNT_dec (hv);	1447	SvREFCNT_dec (hv);
989	DEC_DEC_DEPTH;	1448	DEC_DEC_DEPTH;
990	return 0;	1449	return 0;
991	}	1450	}
992		1451
993	static SV *	1452	static SV *
994	decode_sv (dec_t *dec)	1453	decode_sv (dec_t *dec)
995	{	1454	{
996	decode_ws (dec);	1455	// the beauty of JSON: you need exactly one character lookahead
		1456	// to parse everything.
997	switch (*dec->cur)	1457	switch (*dec->cur)
998	{	1458	{
999	case '"': ++dec->cur; return decode_str (dec);	1459	case '"': ++dec->cur; return decode_str (dec);
1000	case '[': ++dec->cur; return decode_av (dec);	1460	case '[': ++dec->cur; return decode_av (dec);
1001	case '{': ++dec->cur; return decode_hv (dec);	1461	case '{': ++dec->cur; return decode_hv (dec);
1002		1462
1003	case '-':	1463	case '-':
1004	case '0': case '1': case '2': case '3': case '4':	1464	case '0': case '1': case '2': case '3': case '4':
1005	case '5': case '6': case '7': case '8': case '9':	1465	case '5': case '6': case '7': case '8': case '9':
1006	return decode_num (dec);	1466	return decode_num (dec);
1007		1467
1008	case 't':	1468	case 't':
1009	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1469	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1010	{	1470	{
1011	dec->cur += 4;	1471	dec->cur += 4;
		1472	#if JSON_SLOW
		1473	json_true = get_bool ("JSON::XS::true");
		1474	#endif
1012	return newSViv (1);	1475	return newSVsv (json_true);
1013	}	1476	}
1014	else	1477	else
1015	ERR ("'true' expected");	1478	ERR ("'true' expected");
1016		1479
1017	break;	1480	break;
1018		1481
1019	case 'f':	1482	case 'f':
1020	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1483	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1021	{	1484	{
1022	dec->cur += 5;	1485	dec->cur += 5;
		1486	#if JSON_SLOW
		1487	json_false = get_bool ("JSON::XS::false");
		1488	#endif
1023	return newSViv (0);	1489	return newSVsv (json_false);
1024	}	1490	}
1025	else	1491	else
1026	ERR ("'false' expected");	1492	ERR ("'false' expected");
1027		1493
1028	break;	1494	break;
…		…
1046	fail:	1512	fail:
1047	return 0;	1513	return 0;
1048	}	1514	}
1049		1515
1050	static SV *	1516	static SV *
1051	decode_json (SV *string, U32 flags, UV *offset_return)	1517	decode_json (SV *string, JSON *json, char **offset_return)
1052	{	1518	{
1053	dec_t dec;	1519	dec_t dec;
1054	UV offset;
1055	SV *sv;	1520	SV *sv;
1056		1521
		1522	/* work around bugs in 5.10 where manipulating magic values
		1523	* will perl ignore the magic in subsequent accesses
		1524	*/
1057	SvGETMAGIC (string);	1525	/*SvGETMAGIC (string);*/
		1526	if (SvMAGICAL (string))
		1527	string = sv_2mortal (newSVsv (string));
		1528
1058	SvUPGRADE (string, SVt_PV);	1529	SvUPGRADE (string, SVt_PV);
1059		1530
		1531	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1532	* assertion with -DDEBUGGING, although SvCUR is documented to
		1533	* return the xpv_cur field which certainly exists after upgrading.
		1534	* according to nicholas clark, calling SvPOK fixes this.
		1535	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1536	* and hope for the best.
		1537	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1538	* assertion business is seriously broken, try yet another workaround
		1539	* for the broken -DDEBUGGING.
		1540	*/
		1541	{
		1542	#ifdef DEBUGGING
		1543	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1544	#else
		1545	STRLEN offset = SvCUR (string);
		1546	#endif
		1547
		1548	if (offset > json->max_size && json->max_size)
		1549	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1550	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1551	}
		1552
1060	if (flags & F_UTF8)	1553	if (json->flags & F_UTF8)
1061	sv_utf8_downgrade (string, 0);	1554	sv_utf8_downgrade (string, 0);
1062	else	1555	else
1063	sv_utf8_upgrade (string);	1556	sv_utf8_upgrade (string);
1064		1557
1065	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1558	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1066		1559
1067	dec.flags = flags;	1560	dec.json = *json;
1068	dec.cur = SvPVX (string);	1561	dec.cur = SvPVX (string);
1069	dec.end = SvEND (string);	1562	dec.end = SvEND (string);
1070	dec.err = 0;	1563	dec.err = 0;
1071	dec.depth = 0;	1564	dec.depth = 0;
1072	dec.maxdepth = DEC_DEPTH (dec.flags);	1565
		1566	if (dec.json.cb_object || dec.json.cb_sk_object)
		1567	dec.json.flags |= F_HOOK;
1073		1568
1074	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1569	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1570
		1571	decode_ws (&dec);
1075	sv = decode_sv (&dec);	1572	sv = decode_sv (&dec);
		1573
		1574	if (offset_return)
		1575	*offset_return = dec.cur;
1076		1576
1077	if (!(offset_return || !sv))	1577	if (!(offset_return || !sv))
1078	{	1578	{
1079	// check for trailing garbage	1579	// check for trailing garbage
1080	decode_ws (&dec);	1580	decode_ws (&dec);
…		…
1083	{	1583	{
1084	dec.err = "garbage after JSON object";	1584	dec.err = "garbage after JSON object";
1085	SvREFCNT_dec (sv);	1585	SvREFCNT_dec (sv);
1086	sv = 0;	1586	sv = 0;
1087	}	1587	}
1088	}
1089
1090	if (offset_return || !sv)
1091	{
1092	offset = dec.flags & F_UTF8
1093	? dec.cur - SvPVX (string)
1094	: utf8_distance (dec.cur, SvPVX (string));
1095
1096	if (offset_return)
1097	*offset_return = offset;
1098	}	1588	}
1099		1589
1100	if (!sv)	1590	if (!sv)
1101	{	1591	{
1102	SV *uni = sv_newmortal ();	1592	SV *uni = sv_newmortal ();
…		…
1108	SAVEVPTR (PL_curcop);	1598	SAVEVPTR (PL_curcop);
1109	PL_curcop = &cop;	1599	PL_curcop = &cop;
1110	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1600	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1111	LEAVE;	1601	LEAVE;
1112		1602
1113	croak ("%s, at character offset %d [\"%s\"]",	1603	croak ("%s, at character offset %d (before \"%s\")",
1114	dec.err,	1604	dec.err,
1115	(int)offset,	1605	ptr_to_index (string, dec.cur),
1116	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1606	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1117	}	1607	}
1118		1608
1119	sv = sv_2mortal (sv);	1609	sv = sv_2mortal (sv);
1120		1610
1121	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1611	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1122	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1612	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1123		1613
1124	return sv;	1614	return sv;
		1615	}
		1616
		1617	/////////////////////////////////////////////////////////////////////////////
		1618	// incremental parser
		1619
		1620	static void
		1621	incr_parse (JSON *self)
		1622	{
		1623	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1624
		1625	// the state machine here is a bit convoluted and could be simplified a lot
		1626	// but this would make it slower, so...
		1627
		1628	for (;;)
		1629	{
		1630	//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
		1631	switch (self->incr_mode)
		1632	{
		1633	// only used for initial whitespace skipping
		1634	case INCR_M_WS:
		1635	for (;;)
		1636	{
		1637	if (*p > 0x20)
		1638	{
		1639	if (*p == '#')
		1640	{
		1641	self->incr_mode = INCR_M_C0;
		1642	goto incr_m_c;
		1643	}
		1644	else
		1645	{
		1646	self->incr_mode = INCR_M_JSON;
		1647	goto incr_m_json;
		1648	}
		1649	}
		1650	else if (!*p)
		1651	goto interrupt;
		1652
		1653	++p;
		1654	}
		1655
		1656	// skip a single char inside a string (for \\-processing)
		1657	case INCR_M_BS:
		1658	if (!*p)
		1659	goto interrupt;
		1660
		1661	++p;
		1662	self->incr_mode = INCR_M_STR;
		1663	goto incr_m_str;
		1664
		1665	// inside #-style comments
		1666	case INCR_M_C0:
		1667	case INCR_M_C1:
		1668	incr_m_c:
		1669	for (;;)
		1670	{
		1671	if (*p == '\n')
		1672	{
		1673	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1674	break;
		1675	}
		1676	else if (!*p)
		1677	goto interrupt;
		1678
		1679	++p;
		1680	}
		1681
		1682	break;
		1683
		1684	// inside a string
		1685	case INCR_M_STR:
		1686	incr_m_str:
		1687	for (;;)
		1688	{
		1689	if (*p == '"')
		1690	{
		1691	++p;
		1692	self->incr_mode = INCR_M_JSON;
		1693
		1694	if (!self->incr_nest)
		1695	goto interrupt;
		1696
		1697	goto incr_m_json;
		1698	}
		1699	else if (*p == '\\')
		1700	{
		1701	++p; // "virtually" consumes character after \
		1702
		1703	if (!*p) // if at end of string we have to switch modes
		1704	{
		1705	self->incr_mode = INCR_M_BS;
		1706	goto interrupt;
		1707	}
		1708	}
		1709	else if (!*p)
		1710	goto interrupt;
		1711
		1712	++p;
		1713	}
		1714
		1715	// after initial ws, outside string
		1716	case INCR_M_JSON:
		1717	incr_m_json:
		1718	for (;;)
		1719	{
		1720	switch (*p++)
		1721	{
		1722	case 0:
		1723	--p;
		1724	goto interrupt;
		1725
		1726	case 0x09:
		1727	case 0x0a:
		1728	case 0x0d:
		1729	case 0x20:
		1730	if (!self->incr_nest)
		1731	{
		1732	--p; // do not eat the whitespace, let the next round do it
		1733	goto interrupt;
		1734	}
		1735	break;
		1736
		1737	case '"':
		1738	self->incr_mode = INCR_M_STR;
		1739	goto incr_m_str;
		1740
		1741	case '[':
		1742	case '{':
		1743	if (++self->incr_nest > self->max_depth)
		1744	croak (ERR_NESTING_EXCEEDED);
		1745	break;
		1746
		1747	case ']':
		1748	case '}':
		1749	if (--self->incr_nest <= 0)
		1750	goto interrupt;
		1751	break;
		1752
		1753	case '#':
		1754	self->incr_mode = INCR_M_C1;
		1755	goto incr_m_c;
		1756	}
		1757	}
		1758	}
		1759
		1760	modechange:
		1761	;
		1762	}
		1763
		1764	interrupt:
		1765	self->incr_pos = p - SvPVX (self->incr_text);
		1766	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1767	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1125	}	1768	}
1126		1769
1127	/////////////////////////////////////////////////////////////////////////////	1770	/////////////////////////////////////////////////////////////////////////////
1128	// XS interface functions	1771	// XS interface functions
1129		1772
1130	MODULE = JSON::XS PACKAGE = JSON::XS	1773	MODULE = JSON::XS PACKAGE = JSON::XS
1131		1774
1132	BOOT:	1775	BOOT:
1133	{	1776	{
1134	int i;	1777	int i;
1135
1136	memset (decode_hexdigit, 0xff, 256);
1137		1778
1138	for (i = 0; i < 256; ++i)	1779	for (i = 0; i < 256; ++i)
1139	decode_hexdigit [i] =	1780	decode_hexdigit [i] =
1140	i >= '0' && i <= '9' ? i - '0'	1781	i >= '0' && i <= '9' ? i - '0'
1141	: i >= 'a' && i <= 'f' ? i - 'a' + 10	1782	: i >= 'a' && i <= 'f' ? i - 'a' + 10
1142	: i >= 'A' && i <= 'F' ? i - 'A' + 10	1783	: i >= 'A' && i <= 'F' ? i - 'A' + 10
1143	: -1;	1784	: -1;
1144		1785
1145	json_stash = gv_stashpv ("JSON::XS", 1);	1786	json_stash = gv_stashpv ("JSON::XS" , 1);
		1787	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1788
		1789	json_true = get_bool ("JSON::XS::true");
		1790	json_false = get_bool ("JSON::XS::false");
		1791
		1792	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1146	}	1793	}
1147		1794
1148	PROTOTYPES: DISABLE	1795	PROTOTYPES: DISABLE
1149		1796
1150	SV *new (char *dummy)	1797	void CLONE (...)
1151	CODE:	1798	CODE:
1152	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1799	json_stash = 0;
		1800	json_boolean_stash = 0;
		1801
		1802	void new (char *klass)
		1803	PPCODE:
		1804	{
		1805	SV *pv = NEWSV (0, sizeof (JSON));
		1806	SvPOK_only (pv);
		1807	json_init ((JSON *)SvPVX (pv));
		1808	XPUSHs (sv_2mortal (sv_bless (
		1809	newRV_noinc (pv),
		1810	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1811	)));
		1812	}
		1813
		1814	void ascii (JSON *self, int enable = 1)
		1815	ALIAS:
		1816	ascii = F_ASCII
		1817	latin1 = F_LATIN1
		1818	utf8 = F_UTF8
		1819	indent = F_INDENT
		1820	canonical = F_CANONICAL
		1821	space_before = F_SPACE_BEFORE
		1822	space_after = F_SPACE_AFTER
		1823	pretty = F_PRETTY
		1824	allow_nonref = F_ALLOW_NONREF
		1825	shrink = F_SHRINK
		1826	allow_blessed = F_ALLOW_BLESSED
		1827	convert_blessed = F_CONV_BLESSED
		1828	relaxed = F_RELAXED
		1829	allow_unknown = F_ALLOW_UNKNOWN
		1830	PPCODE:
		1831	{
		1832	if (enable)
		1833	self->flags |= ix;
		1834	else
		1835	self->flags &= ~ix;
		1836
		1837	XPUSHs (ST (0));
		1838	}
		1839
		1840	void get_ascii (JSON *self)
		1841	ALIAS:
		1842	get_ascii = F_ASCII
		1843	get_latin1 = F_LATIN1
		1844	get_utf8 = F_UTF8
		1845	get_indent = F_INDENT
		1846	get_canonical = F_CANONICAL
		1847	get_space_before = F_SPACE_BEFORE
		1848	get_space_after = F_SPACE_AFTER
		1849	get_allow_nonref = F_ALLOW_NONREF
		1850	get_shrink = F_SHRINK
		1851	get_allow_blessed = F_ALLOW_BLESSED
		1852	get_convert_blessed = F_CONV_BLESSED
		1853	get_relaxed = F_RELAXED
		1854	get_allow_unknown = F_ALLOW_UNKNOWN
		1855	PPCODE:
		1856	XPUSHs (boolSV (self->flags & ix));
		1857
		1858	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		1859	PPCODE:
		1860	self->max_depth = max_depth;
		1861	XPUSHs (ST (0));
		1862
		1863	U32 get_max_depth (JSON *self)
		1864	CODE:
		1865	RETVAL = self->max_depth;
1153	OUTPUT:	1866	OUTPUT:
1154	RETVAL	1867	RETVAL
1155		1868
1156	SV *ascii (SV *self, int enable = 1)	1869	void max_size (JSON *self, U32 max_size = 0)
1157	ALIAS:	1870	PPCODE:
1158	ascii = F_ASCII	1871	self->max_size = max_size;
1159	latin1 = F_LATIN1	1872	XPUSHs (ST (0));
1160	utf8 = F_UTF8	1873
1161	indent = F_INDENT	1874	int get_max_size (JSON *self)
1162	canonical = F_CANONICAL
1163	space_before = F_SPACE_BEFORE
1164	space_after = F_SPACE_AFTER
1165	pretty = F_PRETTY
1166	allow_nonref = F_ALLOW_NONREF
1167	shrink = F_SHRINK
1168	CODE:	1875	CODE:
1169	{	1876	RETVAL = self->max_size;
1170	UV *uv = SvJSON (self);
1171	if (enable)
1172	*uv |= ix;
1173	else
1174	*uv &= ~ix;
1175
1176	RETVAL = newSVsv (self);
1177	}
1178	OUTPUT:	1877	OUTPUT:
1179	RETVAL	1878	RETVAL
1180		1879
1181	SV *max_depth (SV *self, UV max_depth = 0x80000000UL)	1880	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
		1881	PPCODE:
		1882	{
		1883	SvREFCNT_dec (self->cb_object);
		1884	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1885
		1886	XPUSHs (ST (0));
		1887	}
		1888
		1889	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
		1890	PPCODE:
		1891	{
		1892	if (!self->cb_sk_object)
		1893	self->cb_sk_object = newHV ();
		1894
		1895	if (SvOK (cb))
		1896	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1897	else
		1898	{
		1899	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1900
		1901	if (!HvKEYS (self->cb_sk_object))
		1902	{
		1903	SvREFCNT_dec (self->cb_sk_object);
		1904	self->cb_sk_object = 0;
		1905	}
		1906	}
		1907
		1908	XPUSHs (ST (0));
		1909	}
		1910
		1911	void encode (JSON *self, SV *scalar)
		1912	PPCODE:
		1913	XPUSHs (encode_json (scalar, self));
		1914
		1915	void decode (JSON *self, SV *jsonstr)
		1916	PPCODE:
		1917	XPUSHs (decode_json (jsonstr, self, 0));
		1918
		1919	void decode_prefix (JSON *self, SV *jsonstr)
		1920	PPCODE:
		1921	{
		1922	char *offset;
		1923	EXTEND (SP, 2);
		1924	PUSHs (decode_json (jsonstr, self, &offset));
		1925	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		1926	}
		1927
		1928	void incr_parse (JSON *self, SV *jsonstr = 0)
		1929	PPCODE:
		1930	{
		1931	if (!self->incr_text)
		1932	self->incr_text = newSVpvn ("", 0);
		1933
		1934	// append data, if any
		1935	if (jsonstr)
		1936	{
		1937	if (SvUTF8 (jsonstr))
		1938	{
		1939	if (!SvUTF8 (self->incr_text))
		1940	{
		1941	/* utf-8-ness differs, need to upgrade */
		1942	sv_utf8_upgrade (self->incr_text);
		1943
		1944	if (self->incr_pos)
		1945	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1946	- (U8 *)SvPVX (self->incr_text);
		1947	}
		1948	}
		1949	else if (SvUTF8 (self->incr_text))
		1950	sv_utf8_upgrade (jsonstr);
		1951
		1952	{
		1953	STRLEN len;
		1954	const char *str = SvPV (jsonstr, len);
		1955	STRLEN cur = SvCUR (self->incr_text);
		1956
		1957	if (SvLEN (self->incr_text) <= cur + len)
		1958	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1959
		1960	Move (str, SvEND (self->incr_text), len, char);
		1961	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1962	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1963	}
		1964	}
		1965
		1966	if (GIMME_V != G_VOID)
		1967	do
		1968	{
		1969	char *offset;
		1970
		1971	if (!INCR_DONE (self))
		1972	{
		1973	incr_parse (self);
		1974
		1975	if (self->incr_pos > self->max_size && self->max_size)
		1976	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1977	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1978
		1979	if (!INCR_DONE (self))
		1980	break;
		1981	}
		1982
		1983	XPUSHs (decode_json (self->incr_text, self, &offset));
		1984
		1985	self->incr_pos -= offset - SvPVX (self->incr_text);
		1986	self->incr_nest = 0;
		1987	self->incr_mode = 0;
		1988
		1989	sv_chop (self->incr_text, offset);
		1990	}
		1991	while (GIMME_V == G_ARRAY);
		1992	}
		1993
		1994	SV *incr_text (JSON *self)
		1995	ATTRS: lvalue
1182	CODE:	1996	CODE:
1183	{	1997	{
1184	UV *uv = SvJSON (self);	1998	if (self->incr_pos)
1185	UV log2 = 0;	1999	croak ("incr_text can not be called when the incremental parser already started parsing");
1186		2000
1187	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	2001	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
1188
1189	while ((1UL << log2) < max_depth)
1190	++log2;
1191
1192	*uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1193
1194	RETVAL = newSVsv (self);
1195	}	2002	}
1196	OUTPUT:	2003	OUTPUT:
1197	RETVAL	2004	RETVAL
1198		2005
		2006	void incr_skip (JSON *self)
		2007	CODE:
		2008	{
		2009	if (self->incr_pos)
		2010	{
		2011	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2012	self->incr_pos = 0;
		2013	self->incr_nest = 0;
		2014	self->incr_mode = 0;
		2015	}
		2016	}
		2017
		2018	void incr_reset (JSON *self)
		2019	CODE:
		2020	{
		2021	SvREFCNT_dec (self->incr_text);
		2022	self->incr_text = 0;
		2023	self->incr_pos = 0;
		2024	self->incr_nest = 0;
		2025	self->incr_mode = 0;
		2026	}
		2027
		2028	void DESTROY (JSON *self)
		2029	CODE:
		2030	SvREFCNT_dec (self->cb_sk_object);
		2031	SvREFCNT_dec (self->cb_object);
		2032	SvREFCNT_dec (self->incr_text);
		2033
		2034	PROTOTYPES: ENABLE
		2035
1199	void encode (SV *self, SV *scalar)	2036	void encode_json (SV *scalar)
		2037	ALIAS:
		2038	to_json_ = 0
		2039	encode_json = F_UTF8
1200	PPCODE:	2040	PPCODE:
		2041	{
		2042	JSON json;
		2043	json_init (&json);
		2044	json.flags |= ix;
1201	XPUSHs (encode_json (scalar, *SvJSON (self)));	2045	XPUSHs (encode_json (scalar, &json));
		2046	}
1202		2047
1203	void decode (SV *self, SV *jsonstr)	2048	void decode_json (SV *jsonstr)
		2049	ALIAS:
		2050	from_json_ = 0
		2051	decode_json = F_UTF8
1204	PPCODE:	2052	PPCODE:
		2053	{
		2054	JSON json;
		2055	json_init (&json);
		2056	json.flags |= ix;
1205	XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));	2057	XPUSHs (decode_json (jsonstr, &json, 0));
1206
1207	void decode_prefix (SV *self, SV *jsonstr)
1208	PPCODE:
1209	{
1210	UV offset;
1211	EXTEND (SP, 2);
1212	PUSHs (decode_json (jsonstr, *SvJSON (self), &offset));
1213	PUSHs (sv_2mortal (newSVuv (offset)));
1214	}	2058	}
1215		2059
1216	PROTOTYPES: ENABLE
1217
1218	void to_json (SV *scalar)
1219	ALIAS:
1220	objToJson = 0
1221	PPCODE:
1222	XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8));
1223
1224	void from_json (SV *jsonstr)
1225	ALIAS:
1226	jsonToObj = 0
1227	PPCODE:
1228	XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8, 0));
1229

Diff Legend

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