ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.27 by root, Mon Apr 9 05:09:57 2007 UTC vs.
Revision 1.103 by root, Wed Jan 6 08:01:39 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
		26	#define F_LATIN1 0x00000002UL
15	#define F_UTF8 0x00000002UL	27	#define F_UTF8 0x00000004UL
16	#define F_INDENT 0x00000004UL	28	#define F_INDENT 0x00000008UL
17	#define F_CANONICAL 0x00000008UL	29	#define F_CANONICAL 0x00000010UL
18	#define F_SPACE_BEFORE 0x00000010UL	30	#define F_SPACE_BEFORE 0x00000020UL
19	#define F_SPACE_AFTER 0x00000020UL	31	#define F_SPACE_AFTER 0x00000040UL
20	#define F_ALLOW_NONREF 0x00000080UL	32	#define F_ALLOW_NONREF 0x00000100UL
21	#define F_SHRINK 0x00000100UL	33	#define F_SHRINK 0x00000200UL
		34	#define F_ALLOW_BLESSED 0x00000400UL
		35	#define F_CONV_BLESSED 0x00000800UL
22	#define F_MAXDEPTH 0xf8000000UL	36	#define F_RELAXED 0x00001000UL
23	#define S_MAXDEPTH 27	37	#define F_ALLOW_UNKNOWN 0x00002000UL
24		38	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
25	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
26
27	// F_SELFCONVERT? <=> to_json/toJson
28	// F_BLESSED? <=> { $__class__$ => }
29		39
30	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	40	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
31	#define F_DEFAULT (9UL << S_MAXDEPTH)
32		41
33	#define INIT_SIZE 32 // initial scalar size to be allocated	42	#define INIT_SIZE 32 // initial scalar size to be allocated
34	#define INDENT_STEP 3 // spaces per indentation level	43	#define INDENT_STEP 3 // spaces per indentation level
35		44
36	#define UTF8_MAX_LEN 11 // for perls UTF-X: max. number of octets per character
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
		50	#if __GNUC__ >= 3
		51	# define expect(expr,value) __builtin_expect ((expr), (value))
		52	# define INLINE static inline
		53	#else
		54	# define expect(expr,value) (expr)
		55	# define INLINE static
		56	#endif
		57
		58	#define expect_false(expr) expect ((expr) != 0, 0)
		59	#define expect_true(expr) expect ((expr) != 0, 1)
		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
42	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	}
43		110
44	/////////////////////////////////////////////////////////////////////////////	111	/////////////////////////////////////////////////////////////////////////////
45	// utility functions	112	// utility functions
46		113
47	static UV *	114	INLINE SV *
48	SvJSON (SV *sv)	115	get_bool (const char *name)
49	{	116	{
50	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))	117	SV *sv = get_sv (name, 1);
51	croak ("object is not of type JSON::XS");
52		118
53	return &SvUVX (SvRV (sv));	119	SvREADONLY_on (sv);
54	}	120	SvREADONLY_on (SvRV (sv));
55		121
56	static void	122	return sv;
		123	}
		124
		125	INLINE void
57	shrink (SV *sv)	126	shrink (SV *sv)
58	{	127	{
59	sv_utf8_downgrade (sv, 1);	128	sv_utf8_downgrade (sv, 1);
		129
60	if (SvLEN (sv) > SvCUR (sv) + 1)	130	if (SvLEN (sv) > SvCUR (sv) + 1)
61	{	131	{
62	#ifdef SvPV_shrink_to_cur	132	#ifdef SvPV_shrink_to_cur
63	SvPV_shrink_to_cur (sv);	133	SvPV_shrink_to_cur (sv);
64	#elif defined (SvPV_renew)	134	#elif defined (SvPV_renew)
…		…
70	// 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
71	// case of an error.	141	// case of an error.
72	// we special-case "safe" characters from U+80 .. U+7FF,	142	// we special-case "safe" characters from U+80 .. U+7FF,
73	// but use the very good perl function to parse anything else.	143	// but use the very good perl function to parse anything else.
74	// note that we never call this function for a ascii codepoints	144	// note that we never call this function for a ascii codepoints
75	static UV	145	INLINE UV
76	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)	146	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
77	{	147	{
78	if (s[0] > 0xdf || s[0] < 0xc2)	148	if (expect_true (len >= 2
79	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	149	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
80	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	150	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
81	{	151	{
82	*clen = 2;	152	*clen = 2;
83	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);	153	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
84	}	154	}
85	else	155	else
86	{	156	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
87	*clen = (STRLEN)-1;	157	}
88	return (UV)-1;	158
89	}	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);
90	}	190	}
91		191
92	/////////////////////////////////////////////////////////////////////////////	192	/////////////////////////////////////////////////////////////////////////////
93	// encoder	193	// encoder
94		194
…		…
96	typedef struct	196	typedef struct
97	{	197	{
98	char *cur; // SvPVX (sv) + current output position	198	char *cur; // SvPVX (sv) + current output position
99	char *end; // SvEND (sv)	199	char *end; // SvEND (sv)
100	SV *sv; // result scalar	200	SV *sv; // result scalar
101	U32 flags; // F_*	201	JSON json;
102	U32 indent; // indentation level	202	U32 indent; // indentation level
103	U32 maxdepth; // max. indentation/recursion level	203	UV limit; // escape character values >= this value when encoding
104	} enc_t;	204	} enc_t;
105		205
106	static void	206	INLINE void
107	need (enc_t *enc, STRLEN len)	207	need (enc_t *enc, STRLEN len)
108	{	208	{
109	if (enc->cur + len >= enc->end)	209	if (expect_false (enc->cur + len >= enc->end))
110	{	210	{
111	STRLEN cur = enc->cur - SvPVX (enc->sv);	211	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
112	SvGROW (enc->sv, cur + len + 1);	212	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
113	enc->cur = SvPVX (enc->sv) + cur;	213	enc->cur = SvPVX (enc->sv) + cur;
114	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	214	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
115	}	215	}
116	}	216	}
117		217
118	static void	218	INLINE void
119	encode_ch (enc_t *enc, char ch)	219	encode_ch (enc_t *enc, char ch)
120	{	220	{
121	need (enc, 1);	221	need (enc, 1);
122	*enc->cur++ = ch;	222	*enc->cur++ = ch;
123	}	223	}
…		…
131		231
132	while (str < end)	232	while (str < end)
133	{	233	{
134	unsigned char ch = *(unsigned char *)str;	234	unsigned char ch = *(unsigned char *)str;
135		235
136	if (ch >= 0x20 && ch < 0x80) // most common case	236	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
137	{	237	{
138	if (ch == '"') // but with slow exceptions	238	if (expect_false (ch == '"')) // but with slow exceptions
139	{	239	{
140	need (enc, len += 1);	240	need (enc, len += 1);
141	*enc->cur++ = '\\';	241	*enc->cur++ = '\\';
142	*enc->cur++ = '"';	242	*enc->cur++ = '"';
143	}	243	}
144	else if (ch == '\\')	244	else if (expect_false (ch == '\\'))
145	{	245	{
146	need (enc, len += 1);	246	need (enc, len += 1);
147	*enc->cur++ = '\\';	247	*enc->cur++ = '\\';
148	*enc->cur++ = '\\';	248	*enc->cur++ = '\\';
149	}	249	}
…		…
167	STRLEN clen;	267	STRLEN clen;
168	UV uch;	268	UV uch;
169		269
170	if (is_utf8)	270	if (is_utf8)
171	{	271	{
172	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
173	uch = decode_utf8 (str, end - str, &clen);	272	uch = decode_utf8 (str, end - str, &clen);
174	if (clen == (STRLEN)-1)	273	if (clen == (STRLEN)-1)
175	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	274	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
176	}	275	}
177	else	276	else
178	{	277	{
179	uch = ch;	278	uch = ch;
180	clen = 1;	279	clen = 1;
181	}	280	}
182		281
183	if (uch > 0x10FFFFUL)	282	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
184	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
185
186	if (uch < 0x80 || enc->flags & F_ASCII)
187	{	283	{
188	if (uch > 0xFFFFUL)	284	if (uch >= 0x10000UL)
189	{	285	{
		286	if (uch >= 0x110000UL)
		287	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		288
190	need (enc, len += 11);	289	need (enc, len += 11);
191	sprintf (enc->cur, "\\u%04x\\u%04x",	290	sprintf (enc->cur, "\\u%04x\\u%04x",
192	(int)((uch - 0x10000) / 0x400 + 0xD800),	291	(int)((uch - 0x10000) / 0x400 + 0xD800),
193	(int)((uch - 0x10000) % 0x400 + 0xDC00));	292	(int)((uch - 0x10000) % 0x400 + 0xDC00));
194	enc->cur += 12;	293	enc->cur += 12;
195	}	294	}
196	else	295	else
197	{	296	{
198	static char hexdigit [16] = "0123456789abcdef";
199	need (enc, len += 5);	297	need (enc, len += 5);
200	*enc->cur++ = '\\';	298	*enc->cur++ = '\\';
201	*enc->cur++ = 'u';	299	*enc->cur++ = 'u';
202	*enc->cur++ = hexdigit [ uch >> 12 ];	300	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
203	*enc->cur++ = hexdigit [(uch >> 8) & 15];	301	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
204	*enc->cur++ = hexdigit [(uch >> 4) & 15];	302	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
205	*enc->cur++ = hexdigit [(uch >> 0) & 15];	303	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
206	}	304	}
207		305
		306	str += clen;
		307	}
		308	else if (enc->json.flags & F_LATIN1)
		309	{
		310	*enc->cur++ = uch;
208	str += clen;	311	str += clen;
209	}	312	}
210	else if (is_utf8)	313	else if (is_utf8)
211	{	314	{
212	need (enc, len += clen);	315	need (enc, len += clen);
…		…
216	}	319	}
217	while (--clen);	320	while (--clen);
218	}	321	}
219	else	322	else
220	{	323	{
221	need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed	324	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
222	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	325	enc->cur = encode_utf8 (enc->cur, uch);
223	++str;	326	++str;
224	}	327	}
225	}	328	}
226	}	329	}
227	}	330	}
228		331
229	--len;	332	--len;
230	}	333	}
231	}	334	}
232		335
233	static void	336	INLINE void
234	encode_indent (enc_t *enc)	337	encode_indent (enc_t *enc)
235	{	338	{
236	if (enc->flags & F_INDENT)	339	if (enc->json.flags & F_INDENT)
237	{	340	{
238	int spaces = enc->indent * INDENT_STEP;	341	int spaces = enc->indent * INDENT_STEP;
239		342
240	need (enc, spaces);	343	need (enc, spaces);
241	memset (enc->cur, ' ', spaces);	344	memset (enc->cur, ' ', spaces);
242	enc->cur += spaces;	345	enc->cur += spaces;
243	}	346	}
244	}	347	}
245		348
246	static void	349	INLINE void
247	encode_space (enc_t *enc)	350	encode_space (enc_t *enc)
248	{	351	{
249	need (enc, 1);	352	need (enc, 1);
250	encode_ch (enc, ' ');	353	encode_ch (enc, ' ');
251	}	354	}
252		355
253	static void	356	INLINE void
254	encode_nl (enc_t *enc)	357	encode_nl (enc_t *enc)
255	{	358	{
256	if (enc->flags & F_INDENT)	359	if (enc->json.flags & F_INDENT)
257	{	360	{
258	need (enc, 1);	361	need (enc, 1);
259	encode_ch (enc, '\n');	362	encode_ch (enc, '\n');
260	}	363	}
261	}	364	}
262		365
263	static void	366	INLINE void
264	encode_comma (enc_t *enc)	367	encode_comma (enc_t *enc)
265	{	368	{
266	encode_ch (enc, ',');	369	encode_ch (enc, ',');
267		370
268	if (enc->flags & F_INDENT)	371	if (enc->json.flags & F_INDENT)
269	encode_nl (enc);	372	encode_nl (enc);
270	else if (enc->flags & F_SPACE_AFTER)	373	else if (enc->json.flags & F_SPACE_AFTER)
271	encode_space (enc);	374	encode_space (enc);
272	}	375	}
273		376
274	static void encode_sv (enc_t *enc, SV *sv);	377	static void encode_sv (enc_t *enc, SV *sv);
275		378
276	static void	379	static void
277	encode_av (enc_t *enc, AV *av)	380	encode_av (enc_t *enc, AV *av)
278	{	381	{
279	int i, len = av_len (av);	382	int i, len = av_len (av);
280		383
281	if (enc->indent >= enc->maxdepth)	384	if (enc->indent >= enc->json.max_depth)
282	croak ("data structure too deep (hit recursion limit)");	385	croak (ERR_NESTING_EXCEEDED);
283		386
284	encode_ch (enc, '['); encode_nl (enc);	387	encode_ch (enc, '[');
285	++enc->indent;	388
		389	if (len >= 0)
		390	{
		391	encode_nl (enc); ++enc->indent;
286		392
287	for (i = 0; i <= len; ++i)	393	for (i = 0; i <= len; ++i)
288	{	394	{
		395	SV **svp = av_fetch (av, i, 0);
		396
289	encode_indent (enc);	397	encode_indent (enc);
290	encode_sv (enc, *av_fetch (av, i, 0));
291		398
		399	if (svp)
		400	encode_sv (enc, *svp);
		401	else
		402	encode_str (enc, "null", 4, 0);
		403
292	if (i < len)	404	if (i < len)
293	encode_comma (enc);	405	encode_comma (enc);
294	}	406	}
295		407
		408	encode_nl (enc); --enc->indent; encode_indent (enc);
		409	}
		410
296	encode_nl (enc);	411	encode_ch (enc, ']');
297
298	--enc->indent;
299	encode_indent (enc); encode_ch (enc, ']');
300	}	412	}
301		413
302	static void	414	static void
303	encode_he (enc_t *enc, HE *he)	415	encode_hk (enc_t *enc, HE *he)
304	{	416	{
305	encode_ch (enc, '"');	417	encode_ch (enc, '"');
306		418
307	if (HeKLEN (he) == HEf_SVKEY)	419	if (HeKLEN (he) == HEf_SVKEY)
308	{	420	{
…		…
318	else	430	else
319	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	431	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
320		432
321	encode_ch (enc, '"');	433	encode_ch (enc, '"');
322		434
323	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	435	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
324	encode_ch (enc, ':');	436	encode_ch (enc, ':');
325	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	437	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
326	encode_sv (enc, HeVAL (he));
327	}	438	}
328		439
329	// compare hash entries, used when all keys are bytestrings	440	// compare hash entries, used when all keys are bytestrings
330	static int	441	static int
331	he_cmp_fast (const void *a_, const void *b_)	442	he_cmp_fast (const void *a_, const void *b_)
…		…
336	HE *b = *(HE **)b_;	447	HE *b = *(HE **)b_;
337		448
338	STRLEN la = HeKLEN (a);	449	STRLEN la = HeKLEN (a);
339	STRLEN lb = HeKLEN (b);	450	STRLEN lb = HeKLEN (b);
340		451
341	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	452	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
342	cmp = la - lb;	453	cmp = lb - la;
343		454
344	return cmp;	455	return cmp;
345	}	456	}
346		457
347	// compare hash entries, used when some keys are sv's or utf-x	458	// compare hash entries, used when some keys are sv's or utf-x
348	static int	459	static int
349	he_cmp_slow (const void *a, const void *b)	460	he_cmp_slow (const void *a, const void *b)
350	{	461	{
351	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	462	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
352	}	463	}
353		464
354	static void	465	static void
355	encode_hv (enc_t *enc, HV *hv)	466	encode_hv (enc_t *enc, HV *hv)
356	{	467	{
357	int count, i;	468	HE *he;
358		469
359	if (enc->indent >= enc->maxdepth)	470	if (enc->indent >= enc->json.max_depth)
360	croak ("data structure too deep (hit recursion limit)");	471	croak (ERR_NESTING_EXCEEDED);
361		472
362	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	473	encode_ch (enc, '{');
363		474
364	if ((count = hv_iterinit (hv)))
365	{
366	// for canonical output we have to sort by keys first	475	// for canonical output we have to sort by keys first
367	// actually, this is mostly due to the stupid so-called	476	// actually, this is mostly due to the stupid so-called
368	// security workaround added somewhere in 5.8.x.	477	// security workaround added somewhere in 5.8.x
369	// that randomises hash orderings	478	// that randomises hash orderings
370	if (enc->flags & F_CANONICAL)	479	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
		480	{
		481	int count = hv_iterinit (hv);
		482
		483	if (SvMAGICAL (hv))
371	{	484	{
		485	// need to count by iterating. could improve by dynamically building the vector below
		486	// but I don't care for the speed of this special case.
		487	// note also that we will run into undefined behaviour when the two iterations
		488	// do not result in the same count, something I might care for in some later release.
		489
		490	count = 0;
		491	while (hv_iternext (hv))
		492	++count;
		493
		494	hv_iterinit (hv);
		495	}
		496
		497	if (count)
		498	{
		499	int i, fast = 1;
		500	#if defined(__BORLANDC__) || defined(_MSC_VER)
		501	HE **hes = _alloca (count * sizeof (HE));
		502	#else
372	HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode	503	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
373	int fast = 1;	504	#endif
374		505
375	i = 0;	506	i = 0;
376	while ((he = hv_iternext (hv)))	507	while ((he = hv_iternext (hv)))
377	{	508	{
378	hes [i++] = he;	509	hes [i++] = he;
…		…
400		531
401	FREETMPS;	532	FREETMPS;
402	LEAVE;	533	LEAVE;
403	}	534	}
404		535
405	for (i = 0; i < count; ++i)	536	encode_nl (enc); ++enc->indent;
		537
		538	while (count--)
406	{	539	{
407	encode_indent (enc);	540	encode_indent (enc);
		541	he = hes [count];
408	encode_he (enc, hes [i]);	542	encode_hk (enc, he);
		543	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
409		544
410	if (i < count - 1)	545	if (count)
411	encode_comma (enc);	546	encode_comma (enc);
412	}	547	}
413		548
		549	encode_nl (enc); --enc->indent; encode_indent (enc);
		550	}
		551	}
		552	else
		553	{
		554	if (hv_iterinit (hv) || SvMAGICAL (hv))
		555	if ((he = hv_iternext (hv)))
		556	{
		557	encode_nl (enc); ++enc->indent;
		558
		559	for (;;)
		560	{
414	encode_nl (enc);	561	encode_indent (enc);
		562	encode_hk (enc, he);
		563	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		564
		565	if (!(he = hv_iternext (hv)))
		566	break;
		567
		568	encode_comma (enc);
		569	}
		570
		571	encode_nl (enc); --enc->indent; encode_indent (enc);
		572	}
		573	}
		574
		575	encode_ch (enc, '}');
		576	}
		577
		578	// encode objects, arrays and special \0=false and \1=true values.
		579	static void
		580	encode_rv (enc_t *enc, SV *sv)
		581	{
		582	svtype svt;
		583
		584	SvGETMAGIC (sv);
		585	svt = SvTYPE (sv);
		586
		587	if (expect_false (SvOBJECT (sv)))
		588	{
		589	HV *stash = !JSON_SLOW || json_boolean_stash
		590	? json_boolean_stash
		591	: gv_stashpv ("JSON::XS::Boolean", 1);
		592
		593	if (SvSTASH (sv) == stash)
		594	{
		595	if (SvIV (sv))
		596	encode_str (enc, "true", 4, 0);
		597	else
		598	encode_str (enc, "false", 5, 0);
415	}	599	}
416	else	600	else
417	{	601	{
418	HE *he = hv_iternext (hv);	602	#if 0
419		603	if (0 && sv_derived_from (rv, "JSON::Literal"))
420	for (;;)
421	{	604	{
422	encode_indent (enc);	605	// not yet
423	encode_he (enc, he);
424
425	if (!(he = hv_iternext (hv)))
426	break;
427
428	encode_comma (enc);
429	}	606	}
		607	#endif
		608	if (enc->json.flags & F_CONV_BLESSED)
		609	{
		610	// we re-bless the reference to get overload and other niceties right
		611	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
430		612
		613	if (to_json)
		614	{
		615	dSP;
		616
		617	ENTER; SAVETMPS; PUSHMARK (SP);
		618	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		619
		620	// calling with G_SCALAR ensures that we always get a 1 return value
		621	PUTBACK;
		622	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		623	SPAGAIN;
		624
		625	// catch this surprisingly common error
		626	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		627	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		628
		629	sv = POPs;
		630	PUTBACK;
		631
431	encode_nl (enc);	632	encode_sv (enc, sv);
		633
		634	FREETMPS; LEAVE;
		635	}
		636	else if (enc->json.flags & F_ALLOW_BLESSED)
		637	encode_str (enc, "null", 4, 0);
		638	else
		639	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		640	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		641	}
		642	else if (enc->json.flags & F_ALLOW_BLESSED)
		643	encode_str (enc, "null", 4, 0);
		644	else
		645	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		646	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
432	}	647	}
433	}	648	}
434
435	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
436	}
437
438	// encode objects, arrays and special \0=false and \1=true values.
439	static void
440	encode_rv (enc_t *enc, SV *sv)
441	{
442	svtype svt;
443
444	SvGETMAGIC (sv);
445	svt = SvTYPE (sv);
446
447	if (svt == SVt_PVHV)	649	else if (svt == SVt_PVHV)
448	encode_hv (enc, (HV *)sv);	650	encode_hv (enc, (HV *)sv);
449	else if (svt == SVt_PVAV)	651	else if (svt == SVt_PVAV)
450	encode_av (enc, (AV *)sv);	652	encode_av (enc, (AV *)sv);
451	else if (svt < SVt_PVAV)	653	else if (svt < SVt_PVAV)
452	{	654	{
453	if (SvNIOK (sv) && SvIV (sv) == 0)	655	STRLEN len = 0;
		656	char *pv = svt ? SvPV (sv, len) : 0;
		657
		658	if (len == 1 && *pv == '1')
		659	encode_str (enc, "true", 4, 0);
		660	else if (len == 1 && *pv == '0')
454	encode_str (enc, "false", 5, 0);	661	encode_str (enc, "false", 5, 0);
455	else if (SvNIOK (sv) && SvIV (sv) == 1)	662	else if (enc->json.flags & F_ALLOW_UNKNOWN)
456	encode_str (enc, "true", 4, 0);	663	encode_str (enc, "null", 4, 0);
457	else	664	else
458	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	665	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
459	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	666	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
460	}	667	}
		668	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		669	encode_str (enc, "null", 4, 0);
461	else	670	else
462	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	671	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
463	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	672	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
464	}	673	}
465		674
…		…
476	encode_str (enc, str, len, SvUTF8 (sv));	685	encode_str (enc, str, len, SvUTF8 (sv));
477	encode_ch (enc, '"');	686	encode_ch (enc, '"');
478	}	687	}
479	else if (SvNOKp (sv))	688	else if (SvNOKp (sv))
480	{	689	{
		690	// trust that perl will do the right thing w.r.t. JSON syntax.
481	need (enc, NV_DIG + 32);	691	need (enc, NV_DIG + 32);
482	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	692	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
483	enc->cur += strlen (enc->cur);	693	enc->cur += strlen (enc->cur);
484	}	694	}
485	else if (SvIOKp (sv))	695	else if (SvIOKp (sv))
486	{	696	{
		697	// we assume we can always read an IV as a UV and vice versa
		698	// we assume two's complement
		699	// we assume no aliasing issues in the union
		700	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		701	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		702	{
		703	// optimise the "small number case"
		704	// code will likely be branchless and use only a single multiplication
		705	// works for numbers up to 59074
		706	I32 i = SvIVX (sv);
		707	U32 u;
		708	char digit, nz = 0;
		709
487	need (enc, 64);	710	need (enc, 6);
		711
		712	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		713	u = i < 0 ? -i : i;
		714
		715	// convert to 4.28 fixed-point representation
		716	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		717
		718	// now output digit by digit, each time masking out the integer part
		719	// and multiplying by 5 while moving the decimal point one to the right,
		720	// resulting in a net multiplication by 10.
		721	// we always write the digit to memory but conditionally increment
		722	// the pointer, to enable the use of conditional move instructions.
		723	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		724	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		725	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		726	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		727	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		728	}
		729	else
		730	{
		731	// large integer, use the (rather slow) snprintf way.
		732	need (enc, IVUV_MAXCHARS);
488	enc->cur +=	733	enc->cur +=
489	SvIsUV(sv)	734	SvIsUV(sv)
490	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	735	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
491	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	736	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		737	}
492	}	738	}
493	else if (SvROK (sv))	739	else if (SvROK (sv))
494	encode_rv (enc, SvRV (sv));	740	encode_rv (enc, SvRV (sv));
495	else if (!SvOK (sv))	741	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
496	encode_str (enc, "null", 4, 0);	742	encode_str (enc, "null", 4, 0);
497	else	743	else
498	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	744	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
499	SvPV_nolen (sv), SvFLAGS (sv));	745	SvPV_nolen (sv), SvFLAGS (sv));
500	}	746	}
501		747
502	static SV *	748	static SV *
503	encode_json (SV *scalar, U32 flags)	749	encode_json (SV *scalar, JSON *json)
504	{	750	{
505	enc_t enc;	751	enc_t enc;
506		752
507	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	753	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
508	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	754	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
509		755
510	enc.flags = flags;	756	enc.json = *json;
511	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	757	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
512	enc.cur = SvPVX (enc.sv);	758	enc.cur = SvPVX (enc.sv);
513	enc.end = SvEND (enc.sv);	759	enc.end = SvEND (enc.sv);
514	enc.indent = 0;	760	enc.indent = 0;
515	enc.maxdepth = DEC_DEPTH (flags);	761	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		762	: enc.json.flags & F_LATIN1 ? 0x000100UL
		763	: 0x110000UL;
516		764
517	SvPOK_only (enc.sv);	765	SvPOK_only (enc.sv);
518	encode_sv (&enc, scalar);	766	encode_sv (&enc, scalar);
519		767	encode_nl (&enc);
520	if (!(flags & (F_ASCII | F_UTF8)))
521	SvUTF8_on (enc.sv);
522		768
523	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	769	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
524	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	770	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
525		771
		772	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
		773	SvUTF8_on (enc.sv);
		774
526	if (enc.flags & F_SHRINK)	775	if (enc.json.flags & F_SHRINK)
527	shrink (enc.sv);	776	shrink (enc.sv);
528		777
529	return enc.sv;	778	return enc.sv;
530	}	779	}
531		780
…		…
536	typedef struct	785	typedef struct
537	{	786	{
538	char *cur; // current parser pointer	787	char *cur; // current parser pointer
539	char *end; // end of input string	788	char *end; // end of input string
540	const char *err; // parse error, if != 0	789	const char *err; // parse error, if != 0
541	U32 flags; // F_*	790	JSON json;
542	U32 depth; // recursion depth	791	U32 depth; // recursion depth
543	U32 maxdepth; // recursion depth limit	792	U32 maxdepth; // recursion depth limit
544	} dec_t;	793	} dec_t;
545		794
546	static void	795	INLINE void
		796	decode_comment (dec_t *dec)
		797	{
		798	// only '#'-style comments allowed a.t.m.
		799
		800	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		801	++dec->cur;
		802	}
		803
		804	INLINE void
547	decode_ws (dec_t *dec)	805	decode_ws (dec_t *dec)
548	{	806	{
549	for (;;)	807	for (;;)
550	{	808	{
551	char ch = *dec->cur;	809	char ch = *dec->cur;
552		810
553	if (ch > 0x20	811	if (ch > 0x20)
		812	{
		813	if (expect_false (ch == '#'))
		814	{
		815	if (dec->json.flags & F_RELAXED)
		816	decode_comment (dec);
		817	else
		818	break;
		819	}
		820	else
		821	break;
		822	}
554	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	823	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
555	break;	824	break; // parse error, but let higher level handle it, gives better error messages
556		825
557	++dec->cur;	826	++dec->cur;
558	}	827	}
559	}	828	}
560		829
…		…
564	if (*dec->cur != ch) \	833	if (*dec->cur != ch) \
565	ERR (# ch " expected"); \	834	ERR (# ch " expected"); \
566	++dec->cur; \	835	++dec->cur; \
567	SE	836	SE
568		837
569	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	838	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
570	#define DEC_DEC_DEPTH --dec->depth	839	#define DEC_DEC_DEPTH --dec->depth
571		840
572	static SV *decode_sv (dec_t *dec);	841	static SV *decode_sv (dec_t *dec);
573		842
574	static signed char decode_hexdigit[256];	843	static signed char decode_hexdigit[256];
…		…
577	decode_4hex (dec_t *dec)	846	decode_4hex (dec_t *dec)
578	{	847	{
579	signed char d1, d2, d3, d4;	848	signed char d1, d2, d3, d4;
580	unsigned char *cur = (unsigned char *)dec->cur;	849	unsigned char *cur = (unsigned char *)dec->cur;
581		850
582	d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected");	851	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
583	d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected");	852	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
584	d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected");	853	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
585	d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected");	854	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
586		855
587	dec->cur += 4;	856	dec->cur += 4;
588		857
589	return ((UV)d1) << 12	858	return ((UV)d1) << 12
590	| ((UV)d2) << 8	859	| ((UV)d2) << 8
…		…
598	static SV *	867	static SV *
599	decode_str (dec_t *dec)	868	decode_str (dec_t *dec)
600	{	869	{
601	SV *sv = 0;	870	SV *sv = 0;
602	int utf8 = 0;	871	int utf8 = 0;
		872	char *dec_cur = dec->cur;
603		873
604	do	874	do
605	{	875	{
606	char buf [SHORT_STRING_LEN + UTF8_MAX_LEN];	876	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
607	char *cur = buf;	877	char *cur = buf;
608		878
609	do	879	do
610	{	880	{
611	unsigned char ch = *(unsigned char *)dec->cur++;	881	unsigned char ch = *(unsigned char *)dec_cur++;
612		882
613	if (ch == '"')	883	if (expect_false (ch == '"'))
614	{	884	{
615	--dec->cur;	885	--dec_cur;
616	break;	886	break;
617	}	887	}
618	else if (ch == '\\')	888	else if (expect_false (ch == '\\'))
619	{	889	{
620	switch (*dec->cur)	890	switch (*dec_cur)
621	{	891	{
622	case '\\':	892	case '\\':
623	case '/':	893	case '/':
624	case '"': *cur++ = *dec->cur++; break;	894	case '"': *cur++ = *dec_cur++; break;
625		895
626	case 'b': ++dec->cur; *cur++ = '\010'; break;	896	case 'b': ++dec_cur; *cur++ = '\010'; break;
627	case 't': ++dec->cur; *cur++ = '\011'; break;	897	case 't': ++dec_cur; *cur++ = '\011'; break;
628	case 'n': ++dec->cur; *cur++ = '\012'; break;	898	case 'n': ++dec_cur; *cur++ = '\012'; break;
629	case 'f': ++dec->cur; *cur++ = '\014'; break;	899	case 'f': ++dec_cur; *cur++ = '\014'; break;
630	case 'r': ++dec->cur; *cur++ = '\015'; break;	900	case 'r': ++dec_cur; *cur++ = '\015'; break;
631		901
632	case 'u':	902	case 'u':
633	{	903	{
634	UV lo, hi;	904	UV lo, hi;
635	++dec->cur;	905	++dec_cur;
636		906
		907	dec->cur = dec_cur;
637	hi = decode_4hex (dec);	908	hi = decode_4hex (dec);
		909	dec_cur = dec->cur;
638	if (hi == (UV)-1)	910	if (hi == (UV)-1)
639	goto fail;	911	goto fail;
640		912
641	// possibly a surrogate pair	913	// possibly a surrogate pair
642	if (hi >= 0xd800)	914	if (hi >= 0xd800)
643	if (hi < 0xdc00)	915	if (hi < 0xdc00)
644	{	916	{
645	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	917	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
646	ERR ("missing low surrogate character in surrogate pair");	918	ERR ("missing low surrogate character in surrogate pair");
647		919
648	dec->cur += 2;	920	dec_cur += 2;
649		921
		922	dec->cur = dec_cur;
650	lo = decode_4hex (dec);	923	lo = decode_4hex (dec);
		924	dec_cur = dec->cur;
651	if (lo == (UV)-1)	925	if (lo == (UV)-1)
652	goto fail;	926	goto fail;
653		927
654	if (lo < 0xdc00 || lo >= 0xe000)	928	if (lo < 0xdc00 || lo >= 0xe000)
655	ERR ("surrogate pair expected");	929	ERR ("surrogate pair expected");
…		…
661		935
662	if (hi >= 0x80)	936	if (hi >= 0x80)
663	{	937	{
664	utf8 = 1;	938	utf8 = 1;
665		939
666	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	940	cur = encode_utf8 (cur, hi);
667	}	941	}
668	else	942	else
669	*cur++ = hi;	943	*cur++ = hi;
670	}	944	}
671	break;	945	break;
672		946
673	default:	947	default:
674	--dec->cur;	948	--dec_cur;
675	ERR ("illegal backslash escape sequence in string");	949	ERR ("illegal backslash escape sequence in string");
676	}	950	}
677	}	951	}
678	else if (ch >= 0x20 && ch <= 0x7f)	952	else if (expect_true (ch >= 0x20 && ch < 0x80))
679	*cur++ = ch;	953	*cur++ = ch;
680	else if (ch >= 0x80)	954	else if (ch >= 0x80)
681	{	955	{
682	STRLEN clen;	956	STRLEN clen;
683	UV uch;
684		957
685	--dec->cur;	958	--dec_cur;
686		959
687	uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	960	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
688	if (clen == (STRLEN)-1)	961	if (clen == (STRLEN)-1)
689	ERR ("malformed UTF-8 character in JSON string");	962	ERR ("malformed UTF-8 character in JSON string");
690		963
691	do	964	do
692	*cur++ = *dec->cur++;	965	*cur++ = *dec_cur++;
693	while (--clen);	966	while (--clen);
694		967
695	utf8 = 1;	968	utf8 = 1;
696	}	969	}
697	else	970	else
698	{	971	{
699	--dec->cur;	972	--dec_cur;
700		973
701	if (!ch)	974	if (!ch)
702	ERR ("unexpected end of string while parsing JSON string");	975	ERR ("unexpected end of string while parsing JSON string");
703	else	976	else
704	ERR ("invalid character encountered while parsing JSON string");	977	ERR ("invalid character encountered while parsing JSON string");
…		…
709	{	982	{
710	STRLEN len = cur - buf;	983	STRLEN len = cur - buf;
711		984
712	if (sv)	985	if (sv)
713	{	986	{
714	SvGROW (sv, SvCUR (sv) + len + 1);	987	STRLEN cur = SvCUR (sv);
		988
		989	if (SvLEN (sv) <= cur + len)
		990	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		991
715	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	992	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
716	SvCUR_set (sv, SvCUR (sv) + len);	993	SvCUR_set (sv, SvCUR (sv) + len);
717	}	994	}
718	else	995	else
719	sv = newSVpvn (buf, len);	996	sv = newSVpvn (buf, len);
720	}	997	}
721	}	998	}
722	while (*dec->cur != '"');	999	while (*dec_cur != '"');
723		1000
724	++dec->cur;	1001	++dec_cur;
725		1002
726	if (sv)	1003	if (sv)
727	{	1004	{
728	SvPOK_only (sv);	1005	SvPOK_only (sv);
729	*SvEND (sv) = 0;	1006	*SvEND (sv) = 0;
…		…
732	SvUTF8_on (sv);	1009	SvUTF8_on (sv);
733	}	1010	}
734	else	1011	else
735	sv = newSVpvn ("", 0);	1012	sv = newSVpvn ("", 0);
736		1013
		1014	dec->cur = dec_cur;
737	return sv;	1015	return sv;
738		1016
739	fail:	1017	fail:
		1018	dec->cur = dec_cur;
740	return 0;	1019	return 0;
741	}	1020	}
742		1021
743	static SV *	1022	static SV *
744	decode_num (dec_t *dec)	1023	decode_num (dec_t *dec)
…		…
802	is_nv = 1;	1081	is_nv = 1;
803	}	1082	}
804		1083
805	if (!is_nv)	1084	if (!is_nv)
806	{	1085	{
807	UV uv;	1086	int len = dec->cur - start;
808	int numtype = grok_number (start, dec->cur - start, &uv);	1087
809	if (numtype & IS_NUMBER_IN_UV)	1088	// special case the rather common 1..5-digit-int case
810	if (numtype & IS_NUMBER_NEG)	1089	if (*start == '-')
		1090	switch (len)
811	{	1091	{
812	if (uv < (UV)IV_MIN)	1092	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
813	return newSViv (-(IV)uv);	1093	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
		1094	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1095	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1096	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
814	}	1097	}
		1098	else
		1099	switch (len)
		1100	{
		1101	case 1: return newSViv ( start [0] - '0' * 1);
		1102	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1103	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1104	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1105	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1106	}
		1107
		1108	{
		1109	UV uv;
		1110	int numtype = grok_number (start, len, &uv);
		1111	if (numtype & IS_NUMBER_IN_UV)
		1112	if (numtype & IS_NUMBER_NEG)
		1113	{
		1114	if (uv < (UV)IV_MIN)
		1115	return newSViv (-(IV)uv);
		1116	}
815	else	1117	else
816	return newSVuv (uv);	1118	return newSVuv (uv);
817	}	1119	}
818		1120
		1121	len -= *start == '-' ? 1 : 0;
		1122
		1123	// does not fit into IV or UV, try NV
		1124	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1125	#if defined (LDBL_DIG)
		1126	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1127	#endif
		1128	)
		1129	// fits into NV without loss of precision
		1130	return newSVnv (Atof (start));
		1131
		1132	// everything else fails, convert it to a string
		1133	return newSVpvn (start, dec->cur - start);
		1134	}
		1135
		1136	// loss of precision here
819	return newSVnv (Atof (start));	1137	return newSVnv (Atof (start));
820		1138
821	fail:	1139	fail:
822	return 0;	1140	return 0;
823	}	1141	}
…		…
853		1171
854	if (*dec->cur != ',')	1172	if (*dec->cur != ',')
855	ERR (", or ] expected while parsing array");	1173	ERR (", or ] expected while parsing array");
856		1174
857	++dec->cur;	1175	++dec->cur;
		1176
		1177	decode_ws (dec);
		1178
		1179	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1180	{
		1181	++dec->cur;
		1182	break;
		1183	}
858	}	1184	}
859		1185
860	DEC_DEC_DEPTH;	1186	DEC_DEC_DEPTH;
861	return newRV_noinc ((SV *)av);	1187	return newRV_noinc ((SV *)av);
862		1188
…		…
867	}	1193	}
868		1194
869	static SV *	1195	static SV *
870	decode_hv (dec_t *dec)	1196	decode_hv (dec_t *dec)
871	{	1197	{
		1198	SV *sv;
872	HV *hv = newHV ();	1199	HV *hv = newHV ();
873		1200
874	DEC_INC_DEPTH;	1201	DEC_INC_DEPTH;
875	decode_ws (dec);	1202	decode_ws (dec);
876		1203
877	if (*dec->cur == '}')	1204	if (*dec->cur == '}')
878	++dec->cur;	1205	++dec->cur;
879	else	1206	else
880	for (;;)	1207	for (;;)
881	{	1208	{
		1209	EXPECT_CH ('"');
		1210
		1211	// heuristic: assume that
		1212	// a) decode_str + hv_store_ent are abysmally slow.
		1213	// b) most hash keys are short, simple ascii text.
		1214	// => try to "fast-match" such strings to avoid
		1215	// the overhead of decode_str + hv_store_ent.
		1216	{
882	SV *key, *value;	1217	SV *value;
		1218	char *p = dec->cur;
		1219	char *e = p + 24; // only try up to 24 bytes
883		1220
884	decode_ws (dec); EXPECT_CH ('"');	1221	for (;;)
885
886	key = decode_str (dec);
887	if (!key)
888	goto fail;
889
890	decode_ws (dec); EXPECT_CH (':');
891
892	value = decode_sv (dec);
893	if (!value)
894	{	1222	{
		1223	// the >= 0x80 is false on most architectures
		1224	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1225	{
		1226	// slow path, back up and use decode_str
		1227	SV *key = decode_str (dec);
		1228	if (!key)
		1229	goto fail;
		1230
		1231	decode_ws (dec); EXPECT_CH (':');
		1232
		1233	decode_ws (dec);
		1234	value = decode_sv (dec);
		1235	if (!value)
		1236	{
		1237	SvREFCNT_dec (key);
		1238	goto fail;
		1239	}
		1240
		1241	hv_store_ent (hv, key, value, 0);
895	SvREFCNT_dec (key);	1242	SvREFCNT_dec (key);
		1243
		1244	break;
		1245	}
		1246	else if (*p == '"')
		1247	{
		1248	// fast path, got a simple key
		1249	char *key = dec->cur;
		1250	int len = p - key;
		1251	dec->cur = p + 1;
		1252
		1253	decode_ws (dec); EXPECT_CH (':');
		1254
		1255	decode_ws (dec);
		1256	value = decode_sv (dec);
		1257	if (!value)
896	goto fail;	1258	goto fail;
		1259
		1260	hv_store (hv, key, len, value, 0);
		1261
		1262	break;
		1263	}
		1264
		1265	++p;
897	}	1266	}
898		1267	}
899	hv_store_ent (hv, key, value, 0);
900	SvREFCNT_dec (key);
901		1268
902	decode_ws (dec);	1269	decode_ws (dec);
903		1270
904	if (*dec->cur == '}')	1271	if (*dec->cur == '}')
905	{	1272	{
…		…
909		1276
910	if (*dec->cur != ',')	1277	if (*dec->cur != ',')
911	ERR (", or } expected while parsing object/hash");	1278	ERR (", or } expected while parsing object/hash");
912		1279
913	++dec->cur;	1280	++dec->cur;
		1281
		1282	decode_ws (dec);
		1283
		1284	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1285	{
		1286	++dec->cur;
		1287	break;
		1288	}
914	}	1289	}
915		1290
916	DEC_DEC_DEPTH;	1291	DEC_DEC_DEPTH;
917	return newRV_noinc ((SV *)hv);	1292	sv = newRV_noinc ((SV *)hv);
		1293
		1294	// check filter callbacks
		1295	if (dec->json.flags & F_HOOK)
		1296	{
		1297	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1298	{
		1299	HE *cb, *he;
		1300
		1301	hv_iterinit (hv);
		1302	he = hv_iternext (hv);
		1303	hv_iterinit (hv);
		1304
		1305	// the next line creates a mortal sv each time its called.
		1306	// might want to optimise this for common cases.
		1307	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1308
		1309	if (cb)
		1310	{
		1311	dSP;
		1312	int count;
		1313
		1314	ENTER; SAVETMPS; PUSHMARK (SP);
		1315	XPUSHs (HeVAL (he));
		1316
		1317	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1318
		1319	if (count == 1)
		1320	{
		1321	sv = newSVsv (POPs);
		1322	FREETMPS; LEAVE;
		1323	return sv;
		1324	}
		1325
		1326	FREETMPS; LEAVE;
		1327	}
		1328	}
		1329
		1330	if (dec->json.cb_object)
		1331	{
		1332	dSP;
		1333	int count;
		1334
		1335	ENTER; SAVETMPS; PUSHMARK (SP);
		1336	XPUSHs (sv_2mortal (sv));
		1337
		1338	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1339
		1340	if (count == 1)
		1341	{
		1342	sv = newSVsv (POPs);
		1343	FREETMPS; LEAVE;
		1344	return sv;
		1345	}
		1346
		1347	SvREFCNT_inc (sv);
		1348	FREETMPS; LEAVE;
		1349	}
		1350	}
		1351
		1352	return sv;
918		1353
919	fail:	1354	fail:
920	SvREFCNT_dec (hv);	1355	SvREFCNT_dec (hv);
921	DEC_DEC_DEPTH;	1356	DEC_DEC_DEPTH;
922	return 0;	1357	return 0;
923	}	1358	}
924		1359
925	static SV *	1360	static SV *
926	decode_sv (dec_t *dec)	1361	decode_sv (dec_t *dec)
927	{	1362	{
928	decode_ws (dec);	1363	// the beauty of JSON: you need exactly one character lookahead
		1364	// to parse everything.
929	switch (*dec->cur)	1365	switch (*dec->cur)
930	{	1366	{
931	case '"': ++dec->cur; return decode_str (dec);	1367	case '"': ++dec->cur; return decode_str (dec);
932	case '[': ++dec->cur; return decode_av (dec);	1368	case '[': ++dec->cur; return decode_av (dec);
933	case '{': ++dec->cur; return decode_hv (dec);	1369	case '{': ++dec->cur; return decode_hv (dec);
934		1370
935	case '-':	1371	case '-':
936	case '0': case '1': case '2': case '3': case '4':	1372	case '0': case '1': case '2': case '3': case '4':
937	case '5': case '6': case '7': case '8': case '9':	1373	case '5': case '6': case '7': case '8': case '9':
938	return decode_num (dec);	1374	return decode_num (dec);
939		1375
940	case 't':	1376	case 't':
941	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1377	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
942	{	1378	{
943	dec->cur += 4;	1379	dec->cur += 4;
		1380	#if JSON_SLOW
		1381	json_true = get_bool ("JSON::XS::true");
		1382	#endif
944	return newSViv (1);	1383	return newSVsv (json_true);
945	}	1384	}
946	else	1385	else
947	ERR ("'true' expected");	1386	ERR ("'true' expected");
948		1387
949	break;	1388	break;
950		1389
951	case 'f':	1390	case 'f':
952	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1391	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
953	{	1392	{
954	dec->cur += 5;	1393	dec->cur += 5;
		1394	#if JSON_SLOW
		1395	json_false = get_bool ("JSON::XS::false");
		1396	#endif
955	return newSViv (0);	1397	return newSVsv (json_false);
956	}	1398	}
957	else	1399	else
958	ERR ("'false' expected");	1400	ERR ("'false' expected");
959		1401
960	break;	1402	break;
…		…
978	fail:	1420	fail:
979	return 0;	1421	return 0;
980	}	1422	}
981		1423
982	static SV *	1424	static SV *
983	decode_json (SV *string, U32 flags)	1425	decode_json (SV *string, JSON *json, char **offset_return)
984	{	1426	{
985	dec_t dec;	1427	dec_t dec;
986	SV *sv;	1428	SV *sv;
987		1429
		1430	/* work around bugs in 5.10 where manipulating magic values
		1431	* will perl ignore the magic in subsequent accesses
		1432	*/
		1433	/*SvGETMAGIC (string);*/
		1434	if (SvMAGICAL (string))
		1435	string = sv_2mortal (newSVsv (string));
		1436
988	SvUPGRADE (string, SVt_PV);	1437	SvUPGRADE (string, SVt_PV);
989		1438
		1439	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1440	* assertion with -DDEBUGGING, although SvCUR is documented to
		1441	* return the xpv_cur field which certainly exists after upgrading.
		1442	* according to nicholas clark, calling SvPOK fixes this.
		1443	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1444	* and hope for the best.
		1445	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1446	* assertion business is seriously broken, try yet another workaround
		1447	* for the broken -DDEBUGGING.
		1448	*/
		1449	{
		1450	#ifdef DEBUGGING
		1451	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1452	#else
		1453	STRLEN offset = SvCUR (string);
		1454	#endif
		1455
		1456	if (offset > json->max_size && json->max_size)
		1457	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1458	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1459	}
		1460
990	if (flags & F_UTF8)	1461	if (json->flags & F_UTF8)
991	sv_utf8_downgrade (string, 0);	1462	sv_utf8_downgrade (string, 0);
992	else	1463	else
993	sv_utf8_upgrade (string);	1464	sv_utf8_upgrade (string);
994		1465
995	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1466	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
996		1467
997	dec.flags = flags;	1468	dec.json = *json;
998	dec.cur = SvPVX (string);	1469	dec.cur = SvPVX (string);
999	dec.end = SvEND (string);	1470	dec.end = SvEND (string);
1000	dec.err = 0;	1471	dec.err = 0;
1001	dec.depth = 0;	1472	dec.depth = 0;
1002	dec.maxdepth = DEC_DEPTH (dec.flags);
1003		1473
		1474	if (dec.json.cb_object || dec.json.cb_sk_object)
		1475	dec.json.flags |= F_HOOK;
		1476
1004	*dec.end = 0; // this should basically be a nop, too, but make sure its there	1477	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1478
		1479	decode_ws (&dec);
1005	sv = decode_sv (&dec);	1480	sv = decode_sv (&dec);
1006		1481
		1482	if (offset_return)
		1483	*offset_return = dec.cur;
		1484
		1485	if (!(offset_return || !sv))
		1486	{
		1487	// check for trailing garbage
		1488	decode_ws (&dec);
		1489
		1490	if (*dec.cur)
		1491	{
		1492	dec.err = "garbage after JSON object";
		1493	SvREFCNT_dec (sv);
		1494	sv = 0;
		1495	}
		1496	}
		1497
1007	if (!sv)	1498	if (!sv)
1008	{	1499	{
1009	IV offset = dec.flags & F_UTF8
1010	? dec.cur - SvPVX (string)
1011	: utf8_distance (dec.cur, SvPVX (string));
1012	SV *uni = sv_newmortal ();	1500	SV *uni = sv_newmortal ();
1013		1501
1014	// horrible hack to silence warning inside pv_uni_display	1502	// horrible hack to silence warning inside pv_uni_display
1015	COP cop = *PL_curcop;	1503	COP cop = *PL_curcop;
1016	cop.cop_warnings = pWARN_NONE;	1504	cop.cop_warnings = pWARN_NONE;
…		…
1018	SAVEVPTR (PL_curcop);	1506	SAVEVPTR (PL_curcop);
1019	PL_curcop = &cop;	1507	PL_curcop = &cop;
1020	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1508	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1021	LEAVE;	1509	LEAVE;
1022		1510
1023	croak ("%s, at character offset %d [\"%s\"]",	1511	croak ("%s, at character offset %d (before \"%s\")",
1024	dec.err,	1512	dec.err,
1025	(int)offset,	1513	ptr_to_index (string, dec.cur),
1026	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1514	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1027	}	1515	}
1028		1516
1029	sv = sv_2mortal (sv);	1517	sv = sv_2mortal (sv);
1030		1518
1031	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1519	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1032	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1520	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1033		1521
1034	return sv;	1522	return sv;
		1523	}
		1524
		1525	/////////////////////////////////////////////////////////////////////////////
		1526	// incremental parser
		1527
		1528	static void
		1529	incr_parse (JSON *self)
		1530	{
		1531	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1532
		1533	// the state machine here is a bit convoluted and could be simplified a lot
		1534	// but this would make it slower, so...
		1535
		1536	for (;;)
		1537	{
		1538	//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
		1539	switch (self->incr_mode)
		1540	{
		1541	// only used for initial whitespace skipping
		1542	case INCR_M_WS:
		1543	for (;;)
		1544	{
		1545	if (*p > 0x20)
		1546	{
		1547	if (*p == '#')
		1548	{
		1549	self->incr_mode = INCR_M_C0;
		1550	goto incr_m_c;
		1551	}
		1552	else
		1553	{
		1554	self->incr_mode = INCR_M_JSON;
		1555	goto incr_m_json;
		1556	}
		1557	}
		1558	else if (!*p)
		1559	goto interrupt;
		1560
		1561	++p;
		1562	}
		1563
		1564	// skip a single char inside a string (for \\-processing)
		1565	case INCR_M_BS:
		1566	if (!*p)
		1567	goto interrupt;
		1568
		1569	++p;
		1570	self->incr_mode = INCR_M_STR;
		1571	goto incr_m_str;
		1572
		1573	// inside #-style comments
		1574	case INCR_M_C0:
		1575	case INCR_M_C1:
		1576	incr_m_c:
		1577	for (;;)
		1578	{
		1579	if (*p == '\n')
		1580	{
		1581	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1582	break;
		1583	}
		1584	else if (!*p)
		1585	goto interrupt;
		1586
		1587	++p;
		1588	}
		1589
		1590	break;
		1591
		1592	// inside a string
		1593	case INCR_M_STR:
		1594	incr_m_str:
		1595	for (;;)
		1596	{
		1597	if (*p == '"')
		1598	{
		1599	++p;
		1600	self->incr_mode = INCR_M_JSON;
		1601
		1602	if (!self->incr_nest)
		1603	goto interrupt;
		1604
		1605	goto incr_m_json;
		1606	}
		1607	else if (*p == '\\')
		1608	{
		1609	++p; // "virtually" consumes character after \
		1610
		1611	if (!*p) // if at end of string we have to switch modes
		1612	{
		1613	self->incr_mode = INCR_M_BS;
		1614	goto interrupt;
		1615	}
		1616	}
		1617	else if (!*p)
		1618	goto interrupt;
		1619
		1620	++p;
		1621	}
		1622
		1623	// after initial ws, outside string
		1624	case INCR_M_JSON:
		1625	incr_m_json:
		1626	for (;;)
		1627	{
		1628	switch (*p++)
		1629	{
		1630	case 0:
		1631	--p;
		1632	goto interrupt;
		1633
		1634	case 0x09:
		1635	case 0x0a:
		1636	case 0x0d:
		1637	case 0x20:
		1638	if (!self->incr_nest)
		1639	{
		1640	--p; // do not eat the whitespace, let the next round do it
		1641	goto interrupt;
		1642	}
		1643	break;
		1644
		1645	case '"':
		1646	self->incr_mode = INCR_M_STR;
		1647	goto incr_m_str;
		1648
		1649	case '[':
		1650	case '{':
		1651	if (++self->incr_nest > self->max_depth)
		1652	croak (ERR_NESTING_EXCEEDED);
		1653	break;
		1654
		1655	case ']':
		1656	case '}':
		1657	if (--self->incr_nest <= 0)
		1658	goto interrupt;
		1659	break;
		1660
		1661	case '#':
		1662	self->incr_mode = INCR_M_C1;
		1663	goto incr_m_c;
		1664	}
		1665	}
		1666	}
		1667
		1668	modechange:
		1669	;
		1670	}
		1671
		1672	interrupt:
		1673	self->incr_pos = p - SvPVX (self->incr_text);
		1674	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1675	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1035	}	1676	}
1036		1677
1037	/////////////////////////////////////////////////////////////////////////////	1678	/////////////////////////////////////////////////////////////////////////////
1038	// XS interface functions	1679	// XS interface functions
1039		1680
1040	MODULE = JSON::XS PACKAGE = JSON::XS	1681	MODULE = JSON::XS PACKAGE = JSON::XS
1041		1682
1042	BOOT:	1683	BOOT:
1043	{	1684	{
1044	int i;	1685	int i;
1045
1046	memset (decode_hexdigit, 0xff, 256);
1047		1686
1048	for (i = 0; i < 256; ++i)	1687	for (i = 0; i < 256; ++i)
1049	decode_hexdigit [i] =	1688	decode_hexdigit [i] =
1050	i >= '0' && i <= '9' ? i - '0'	1689	i >= '0' && i <= '9' ? i - '0'
1051	: i >= 'a' && i <= 'f' ? i - 'a' + 10	1690	: i >= 'a' && i <= 'f' ? i - 'a' + 10
1052	: i >= 'A' && i <= 'F' ? i - 'A' + 10	1691	: i >= 'A' && i <= 'F' ? i - 'A' + 10
1053	: -1;	1692	: -1;
1054		1693
1055	json_stash = gv_stashpv ("JSON::XS", 1);	1694	json_stash = gv_stashpv ("JSON::XS" , 1);
		1695	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1696
		1697	json_true = get_bool ("JSON::XS::true");
		1698	json_false = get_bool ("JSON::XS::false");
		1699
		1700	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1056	}	1701	}
1057		1702
1058	PROTOTYPES: DISABLE	1703	PROTOTYPES: DISABLE
1059		1704
1060	SV *new (char *dummy)	1705	void CLONE (...)
1061	CODE:	1706	CODE:
1062	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1707	json_stash = 0;
		1708	json_boolean_stash = 0;
		1709
		1710	void new (char *klass)
		1711	PPCODE:
		1712	{
		1713	SV *pv = NEWSV (0, sizeof (JSON));
		1714	SvPOK_only (pv);
		1715	json_init ((JSON *)SvPVX (pv));
		1716	XPUSHs (sv_2mortal (sv_bless (
		1717	newRV_noinc (pv),
		1718	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1719	)));
		1720	}
		1721
		1722	void ascii (JSON *self, int enable = 1)
		1723	ALIAS:
		1724	ascii = F_ASCII
		1725	latin1 = F_LATIN1
		1726	utf8 = F_UTF8
		1727	indent = F_INDENT
		1728	canonical = F_CANONICAL
		1729	space_before = F_SPACE_BEFORE
		1730	space_after = F_SPACE_AFTER
		1731	pretty = F_PRETTY
		1732	allow_nonref = F_ALLOW_NONREF
		1733	shrink = F_SHRINK
		1734	allow_blessed = F_ALLOW_BLESSED
		1735	convert_blessed = F_CONV_BLESSED
		1736	relaxed = F_RELAXED
		1737	allow_unknown = F_ALLOW_UNKNOWN
		1738	PPCODE:
		1739	{
		1740	if (enable)
		1741	self->flags |= ix;
		1742	else
		1743	self->flags &= ~ix;
		1744
		1745	XPUSHs (ST (0));
		1746	}
		1747
		1748	void get_ascii (JSON *self)
		1749	ALIAS:
		1750	get_ascii = F_ASCII
		1751	get_latin1 = F_LATIN1
		1752	get_utf8 = F_UTF8
		1753	get_indent = F_INDENT
		1754	get_canonical = F_CANONICAL
		1755	get_space_before = F_SPACE_BEFORE
		1756	get_space_after = F_SPACE_AFTER
		1757	get_allow_nonref = F_ALLOW_NONREF
		1758	get_shrink = F_SHRINK
		1759	get_allow_blessed = F_ALLOW_BLESSED
		1760	get_convert_blessed = F_CONV_BLESSED
		1761	get_relaxed = F_RELAXED
		1762	get_allow_unknown = F_ALLOW_UNKNOWN
		1763	PPCODE:
		1764	XPUSHs (boolSV (self->flags & ix));
		1765
		1766	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		1767	PPCODE:
		1768	self->max_depth = max_depth;
		1769	XPUSHs (ST (0));
		1770
		1771	U32 get_max_depth (JSON *self)
		1772	CODE:
		1773	RETVAL = self->max_depth;
1063	OUTPUT:	1774	OUTPUT:
1064	RETVAL	1775	RETVAL
1065		1776
1066	SV *ascii (SV *self, int enable = 1)	1777	void max_size (JSON *self, U32 max_size = 0)
1067	ALIAS:	1778	PPCODE:
1068	ascii = F_ASCII	1779	self->max_size = max_size;
1069	utf8 = F_UTF8	1780	XPUSHs (ST (0));
1070	indent = F_INDENT	1781
1071	canonical = F_CANONICAL	1782	int get_max_size (JSON *self)
1072	space_before = F_SPACE_BEFORE
1073	space_after = F_SPACE_AFTER
1074	pretty = F_PRETTY
1075	allow_nonref = F_ALLOW_NONREF
1076	shrink = F_SHRINK
1077	CODE:	1783	CODE:
1078	{	1784	RETVAL = self->max_size;
1079	UV *uv = SvJSON (self);
1080	if (enable)
1081	*uv |= ix;
1082	else
1083	*uv &= ~ix;
1084
1085	RETVAL = newSVsv (self);
1086	}
1087	OUTPUT:	1785	OUTPUT:
1088	RETVAL	1786	RETVAL
1089		1787
1090	SV *max_depth (SV *self, UV max_depth = 0x80000000UL)	1788	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
		1789	PPCODE:
		1790	{
		1791	SvREFCNT_dec (self->cb_object);
		1792	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1793
		1794	XPUSHs (ST (0));
		1795	}
		1796
		1797	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
		1798	PPCODE:
		1799	{
		1800	if (!self->cb_sk_object)
		1801	self->cb_sk_object = newHV ();
		1802
		1803	if (SvOK (cb))
		1804	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1805	else
		1806	{
		1807	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1808
		1809	if (!HvKEYS (self->cb_sk_object))
		1810	{
		1811	SvREFCNT_dec (self->cb_sk_object);
		1812	self->cb_sk_object = 0;
		1813	}
		1814	}
		1815
		1816	XPUSHs (ST (0));
		1817	}
		1818
		1819	void encode (JSON *self, SV *scalar)
		1820	PPCODE:
		1821	XPUSHs (encode_json (scalar, self));
		1822
		1823	void decode (JSON *self, SV *jsonstr)
		1824	PPCODE:
		1825	XPUSHs (decode_json (jsonstr, self, 0));
		1826
		1827	void decode_prefix (JSON *self, SV *jsonstr)
		1828	PPCODE:
		1829	{
		1830	char *offset;
		1831	EXTEND (SP, 2);
		1832	PUSHs (decode_json (jsonstr, self, &offset));
		1833	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		1834	}
		1835
		1836	void incr_parse (JSON *self, SV *jsonstr = 0)
		1837	PPCODE:
		1838	{
		1839	if (!self->incr_text)
		1840	self->incr_text = newSVpvn ("", 0);
		1841
		1842	// append data, if any
		1843	if (jsonstr)
		1844	{
		1845	if (SvUTF8 (jsonstr))
		1846	{
		1847	if (!SvUTF8 (self->incr_text))
		1848	{
		1849	/* utf-8-ness differs, need to upgrade */
		1850	sv_utf8_upgrade (self->incr_text);
		1851
		1852	if (self->incr_pos)
		1853	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1854	- (U8 *)SvPVX (self->incr_text);
		1855	}
		1856	}
		1857	else if (SvUTF8 (self->incr_text))
		1858	sv_utf8_upgrade (jsonstr);
		1859
		1860	{
		1861	STRLEN len;
		1862	const char *str = SvPV (jsonstr, len);
		1863	STRLEN cur = SvCUR (self->incr_text);
		1864
		1865	if (SvLEN (self->incr_text) <= cur + len)
		1866	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1867
		1868	Move (str, SvEND (self->incr_text), len, char);
		1869	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1870	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1871	}
		1872	}
		1873
		1874	if (GIMME_V != G_VOID)
		1875	do
		1876	{
		1877	char *offset;
		1878
		1879	if (!INCR_DONE (self))
		1880	{
		1881	incr_parse (self);
		1882
		1883	if (self->incr_pos > self->max_size && self->max_size)
		1884	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1885	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1886
		1887	if (!INCR_DONE (self))
		1888	break;
		1889	}
		1890
		1891	XPUSHs (decode_json (self->incr_text, self, &offset));
		1892
		1893	self->incr_pos -= offset - SvPVX (self->incr_text);
		1894	self->incr_nest = 0;
		1895	self->incr_mode = 0;
		1896
		1897	sv_chop (self->incr_text, offset);
		1898	}
		1899	while (GIMME_V == G_ARRAY);
		1900	}
		1901
		1902	SV *incr_text (JSON *self)
		1903	ATTRS: lvalue
1091	CODE:	1904	CODE:
1092	{	1905	{
1093	UV *uv = SvJSON (self);	1906	if (self->incr_pos)
1094	UV log2 = 0;	1907	croak ("incr_text can not be called when the incremental parser already started parsing");
1095		1908
1096	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	1909	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
1097
1098	while ((1UL << log2) < max_depth)
1099	++log2;
1100
1101	*uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1102
1103	RETVAL = newSVsv (self);
1104	}	1910	}
1105	OUTPUT:	1911	OUTPUT:
1106	RETVAL	1912	RETVAL
1107		1913
		1914	void incr_skip (JSON *self)
		1915	CODE:
		1916	{
		1917	if (self->incr_pos)
		1918	{
		1919	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1920	self->incr_pos = 0;
		1921	self->incr_nest = 0;
		1922	self->incr_mode = 0;
		1923	}
		1924	}
		1925
		1926	void incr_reset (JSON *self)
		1927	CODE:
		1928	{
		1929	SvREFCNT_dec (self->incr_text);
		1930	self->incr_text = 0;
		1931	self->incr_pos = 0;
		1932	self->incr_nest = 0;
		1933	self->incr_mode = 0;
		1934	}
		1935
		1936	void DESTROY (JSON *self)
		1937	CODE:
		1938	SvREFCNT_dec (self->cb_sk_object);
		1939	SvREFCNT_dec (self->cb_object);
		1940	SvREFCNT_dec (self->incr_text);
		1941
		1942	PROTOTYPES: ENABLE
		1943
1108	void encode (SV *self, SV *scalar)	1944	void encode_json (SV *scalar)
		1945	ALIAS:
		1946	to_json_ = 0
		1947	encode_json = F_UTF8
1109	PPCODE:	1948	PPCODE:
		1949	{
		1950	JSON json;
		1951	json_init (&json);
		1952	json.flags |= ix;
1110	XPUSHs (encode_json (scalar, *SvJSON (self)));	1953	XPUSHs (encode_json (scalar, &json));
		1954	}
1111		1955
1112	void decode (SV *self, SV *jsonstr)	1956	void decode_json (SV *jsonstr)
		1957	ALIAS:
		1958	from_json_ = 0
		1959	decode_json = F_UTF8
1113	PPCODE:	1960	PPCODE:
		1961	{
		1962	JSON json;
		1963	json_init (&json);
		1964	json.flags |= ix;
1114	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1965	XPUSHs (decode_json (jsonstr, &json, 0));
		1966	}
1115		1967
1116	PROTOTYPES: ENABLE
1117		1968
1118	void to_json (SV *scalar)
1119	ALIAS:
1120	objToJson = 0
1121	PPCODE:
1122	XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8));
1123
1124	void from_json (SV *jsonstr)
1125	ALIAS:
1126	jsonToObj = 0
1127	PPCODE:
1128	XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8));
1129

Diff Legend

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