ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.17 by root, Sun Mar 25 02:46:41 2007 UTC vs.
Revision 1.109 by root, Tue Mar 8 10:18:50 2011 UTC

1	#include "EXTERN.h"	1	#include "EXTERN.h"
2	#include "perl.h"	2	#include "perl.h"
3	#include "XSUB.h"	3	#include "XSUB.h"
4		4
5	#include "assert.h"	5	#include <assert.h>
6	#include "string.h"	6	#include <string.h>
7	#include "stdlib.h"	7	#include <stdlib.h>
		8	#include <stdio.h>
		9	#include <limits.h>
		10	#include <float.h>
8		11
		12	#if defined(__BORLANDC__) || defined(_MSC_VER)
		13	# define snprintf _snprintf // C compilers have this in stdio.h
		14	#endif
		15
		16	// some old perls do not have this, try to make it work, no
		17	// guarantees, though. if it breaks, you get to keep the pieces.
		18	#ifndef UTF8_MAXBYTES
		19	# define UTF8_MAXBYTES 13
		20	#endif
		21
		22	// three extra for rounding, sign, and end of string
		23	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
		24
9	#define F_ASCII 0x00000001	25	#define F_ASCII 0x00000001UL
		26	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	27	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	28	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	29	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	30	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	31	#define F_SPACE_AFTER 0x00000040UL
15	#define F_ALLOW_NONREF 0x00000080	32	#define F_ALLOW_NONREF 0x00000100UL
16	#define F_SHRINK 0x00000100	33	#define F_SHRINK 0x00000200UL
17		34	#define F_ALLOW_BLESSED 0x00000400UL
18	// F_SKIPINVALID?	35	#define F_CONV_BLESSED 0x00000800UL
19	// F_EXECCODEREF?	36	#define F_RELAXED 0x00001000UL
20	// F_SELFCONVERT? <=> { &__class__ => }	37	#define F_ALLOW_UNKNOWN 0x00002000UL
		38	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
21		39
22	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	40	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
23	#define F_DEFAULT 0
24		41
25	#define INIT_SIZE 32 // initial scalar size to be allocated	42	#define INIT_SIZE 32 // initial scalar size to be allocated
26	#define INDENT_STEP 3 // spaces per indentation level	43	#define INDENT_STEP 3 // spaces per indentation level
27		44
28	#define UTF8_MAX_LEN 11 // for perls UTF-X: max. number of octets per character
29	#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
		46
		47	#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
30		48
31	#define SB do {	49	#define SB do {
32	#define SE } while (0)	50	#define SE } while (0)
33		51
		52	#if __GNUC__ >= 3
		53	# define expect(expr,value) __builtin_expect ((expr), (value))
		54	# define INLINE static inline
		55	#else
		56	# define expect(expr,value) (expr)
		57	# define INLINE static
		58	#endif
		59
		60	#define expect_false(expr) expect ((expr) != 0, 0)
		61	#define expect_true(expr) expect ((expr) != 0, 1)
		62
		63	#define IN_RANGE_INC(type,val,beg,end) \
		64	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		65	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		66
		67	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
		68
		69	#ifdef USE_ITHREADS
		70	# define JSON_SLOW 1
		71	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		72	#else
		73	# define JSON_SLOW 0
		74	# define JSON_STASH json_stash
		75	#endif
		76
34	static HV *json_stash; // JSON::XS::	77	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		78	static SV *json_true, *json_false;
		79
		80	enum {
		81	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		82	INCR_M_STR, // inside string
		83	INCR_M_BS, // inside backslash
		84	INCR_M_C0, // inside comment in initial whitespace sequence
		85	INCR_M_C1, // inside comment in other places
		86	INCR_M_JSON // outside anything, count nesting
		87	};
		88
		89	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		90
		91	typedef struct {
		92	U32 flags;
		93	U32 max_depth;
		94	STRLEN max_size;
		95
		96	SV *cb_object;
		97	HV *cb_sk_object;
		98
		99	// for the incremental parser
		100	SV *incr_text; // the source text so far
		101	STRLEN incr_pos; // the current offset into the text
		102	int incr_nest; // {[]}-nesting level
		103	unsigned char incr_mode;
		104	} JSON;
		105
		106	INLINE void
		107	json_init (JSON *json)
		108	{
		109	Zero (json, 1, JSON);
		110	json->max_depth = 512;
		111	}
35		112
36	/////////////////////////////////////////////////////////////////////////////	113	/////////////////////////////////////////////////////////////////////////////
37	// utility functions	114	// utility functions
38		115
39	static UV *	116	INLINE SV *
40	SvJSON (SV *sv)	117	get_bool (const char *name)
41	{	118	{
42	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))	119	SV *sv = get_sv (name, 1);
43	croak ("object is not of type JSON::XS");
44		120
45	return &SvUVX (SvRV (sv));	121	SvREADONLY_on (sv);
46	}	122	SvREADONLY_on (SvRV (sv));
47		123
48	static void	124	return sv;
		125	}
		126
		127	INLINE void
49	shrink (SV *sv)	128	shrink (SV *sv)
50	{	129	{
51	sv_utf8_downgrade (sv, 1);	130	sv_utf8_downgrade (sv, 1);
		131
52	if (SvLEN (sv) > SvCUR (sv) + 1)	132	if (SvLEN (sv) > SvCUR (sv) + 1)
53	{	133	{
54	#ifdef SvPV_shrink_to_cur	134	#ifdef SvPV_shrink_to_cur
55	SvPV_shrink_to_cur (sv);	135	SvPV_shrink_to_cur (sv);
56	#elif defined (SvPV_renew)	136	#elif defined (SvPV_renew)
…		…
62	// decode an utf-8 character and return it, or (UV)-1 in	142	// decode an utf-8 character and return it, or (UV)-1 in
63	// case of an error.	143	// case of an error.
64	// we special-case "safe" characters from U+80 .. U+7FF,	144	// we special-case "safe" characters from U+80 .. U+7FF,
65	// but use the very good perl function to parse anything else.	145	// but use the very good perl function to parse anything else.
66	// note that we never call this function for a ascii codepoints	146	// note that we never call this function for a ascii codepoints
67	static UV	147	INLINE UV
68	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)	148	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
69	{	149	{
70	if (s[0] > 0xdf || s[0] < 0xc2)	150	if (expect_true (len >= 2
71	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	151	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
72	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	152	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
73	{	153	{
74	*clen = 2;	154	*clen = 2;
75	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);	155	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
76	}	156	}
77	else	157	else
78	return (UV)-1;	158	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
79	}	159	}
80		160
		161	// likewise for encoding, also never called for ascii codepoints
		162	// this function takes advantage of this fact, although current gccs
		163	// seem to optimise the check for >= 0x80 away anyways
		164	INLINE unsigned char *
		165	encode_utf8 (unsigned char *s, UV ch)
		166	{
		167	if (expect_false (ch < 0x000080))
		168	*s++ = ch;
		169	else if (expect_true (ch < 0x000800))
		170	*s++ = 0xc0 | ( ch >> 6),
		171	*s++ = 0x80 | ( ch & 0x3f);
		172	else if ( ch < 0x010000)
		173	*s++ = 0xe0 | ( ch >> 12),
		174	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		175	*s++ = 0x80 | ( ch & 0x3f);
		176	else if ( ch < 0x110000)
		177	*s++ = 0xf0 | ( ch >> 18),
		178	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		179	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		180	*s++ = 0x80 | ( ch & 0x3f);
		181
		182	return s;
		183	}
		184
		185	// convert offset pointer to character index, sv must be string
		186	static STRLEN
		187	ptr_to_index (SV *sv, char *offset)
		188	{
		189	return SvUTF8 (sv)
		190	? utf8_distance (offset, SvPVX (sv))
		191	: offset - SvPVX (sv);
		192	}
		193
		194	/////////////////////////////////////////////////////////////////////////////
		195	// fp hell
		196
		197	// scan a group of digits, and a trailing exponent
		198	static void
		199	json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
		200	{
		201	UV uaccum = 0;
		202	int eaccum = 0;
		203
		204	// if we recurse too deep, skip all remaining digits
		205	// to avoid a stack overflow attack
		206	if (expect_false (--maxdepth <= 0))
		207	while (((U8)*s - '0') < 10)
		208	++s;
		209
		210	for (;;)
		211	{
		212	U8 dig = (U8)*s - '0';
		213
		214	if (expect_false (dig >= 10))
		215	{
		216	if (dig == (U8)((U8)'.' - (U8)'0'))
		217	{
		218	++s;
		219	json_atof_scan1 (s, accum, expo, 1, maxdepth);
		220	}
		221	else if ((dig | ' ') == 'e' - '0')
		222	{
		223	int exp2 = 0;
		224	int neg = 0;
		225
		226	++s;
		227
		228	if (*s == '-')
		229	{
		230	++s;
		231	neg = 1;
		232	}
		233	else if (*s == '+')
		234	++s;
		235
		236	while ((dig = (U8)*s - '0') < 10)
		237	exp2 = exp2 * 10 + *s++ - '0';
		238
		239	*expo += neg ? -exp2 : exp2;
		240	}
		241
		242	break;
		243	}
		244
		245	++s;
		246
		247	uaccum = uaccum * 10 + dig;
		248	++eaccum;
		249
		250	// if we have too many digits, then recurse for more
		251	// we actually do this for rather few digits
		252	if (uaccum >= (UV_MAX - 9) / 10)
		253	{
		254	if (postdp) *expo -= eaccum;
		255	json_atof_scan1 (s, accum, expo, postdp, maxdepth);
		256	if (postdp) *expo += eaccum;
		257
		258	break;
		259	}
		260	}
		261
		262	// this relies greatly on the quality of the pow ()
		263	// implementation of the platform, but a good
		264	// implementation is hard to beat.
		265	if (postdp) *expo -= eaccum;
		266	*accum += uaccum * Perl_pow (10., *expo);
		267	*expo += eaccum;
		268	}
		269
		270	static NV
		271	json_atof (const char *s)
		272	{
		273	NV accum = 0.;
		274	int expo = 0;
		275	int neg = 0;
		276
		277	if (*s == '-')
		278	{
		279	++s;
		280	neg = 1;
		281	}
		282
		283	// a recursion depth of ten gives us >>500 bits
		284	json_atof_scan1 (s, &accum, &expo, 0, 10);
		285
		286	return neg ? -accum : accum;
		287	}
81	/////////////////////////////////////////////////////////////////////////////	288	/////////////////////////////////////////////////////////////////////////////
82	// encoder	289	// encoder
83		290
84	// structure used for encoding JSON	291	// structure used for encoding JSON
85	typedef struct	292	typedef struct
86	{	293	{
87	char *cur; // SvPVX (sv) + current output position	294	char *cur; // SvPVX (sv) + current output position
88	char *end; // SvEND (sv)	295	char *end; // SvEND (sv)
89	SV *sv; // result scalar	296	SV *sv; // result scalar
90	UV flags; // F_*	297	JSON json;
91	int indent; // indentation level	298	U32 indent; // indentation level
92	int max_depth; // max. recursion level	299	UV limit; // escape character values >= this value when encoding
93	} enc_t;	300	} enc_t;
94		301
95	static void	302	INLINE void
96	need (enc_t *enc, STRLEN len)	303	need (enc_t *enc, STRLEN len)
97	{	304	{
98	if (enc->cur + len >= enc->end)	305	if (expect_false (enc->cur + len >= enc->end))
99	{	306	{
100	STRLEN cur = enc->cur - SvPVX (enc->sv);	307	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
101	SvGROW (enc->sv, cur + len + 1);	308	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
102	enc->cur = SvPVX (enc->sv) + cur;	309	enc->cur = SvPVX (enc->sv) + cur;
103	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	310	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
104	}	311	}
105	}	312	}
106		313
107	static void	314	INLINE void
108	encode_ch (enc_t *enc, char ch)	315	encode_ch (enc_t *enc, char ch)
109	{	316	{
110	need (enc, 1);	317	need (enc, 1);
111	*enc->cur++ = ch;	318	*enc->cur++ = ch;
112	}	319	}
…		…
120		327
121	while (str < end)	328	while (str < end)
122	{	329	{
123	unsigned char ch = *(unsigned char *)str;	330	unsigned char ch = *(unsigned char *)str;
124		331
125	if (ch >= 0x20 && ch < 0x80) // most common case	332	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
126	{	333	{
127	if (ch == '"') // but with slow exceptions	334	if (expect_false (ch == '"')) // but with slow exceptions
128	{	335	{
129	need (enc, len += 1);	336	need (enc, len += 1);
130	*enc->cur++ = '\\';	337	*enc->cur++ = '\\';
131	*enc->cur++ = '"';	338	*enc->cur++ = '"';
132	}	339	}
133	else if (ch == '\\')	340	else if (expect_false (ch == '\\'))
134	{	341	{
135	need (enc, len += 1);	342	need (enc, len += 1);
136	*enc->cur++ = '\\';	343	*enc->cur++ = '\\';
137	*enc->cur++ = '\\';	344	*enc->cur++ = '\\';
138	}	345	}
…		…
156	STRLEN clen;	363	STRLEN clen;
157	UV uch;	364	UV uch;
158		365
159	if (is_utf8)	366	if (is_utf8)
160	{	367	{
161	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
162	uch = decode_utf8 (str, end - str, &clen);	368	uch = decode_utf8 (str, end - str, &clen);
163	if (clen == (STRLEN)-1)	369	if (clen == (STRLEN)-1)
164	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	370	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
165	}	371	}
166	else	372	else
167	{	373	{
168	uch = ch;	374	uch = ch;
169	clen = 1;	375	clen = 1;
170	}	376	}
171		377
172	if (uch > 0x10FFFFUL)	378	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
173	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
174
175	if (uch < 0x80 || enc->flags & F_ASCII)
176	{	379	{
177	if (uch > 0xFFFFUL)	380	if (uch >= 0x10000UL)
178	{	381	{
		382	if (uch >= 0x110000UL)
		383	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		384
179	need (enc, len += 11);	385	need (enc, len += 11);
180	sprintf (enc->cur, "\\u%04x\\u%04x",	386	sprintf (enc->cur, "\\u%04x\\u%04x",
181	(int)((uch - 0x10000) / 0x400 + 0xD800),	387	(int)((uch - 0x10000) / 0x400 + 0xD800),
182	(int)((uch - 0x10000) % 0x400 + 0xDC00));	388	(int)((uch - 0x10000) % 0x400 + 0xDC00));
183	enc->cur += 12;	389	enc->cur += 12;
184	}	390	}
185	else	391	else
186	{	392	{
187	static char hexdigit [16] = "0123456789abcdef";
188	need (enc, len += 5);	393	need (enc, len += 5);
189	*enc->cur++ = '\\';	394	*enc->cur++ = '\\';
190	*enc->cur++ = 'u';	395	*enc->cur++ = 'u';
191	*enc->cur++ = hexdigit [ uch >> 12 ];	396	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
192	*enc->cur++ = hexdigit [(uch >> 8) & 15];	397	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
193	*enc->cur++ = hexdigit [(uch >> 4) & 15];	398	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
194	*enc->cur++ = hexdigit [(uch >> 0) & 15];	399	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
195	}	400	}
196		401
		402	str += clen;
		403	}
		404	else if (enc->json.flags & F_LATIN1)
		405	{
		406	*enc->cur++ = uch;
197	str += clen;	407	str += clen;
198	}	408	}
199	else if (is_utf8)	409	else if (is_utf8)
200	{	410	{
201	need (enc, len += clen);	411	need (enc, len += clen);
…		…
205	}	415	}
206	while (--clen);	416	while (--clen);
207	}	417	}
208	else	418	else
209	{	419	{
210	need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed	420	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
211	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	421	enc->cur = encode_utf8 (enc->cur, uch);
212	++str;	422	++str;
213	}	423	}
214	}	424	}
215	}	425	}
216	}	426	}
217		427
218	--len;	428	--len;
219	}	429	}
220	}	430	}
221		431
222	static void	432	INLINE void
223	encode_indent (enc_t *enc)	433	encode_indent (enc_t *enc)
224	{	434	{
225	if (enc->flags & F_INDENT)	435	if (enc->json.flags & F_INDENT)
226	{	436	{
227	int spaces = enc->indent * INDENT_STEP;	437	int spaces = enc->indent * INDENT_STEP;
228		438
229	need (enc, spaces);	439	need (enc, spaces);
230	memset (enc->cur, ' ', spaces);	440	memset (enc->cur, ' ', spaces);
231	enc->cur += spaces;	441	enc->cur += spaces;
232	}	442	}
233	}	443	}
234		444
235	static void	445	INLINE void
236	encode_space (enc_t *enc)	446	encode_space (enc_t *enc)
237	{	447	{
238	need (enc, 1);	448	need (enc, 1);
239	encode_ch (enc, ' ');	449	encode_ch (enc, ' ');
240	}	450	}
241		451
242	static void	452	INLINE void
243	encode_nl (enc_t *enc)	453	encode_nl (enc_t *enc)
244	{	454	{
245	if (enc->flags & F_INDENT)	455	if (enc->json.flags & F_INDENT)
246	{	456	{
247	need (enc, 1);	457	need (enc, 1);
248	encode_ch (enc, '\n');	458	encode_ch (enc, '\n');
249	}	459	}
250	}	460	}
251		461
252	static void	462	INLINE void
253	encode_comma (enc_t *enc)	463	encode_comma (enc_t *enc)
254	{	464	{
255	encode_ch (enc, ',');	465	encode_ch (enc, ',');
256		466
257	if (enc->flags & F_INDENT)	467	if (enc->json.flags & F_INDENT)
258	encode_nl (enc);	468	encode_nl (enc);
259	else if (enc->flags & F_SPACE_AFTER)	469	else if (enc->json.flags & F_SPACE_AFTER)
260	encode_space (enc);	470	encode_space (enc);
261	}	471	}
262		472
263	static void encode_sv (enc_t *enc, SV *sv);	473	static void encode_sv (enc_t *enc, SV *sv);
264		474
265	static void	475	static void
266	encode_av (enc_t *enc, AV *av)	476	encode_av (enc_t *enc, AV *av)
267	{	477	{
268	int i, len = av_len (av);	478	int i, len = av_len (av);
269		479
270	encode_ch (enc, '['); encode_nl (enc);	480	if (enc->indent >= enc->json.max_depth)
271	++enc->indent;	481	croak (ERR_NESTING_EXCEEDED);
272		482
		483	encode_ch (enc, '[');
		484
		485	if (len >= 0)
		486	{
		487	encode_nl (enc); ++enc->indent;
		488
273	for (i = 0; i <= len; ++i)	489	for (i = 0; i <= len; ++i)
274	{	490	{
		491	SV **svp = av_fetch (av, i, 0);
		492
275	encode_indent (enc);	493	encode_indent (enc);
276	encode_sv (enc, *av_fetch (av, i, 0));
277		494
		495	if (svp)
		496	encode_sv (enc, *svp);
		497	else
		498	encode_str (enc, "null", 4, 0);
		499
278	if (i < len)	500	if (i < len)
279	encode_comma (enc);	501	encode_comma (enc);
280	}	502	}
281		503
		504	encode_nl (enc); --enc->indent; encode_indent (enc);
		505	}
		506
282	encode_nl (enc);	507	encode_ch (enc, ']');
283
284	--enc->indent;
285	encode_indent (enc); encode_ch (enc, ']');
286	}	508	}
287		509
288	static void	510	static void
289	encode_he (enc_t *enc, HE *he)	511	encode_hk (enc_t *enc, HE *he)
290	{	512	{
291	encode_ch (enc, '"');	513	encode_ch (enc, '"');
292		514
293	if (HeKLEN (he) == HEf_SVKEY)	515	if (HeKLEN (he) == HEf_SVKEY)
294	{	516	{
…		…
304	else	526	else
305	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	527	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
306		528
307	encode_ch (enc, '"');	529	encode_ch (enc, '"');
308		530
309	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	531	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
310	encode_ch (enc, ':');	532	encode_ch (enc, ':');
311	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	533	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
312	encode_sv (enc, HeVAL (he));
313	}	534	}
314		535
315	// compare hash entries, used when all keys are bytestrings	536	// compare hash entries, used when all keys are bytestrings
316	static int	537	static int
317	he_cmp_fast (const void *a_, const void *b_)	538	he_cmp_fast (const void *a_, const void *b_)
…		…
322	HE *b = *(HE **)b_;	543	HE *b = *(HE **)b_;
323		544
324	STRLEN la = HeKLEN (a);	545	STRLEN la = HeKLEN (a);
325	STRLEN lb = HeKLEN (b);	546	STRLEN lb = HeKLEN (b);
326		547
327	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	548	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
328	cmp = la - lb;	549	cmp = lb - la;
329		550
330	return cmp;	551	return cmp;
331	}	552	}
332		553
333	// compare hash entries, used when some keys are sv's or utf-x	554	// compare hash entries, used when some keys are sv's or utf-x
334	static int	555	static int
335	he_cmp_slow (const void *a, const void *b)	556	he_cmp_slow (const void *a, const void *b)
336	{	557	{
337	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	558	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
338	}	559	}
339		560
340	static void	561	static void
341	encode_hv (enc_t *enc, HV *hv)	562	encode_hv (enc_t *enc, HV *hv)
342	{	563	{
343	int count, i;	564	HE *he;
344		565
345	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	566	if (enc->indent >= enc->json.max_depth)
		567	croak (ERR_NESTING_EXCEEDED);
346		568
347	if ((count = hv_iterinit (hv)))	569	encode_ch (enc, '{');
348	{	570
349	// for canonical output we have to sort by keys first	571	// for canonical output we have to sort by keys first
350	// actually, this is mostly due to the stupid so-called	572	// actually, this is mostly due to the stupid so-called
351	// security workaround added somewhere in 5.8.x.	573	// security workaround added somewhere in 5.8.x
352	// that randomises hash orderings	574	// that randomises hash orderings
353	if (enc->flags & F_CANONICAL)	575	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
		576	{
		577	int count = hv_iterinit (hv);
		578
		579	if (SvMAGICAL (hv))
354	{	580	{
		581	// need to count by iterating. could improve by dynamically building the vector below
		582	// but I don't care for the speed of this special case.
		583	// note also that we will run into undefined behaviour when the two iterations
		584	// do not result in the same count, something I might care for in some later release.
		585
		586	count = 0;
		587	while (hv_iternext (hv))
		588	++count;
		589
		590	hv_iterinit (hv);
		591	}
		592
		593	if (count)
		594	{
		595	int i, fast = 1;
		596	#if defined(__BORLANDC__) || defined(_MSC_VER)
		597	HE **hes = _alloca (count * sizeof (HE));
		598	#else
355	HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode	599	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
356	int fast = 1;	600	#endif
357		601
358	i = 0;	602	i = 0;
359	while ((he = hv_iternext (hv)))	603	while ((he = hv_iternext (hv)))
360	{	604	{
361	hes [i++] = he;	605	hes [i++] = he;
…		…
383		627
384	FREETMPS;	628	FREETMPS;
385	LEAVE;	629	LEAVE;
386	}	630	}
387		631
388	for (i = 0; i < count; ++i)	632	encode_nl (enc); ++enc->indent;
		633
		634	while (count--)
389	{	635	{
390	encode_indent (enc);	636	encode_indent (enc);
		637	he = hes [count];
391	encode_he (enc, hes [i]);	638	encode_hk (enc, he);
		639	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
392		640
393	if (i < count - 1)	641	if (count)
394	encode_comma (enc);	642	encode_comma (enc);
395	}	643	}
396		644
		645	encode_nl (enc); --enc->indent; encode_indent (enc);
		646	}
		647	}
		648	else
		649	{
		650	if (hv_iterinit (hv) || SvMAGICAL (hv))
		651	if ((he = hv_iternext (hv)))
		652	{
		653	encode_nl (enc); ++enc->indent;
		654
		655	for (;;)
		656	{
397	encode_nl (enc);	657	encode_indent (enc);
		658	encode_hk (enc, he);
		659	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		660
		661	if (!(he = hv_iternext (hv)))
		662	break;
		663
		664	encode_comma (enc);
		665	}
		666
		667	encode_nl (enc); --enc->indent; encode_indent (enc);
		668	}
		669	}
		670
		671	encode_ch (enc, '}');
		672	}
		673
		674	// encode objects, arrays and special \0=false and \1=true values.
		675	static void
		676	encode_rv (enc_t *enc, SV *sv)
		677	{
		678	svtype svt;
		679
		680	SvGETMAGIC (sv);
		681	svt = SvTYPE (sv);
		682
		683	if (expect_false (SvOBJECT (sv)))
		684	{
		685	HV *stash = !JSON_SLOW || json_boolean_stash
		686	? json_boolean_stash
		687	: gv_stashpv ("JSON::XS::Boolean", 1);
		688
		689	if (SvSTASH (sv) == stash)
		690	{
		691	if (SvIV (sv))
		692	encode_str (enc, "true", 4, 0);
		693	else
		694	encode_str (enc, "false", 5, 0);
398	}	695	}
399	else	696	else
400	{	697	{
401	SV *sv;	698	#if 0
402	HE *he = hv_iternext (hv);	699	if (0 && sv_derived_from (rv, "JSON::Literal"))
403
404	for (;;)
405	{
406	encode_indent (enc);
407	encode_he (enc, he);
408
409	if (!(he = hv_iternext (hv)))
410	break;
411
412	encode_comma (enc);
413	}	700	{
		701	// not yet
		702	}
		703	#endif
		704	if (enc->json.flags & F_CONV_BLESSED)
		705	{
		706	// we re-bless the reference to get overload and other niceties right
		707	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
414		708
		709	if (to_json)
		710	{
		711	dSP;
		712
		713	ENTER; SAVETMPS; PUSHMARK (SP);
		714	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		715
		716	// calling with G_SCALAR ensures that we always get a 1 return value
		717	PUTBACK;
		718	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		719	SPAGAIN;
		720
		721	// catch this surprisingly common error
		722	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		723	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		724
		725	sv = POPs;
		726	PUTBACK;
		727
415	encode_nl (enc);	728	encode_sv (enc, sv);
		729
		730	FREETMPS; LEAVE;
		731	}
		732	else if (enc->json.flags & F_ALLOW_BLESSED)
		733	encode_str (enc, "null", 4, 0);
		734	else
		735	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		736	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		737	}
		738	else if (enc->json.flags & F_ALLOW_BLESSED)
		739	encode_str (enc, "null", 4, 0);
		740	else
		741	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		742	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
416	}	743	}
417	}	744	}
		745	else if (svt == SVt_PVHV)
		746	encode_hv (enc, (HV *)sv);
		747	else if (svt == SVt_PVAV)
		748	encode_av (enc, (AV *)sv);
		749	else if (svt < SVt_PVAV)
		750	{
		751	STRLEN len = 0;
		752	char *pv = svt ? SvPV (sv, len) : 0;
418		753
419	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	754	if (len == 1 && *pv == '1')
		755	encode_str (enc, "true", 4, 0);
		756	else if (len == 1 && *pv == '0')
		757	encode_str (enc, "false", 5, 0);
		758	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		759	encode_str (enc, "null", 4, 0);
		760	else
		761	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		762	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		763	}
		764	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		765	encode_str (enc, "null", 4, 0);
		766	else
		767	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		768	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
420	}	769	}
421		770
422	static void	771	static void
423	encode_sv (enc_t *enc, SV *sv)	772	encode_sv (enc_t *enc, SV *sv)
424	{	773	{
…		…
432	encode_str (enc, str, len, SvUTF8 (sv));	781	encode_str (enc, str, len, SvUTF8 (sv));
433	encode_ch (enc, '"');	782	encode_ch (enc, '"');
434	}	783	}
435	else if (SvNOKp (sv))	784	else if (SvNOKp (sv))
436	{	785	{
		786	// trust that perl will do the right thing w.r.t. JSON syntax.
437	need (enc, NV_DIG + 32);	787	need (enc, NV_DIG + 32);
438	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	788	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
439	enc->cur += strlen (enc->cur);	789	enc->cur += strlen (enc->cur);
440	}	790	}
441	else if (SvIOKp (sv))	791	else if (SvIOKp (sv))
442	{	792	{
		793	// we assume we can always read an IV as a UV and vice versa
		794	// we assume two's complement
		795	// we assume no aliasing issues in the union
		796	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		797	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		798	{
		799	// optimise the "small number case"
		800	// code will likely be branchless and use only a single multiplication
		801	// works for numbers up to 59074
		802	I32 i = SvIVX (sv);
		803	U32 u;
		804	char digit, nz = 0;
		805
443	need (enc, 64);	806	need (enc, 6);
		807
		808	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		809	u = i < 0 ? -i : i;
		810
		811	// convert to 4.28 fixed-point representation
		812	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		813
		814	// now output digit by digit, each time masking out the integer part
		815	// and multiplying by 5 while moving the decimal point one to the right,
		816	// resulting in a net multiplication by 10.
		817	// we always write the digit to memory but conditionally increment
		818	// the pointer, to enable the use of conditional move instructions.
		819	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		820	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		821	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		822	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		823	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		824	}
		825	else
		826	{
		827	// large integer, use the (rather slow) snprintf way.
		828	need (enc, IVUV_MAXCHARS);
444	enc->cur +=	829	enc->cur +=
445	SvIsUV(sv)	830	SvIsUV(sv)
446	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	831	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
447	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	832	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		833	}
448	}	834	}
449	else if (SvROK (sv))	835	else if (SvROK (sv))
450	{	836	encode_rv (enc, SvRV (sv));
451	SV *rv = SvRV (sv);	837	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
452
453	if (enc->indent >= enc->max_depth)
454	croak ("data structure too deep (hit recursion limit)");
455
456	switch (SvTYPE (rv))
457	{
458	case SVt_PVAV: encode_av (enc, (AV *)rv); break;
459	case SVt_PVHV: encode_hv (enc, (HV *)rv); break;
460
461	default:
462	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
463	SvPV_nolen (sv));
464	}
465	}
466	else if (!SvOK (sv))
467	encode_str (enc, "null", 4, 0);	838	encode_str (enc, "null", 4, 0);
468	else	839	else
469	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	840	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
470	SvPV_nolen (sv), SvFLAGS (sv));	841	SvPV_nolen (sv), SvFLAGS (sv));
471	}	842	}
472		843
473	static SV *	844	static SV *
474	encode_json (SV *scalar, UV flags)	845	encode_json (SV *scalar, JSON *json)
475	{	846	{
		847	enc_t enc;
		848
476	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	849	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
477	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	850	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
478		851
479	enc_t enc;	852	enc.json = *json;
480	enc.flags = flags;
481	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	853	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
482	enc.cur = SvPVX (enc.sv);	854	enc.cur = SvPVX (enc.sv);
483	enc.end = SvEND (enc.sv);	855	enc.end = SvEND (enc.sv);
484	enc.indent = 0;	856	enc.indent = 0;
485	enc.max_depth = 0x7fffffffUL;	857	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		858	: enc.json.flags & F_LATIN1 ? 0x000100UL
		859	: 0x110000UL;
486		860
487	SvPOK_only (enc.sv);	861	SvPOK_only (enc.sv);
488	encode_sv (&enc, scalar);	862	encode_sv (&enc, scalar);
		863	encode_nl (&enc);
489		864
		865	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		866	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		867
490	if (!(flags & (F_ASCII | F_UTF8)))	868	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
491	SvUTF8_on (enc.sv);	869	SvUTF8_on (enc.sv);
492		870
493	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
494
495	if (enc.flags & F_SHRINK)	871	if (enc.json.flags & F_SHRINK)
496	shrink (enc.sv);	872	shrink (enc.sv);
497		873
498	return enc.sv;	874	return enc.sv;
499	}	875	}
500		876
…		…
505	typedef struct	881	typedef struct
506	{	882	{
507	char *cur; // current parser pointer	883	char *cur; // current parser pointer
508	char *end; // end of input string	884	char *end; // end of input string
509	const char *err; // parse error, if != 0	885	const char *err; // parse error, if != 0
510	UV flags; // F_*	886	JSON json;
		887	U32 depth; // recursion depth
		888	U32 maxdepth; // recursion depth limit
511	} dec_t;	889	} dec_t;
512		890
513	static void	891	INLINE void
		892	decode_comment (dec_t *dec)
		893	{
		894	// only '#'-style comments allowed a.t.m.
		895
		896	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		897	++dec->cur;
		898	}
		899
		900	INLINE void
514	decode_ws (dec_t *dec)	901	decode_ws (dec_t *dec)
515	{	902	{
516	for (;;)	903	for (;;)
517	{	904	{
518	char ch = *dec->cur;	905	char ch = *dec->cur;
519		906
520	if (ch > 0x20	907	if (ch > 0x20)
		908	{
		909	if (expect_false (ch == '#'))
		910	{
		911	if (dec->json.flags & F_RELAXED)
		912	decode_comment (dec);
		913	else
		914	break;
		915	}
		916	else
		917	break;
		918	}
521	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	919	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
522	break;	920	break; // parse error, but let higher level handle it, gives better error messages
523		921
524	++dec->cur;	922	++dec->cur;
525	}	923	}
526	}	924	}
527		925
528	#define ERR(reason) SB dec->err = reason; goto fail; SE	926	#define ERR(reason) SB dec->err = reason; goto fail; SE
		927
529	#define EXPECT_CH(ch) SB \	928	#define EXPECT_CH(ch) SB \
530	if (*dec->cur != ch) \	929	if (*dec->cur != ch) \
531	ERR (# ch " expected"); \	930	ERR (# ch " expected"); \
532	++dec->cur; \	931	++dec->cur; \
533	SE	932	SE
534		933
		934	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
		935	#define DEC_DEC_DEPTH --dec->depth
		936
535	static SV *decode_sv (dec_t *dec);	937	static SV *decode_sv (dec_t *dec);
536		938
537	static signed char decode_hexdigit[256];	939	static signed char decode_hexdigit[256];
538		940
539	static UV	941	static UV
540	decode_4hex (dec_t *dec)	942	decode_4hex (dec_t *dec)
541	{	943	{
542	signed char d1, d2, d3, d4;	944	signed char d1, d2, d3, d4;
543	unsigned char *cur = (unsigned char *)dec->cur;	945	unsigned char *cur = (unsigned char *)dec->cur;
544		946
545	d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected");	947	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
546	d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected");	948	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
547	d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected");	949	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
548	d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected");	950	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
549		951
550	dec->cur += 4;	952	dec->cur += 4;
551		953
552	return ((UV)d1) << 12	954	return ((UV)d1) << 12
553	| ((UV)d2) << 8	955	| ((UV)d2) << 8
…		…
561	static SV *	963	static SV *
562	decode_str (dec_t *dec)	964	decode_str (dec_t *dec)
563	{	965	{
564	SV *sv = 0;	966	SV *sv = 0;
565	int utf8 = 0;	967	int utf8 = 0;
		968	char *dec_cur = dec->cur;
566		969
567	do	970	do
568	{	971	{
569	char buf [SHORT_STRING_LEN + UTF8_MAX_LEN];	972	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
570	char *cur = buf;	973	char *cur = buf;
571		974
572	do	975	do
573	{	976	{
574	unsigned char ch = *(unsigned char *)dec->cur++;	977	unsigned char ch = *(unsigned char *)dec_cur++;
575		978
576	if (ch == '"')	979	if (expect_false (ch == '"'))
577	{	980	{
578	--dec->cur;	981	--dec_cur;
579	break;	982	break;
580	}	983	}
581	else if (ch == '\\')	984	else if (expect_false (ch == '\\'))
582	{	985	{
583	switch (*dec->cur)	986	switch (*dec_cur)
584	{	987	{
585	case '\\':	988	case '\\':
586	case '/':	989	case '/':
587	case '"': *cur++ = *dec->cur++; break;	990	case '"': *cur++ = *dec_cur++; break;
588		991
589	case 'b': ++dec->cur; *cur++ = '\010'; break;	992	case 'b': ++dec_cur; *cur++ = '\010'; break;
590	case 't': ++dec->cur; *cur++ = '\011'; break;	993	case 't': ++dec_cur; *cur++ = '\011'; break;
591	case 'n': ++dec->cur; *cur++ = '\012'; break;	994	case 'n': ++dec_cur; *cur++ = '\012'; break;
592	case 'f': ++dec->cur; *cur++ = '\014'; break;	995	case 'f': ++dec_cur; *cur++ = '\014'; break;
593	case 'r': ++dec->cur; *cur++ = '\015'; break;	996	case 'r': ++dec_cur; *cur++ = '\015'; break;
594		997
595	case 'u':	998	case 'u':
596	{	999	{
597	UV lo, hi;	1000	UV lo, hi;
598	++dec->cur;	1001	++dec_cur;
599		1002
		1003	dec->cur = dec_cur;
600	hi = decode_4hex (dec);	1004	hi = decode_4hex (dec);
		1005	dec_cur = dec->cur;
601	if (hi == (UV)-1)	1006	if (hi == (UV)-1)
602	goto fail;	1007	goto fail;
603		1008
604	// possibly a surrogate pair	1009	// possibly a surrogate pair
605	if (hi >= 0xd800)	1010	if (hi >= 0xd800)
606	if (hi < 0xdc00)	1011	if (hi < 0xdc00)
607	{	1012	{
608	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	1013	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
609	ERR ("missing low surrogate character in surrogate pair");	1014	ERR ("missing low surrogate character in surrogate pair");
610		1015
611	dec->cur += 2;	1016	dec_cur += 2;
612		1017
		1018	dec->cur = dec_cur;
613	lo = decode_4hex (dec);	1019	lo = decode_4hex (dec);
		1020	dec_cur = dec->cur;
614	if (lo == (UV)-1)	1021	if (lo == (UV)-1)
615	goto fail;	1022	goto fail;
616		1023
617	if (lo < 0xdc00 || lo >= 0xe000)	1024	if (lo < 0xdc00 || lo >= 0xe000)
618	ERR ("surrogate pair expected");	1025	ERR ("surrogate pair expected");
…		…
624		1031
625	if (hi >= 0x80)	1032	if (hi >= 0x80)
626	{	1033	{
627	utf8 = 1;	1034	utf8 = 1;
628		1035
629	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	1036	cur = encode_utf8 (cur, hi);
630	}	1037	}
631	else	1038	else
632	*cur++ = hi;	1039	*cur++ = hi;
633	}	1040	}
634	break;	1041	break;
635		1042
636	default:	1043	default:
637	--dec->cur;	1044	--dec_cur;
638	ERR ("illegal backslash escape sequence in string");	1045	ERR ("illegal backslash escape sequence in string");
639	}	1046	}
640	}	1047	}
641	else if (ch >= 0x20 && ch <= 0x7f)	1048	else if (expect_true (ch >= 0x20 && ch < 0x80))
642	*cur++ = ch;	1049	*cur++ = ch;
643	else if (ch >= 0x80)	1050	else if (ch >= 0x80)
644	{	1051	{
645	--dec->cur;
646
647	STRLEN clen;	1052	STRLEN clen;
		1053
		1054	--dec_cur;
		1055
648	UV uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	1056	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
649	if (clen == (STRLEN)-1)	1057	if (clen == (STRLEN)-1)
650	ERR ("malformed UTF-8 character in JSON string");	1058	ERR ("malformed UTF-8 character in JSON string");
651		1059
652	do	1060	do
653	{
654	*cur++ = *dec->cur++;	1061	*cur++ = *dec_cur++;
655	}
656	while (--clen);	1062	while (--clen);
657		1063
658	utf8 = 1;	1064	utf8 = 1;
659	}	1065	}
660	else if (!ch)
661	ERR ("unexpected end of string while parsing json string");
662	else	1066	else
		1067	{
		1068	--dec_cur;
		1069
		1070	if (!ch)
		1071	ERR ("unexpected end of string while parsing JSON string");
		1072	else
663	ERR ("invalid character encountered");	1073	ERR ("invalid character encountered while parsing JSON string");
664		1074	}
665	}	1075	}
666	while (cur < buf + SHORT_STRING_LEN);	1076	while (cur < buf + SHORT_STRING_LEN);
667		1077
		1078	{
668	STRLEN len = cur - buf;	1079	STRLEN len = cur - buf;
669		1080
670	if (sv)	1081	if (sv)
671	{	1082	{
672	SvGROW (sv, SvCUR (sv) + len + 1);	1083	STRLEN cur = SvCUR (sv);
		1084
		1085	if (SvLEN (sv) <= cur + len)
		1086	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1087
673	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	1088	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
674	SvCUR_set (sv, SvCUR (sv) + len);	1089	SvCUR_set (sv, SvCUR (sv) + len);
675	}	1090	}
676	else	1091	else
677	sv = newSVpvn (buf, len);	1092	sv = newSVpvn (buf, len);
678	}	1093	}
		1094	}
679	while (*dec->cur != '"');	1095	while (*dec_cur != '"');
680		1096
681	++dec->cur;	1097	++dec_cur;
682		1098
683	if (sv)	1099	if (sv)
684	{	1100	{
685	SvPOK_only (sv);	1101	SvPOK_only (sv);
686	*SvEND (sv) = 0;	1102	*SvEND (sv) = 0;
…		…
689	SvUTF8_on (sv);	1105	SvUTF8_on (sv);
690	}	1106	}
691	else	1107	else
692	sv = newSVpvn ("", 0);	1108	sv = newSVpvn ("", 0);
693		1109
		1110	dec->cur = dec_cur;
694	return sv;	1111	return sv;
695		1112
696	fail:	1113	fail:
		1114	dec->cur = dec_cur;
697	return 0;	1115	return 0;
698	}	1116	}
699		1117
700	static SV *	1118	static SV *
701	decode_num (dec_t *dec)	1119	decode_num (dec_t *dec)
…		…
759	is_nv = 1;	1177	is_nv = 1;
760	}	1178	}
761		1179
762	if (!is_nv)	1180	if (!is_nv)
763	{	1181	{
764	UV uv;	1182	int len = dec->cur - start;
765	int numtype = grok_number (start, dec->cur - start, &uv);	1183
766	if (numtype & IS_NUMBER_IN_UV)	1184	// special case the rather common 1..5-digit-int case
767	if (numtype & IS_NUMBER_NEG)	1185	if (*start == '-')
		1186	switch (len)
768	{	1187	{
769	if (uv < (UV)IV_MIN)	1188	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
770	return newSViv (-(IV)uv);	1189	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
		1190	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1191	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1192	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
771	}	1193	}
		1194	else
		1195	switch (len)
		1196	{
		1197	case 1: return newSViv ( start [0] - '0' * 1);
		1198	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1199	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1200	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1201	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1202	}
		1203
		1204	{
		1205	UV uv;
		1206	int numtype = grok_number (start, len, &uv);
		1207	if (numtype & IS_NUMBER_IN_UV)
		1208	if (numtype & IS_NUMBER_NEG)
		1209	{
		1210	if (uv < (UV)IV_MIN)
		1211	return newSViv (-(IV)uv);
		1212	}
772	else	1213	else
773	return newSVuv (uv);	1214	return newSVuv (uv);
774	}	1215	}
775		1216
		1217	len -= *start == '-' ? 1 : 0;
		1218
		1219	// does not fit into IV or UV, try NV
		1220	if (len <= NV_DIG)
		1221	// fits into NV without loss of precision
		1222	return newSVnv (json_atof (start));
		1223
		1224	// everything else fails, convert it to a string
		1225	return newSVpvn (start, dec->cur - start);
		1226	}
		1227
		1228	// loss of precision here
776	return newSVnv (Atof (start));	1229	return newSVnv (json_atof (start));
777		1230
778	fail:	1231	fail:
779	return 0;	1232	return 0;
780	}	1233	}
781		1234
782	static SV *	1235	static SV *
783	decode_av (dec_t *dec)	1236	decode_av (dec_t *dec)
784	{	1237	{
785	AV *av = newAV ();	1238	AV *av = newAV ();
786		1239
		1240	DEC_INC_DEPTH;
787	decode_ws (dec);	1241	decode_ws (dec);
		1242
788	if (*dec->cur == ']')	1243	if (*dec->cur == ']')
789	++dec->cur;	1244	++dec->cur;
790	else	1245	else
791	for (;;)	1246	for (;;)
792	{	1247	{
…		…
808		1263
809	if (*dec->cur != ',')	1264	if (*dec->cur != ',')
810	ERR (", or ] expected while parsing array");	1265	ERR (", or ] expected while parsing array");
811		1266
812	++dec->cur;	1267	++dec->cur;
		1268
		1269	decode_ws (dec);
		1270
		1271	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1272	{
		1273	++dec->cur;
		1274	break;
		1275	}
813	}	1276	}
814		1277
		1278	DEC_DEC_DEPTH;
815	return newRV_noinc ((SV *)av);	1279	return newRV_noinc ((SV *)av);
816		1280
817	fail:	1281	fail:
818	SvREFCNT_dec (av);	1282	SvREFCNT_dec (av);
		1283	DEC_DEC_DEPTH;
819	return 0;	1284	return 0;
820	}	1285	}
821		1286
822	static SV *	1287	static SV *
823	decode_hv (dec_t *dec)	1288	decode_hv (dec_t *dec)
824	{	1289	{
		1290	SV *sv;
825	HV *hv = newHV ();	1291	HV *hv = newHV ();
826		1292
		1293	DEC_INC_DEPTH;
827	decode_ws (dec);	1294	decode_ws (dec);
		1295
828	if (*dec->cur == '}')	1296	if (*dec->cur == '}')
829	++dec->cur;	1297	++dec->cur;
830	else	1298	else
831	for (;;)	1299	for (;;)
832	{	1300	{
		1301	EXPECT_CH ('"');
		1302
		1303	// heuristic: assume that
		1304	// a) decode_str + hv_store_ent are abysmally slow.
		1305	// b) most hash keys are short, simple ascii text.
		1306	// => try to "fast-match" such strings to avoid
		1307	// the overhead of decode_str + hv_store_ent.
		1308	{
833	SV *key, *value;	1309	SV *value;
		1310	char *p = dec->cur;
		1311	char *e = p + 24; // only try up to 24 bytes
834		1312
835	decode_ws (dec); EXPECT_CH ('"');	1313	for (;;)
836
837	key = decode_str (dec);
838	if (!key)
839	goto fail;
840
841	decode_ws (dec); EXPECT_CH (':');
842
843	value = decode_sv (dec);
844	if (!value)
845	{	1314	{
		1315	// the >= 0x80 is false on most architectures
		1316	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1317	{
		1318	// slow path, back up and use decode_str
		1319	SV *key = decode_str (dec);
		1320	if (!key)
		1321	goto fail;
		1322
		1323	decode_ws (dec); EXPECT_CH (':');
		1324
		1325	decode_ws (dec);
		1326	value = decode_sv (dec);
		1327	if (!value)
		1328	{
		1329	SvREFCNT_dec (key);
		1330	goto fail;
		1331	}
		1332
		1333	hv_store_ent (hv, key, value, 0);
846	SvREFCNT_dec (key);	1334	SvREFCNT_dec (key);
		1335
		1336	break;
		1337	}
		1338	else if (*p == '"')
		1339	{
		1340	// fast path, got a simple key
		1341	char *key = dec->cur;
		1342	int len = p - key;
		1343	dec->cur = p + 1;
		1344
		1345	decode_ws (dec); EXPECT_CH (':');
		1346
		1347	decode_ws (dec);
		1348	value = decode_sv (dec);
		1349	if (!value)
847	goto fail;	1350	goto fail;
		1351
		1352	hv_store (hv, key, len, value, 0);
		1353
		1354	break;
		1355	}
		1356
		1357	++p;
848	}	1358	}
849		1359	}
850	//TODO: optimise
851	hv_store_ent (hv, key, value, 0);
852		1360
853	decode_ws (dec);	1361	decode_ws (dec);
854		1362
855	if (*dec->cur == '}')	1363	if (*dec->cur == '}')
856	{	1364	{
…		…
860		1368
861	if (*dec->cur != ',')	1369	if (*dec->cur != ',')
862	ERR (", or } expected while parsing object/hash");	1370	ERR (", or } expected while parsing object/hash");
863		1371
864	++dec->cur;	1372	++dec->cur;
		1373
		1374	decode_ws (dec);
		1375
		1376	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1377	{
		1378	++dec->cur;
		1379	break;
		1380	}
865	}	1381	}
866		1382
		1383	DEC_DEC_DEPTH;
867	return newRV_noinc ((SV *)hv);	1384	sv = newRV_noinc ((SV *)hv);
		1385
		1386	// check filter callbacks
		1387	if (dec->json.flags & F_HOOK)
		1388	{
		1389	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1390	{
		1391	HE *cb, *he;
		1392
		1393	hv_iterinit (hv);
		1394	he = hv_iternext (hv);
		1395	hv_iterinit (hv);
		1396
		1397	// the next line creates a mortal sv each time its called.
		1398	// might want to optimise this for common cases.
		1399	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1400
		1401	if (cb)
		1402	{
		1403	dSP;
		1404	int count;
		1405
		1406	ENTER; SAVETMPS; PUSHMARK (SP);
		1407	XPUSHs (HeVAL (he));
		1408	sv_2mortal (sv);
		1409
		1410	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1411
		1412	if (count == 1)
		1413	{
		1414	sv = newSVsv (POPs);
		1415	FREETMPS; LEAVE;
		1416	return sv;
		1417	}
		1418
		1419	SvREFCNT_inc (sv);
		1420	FREETMPS; LEAVE;
		1421	}
		1422	}
		1423
		1424	if (dec->json.cb_object)
		1425	{
		1426	dSP;
		1427	int count;
		1428
		1429	ENTER; SAVETMPS; PUSHMARK (SP);
		1430	XPUSHs (sv_2mortal (sv));
		1431
		1432	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1433
		1434	if (count == 1)
		1435	{
		1436	sv = newSVsv (POPs);
		1437	FREETMPS; LEAVE;
		1438	return sv;
		1439	}
		1440
		1441	SvREFCNT_inc (sv);
		1442	FREETMPS; LEAVE;
		1443	}
		1444	}
		1445
		1446	return sv;
868		1447
869	fail:	1448	fail:
870	SvREFCNT_dec (hv);	1449	SvREFCNT_dec (hv);
		1450	DEC_DEC_DEPTH;
871	return 0;	1451	return 0;
872	}	1452	}
873		1453
874	static SV *	1454	static SV *
875	decode_sv (dec_t *dec)	1455	decode_sv (dec_t *dec)
876	{	1456	{
877	decode_ws (dec);	1457	// the beauty of JSON: you need exactly one character lookahead
		1458	// to parse everything.
878	switch (*dec->cur)	1459	switch (*dec->cur)
879	{	1460	{
880	case '"': ++dec->cur; return decode_str (dec);	1461	case '"': ++dec->cur; return decode_str (dec);
881	case '[': ++dec->cur; return decode_av (dec);	1462	case '[': ++dec->cur; return decode_av (dec);
882	case '{': ++dec->cur; return decode_hv (dec);	1463	case '{': ++dec->cur; return decode_hv (dec);
883		1464
884	case '-':	1465	case '-':
885	case '0': case '1': case '2': case '3': case '4':	1466	case '0': case '1': case '2': case '3': case '4':
886	case '5': case '6': case '7': case '8': case '9':	1467	case '5': case '6': case '7': case '8': case '9':
887	return decode_num (dec);	1468	return decode_num (dec);
888		1469
889	case 't':	1470	case 't':
890	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1471	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
891	{	1472	{
892	dec->cur += 4;	1473	dec->cur += 4;
		1474	#if JSON_SLOW
		1475	json_true = get_bool ("JSON::XS::true");
		1476	#endif
893	return newSViv (1);	1477	return newSVsv (json_true);
894	}	1478	}
895	else	1479	else
896	ERR ("'true' expected");	1480	ERR ("'true' expected");
897		1481
898	break;	1482	break;
899		1483
900	case 'f':	1484	case 'f':
901	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1485	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
902	{	1486	{
903	dec->cur += 5;	1487	dec->cur += 5;
		1488	#if JSON_SLOW
		1489	json_false = get_bool ("JSON::XS::false");
		1490	#endif
904	return newSViv (0);	1491	return newSVsv (json_false);
905	}	1492	}
906	else	1493	else
907	ERR ("'false' expected");	1494	ERR ("'false' expected");
908		1495
909	break;	1496	break;
…		…
918	ERR ("'null' expected");	1505	ERR ("'null' expected");
919		1506
920	break;	1507	break;
921		1508
922	default:	1509	default:
923	ERR ("malformed json string, neither array, object, number, string or atom");	1510	ERR ("malformed JSON string, neither array, object, number, string or atom");
924	break;	1511	break;
925	}	1512	}
926		1513
927	fail:	1514	fail:
928	return 0;	1515	return 0;
929	}	1516	}
930		1517
931	static SV *	1518	static SV *
932	decode_json (SV *string, UV flags)	1519	decode_json (SV *string, JSON *json, char **offset_return)
933	{	1520	{
		1521	dec_t dec;
934	SV *sv;	1522	SV *sv;
935		1523
936	if (flags & F_UTF8)	1524	/* work around bugs in 5.10 where manipulating magic values
		1525	* will perl ignore the magic in subsequent accesses.
		1526	* also make a copy of non-PV values, to get them into a clean
		1527	* state (SvPV should do that, but it's buggy, see below).
		1528	*/
		1529	/*SvGETMAGIC (string);*/
		1530	if (SvMAGICAL (string) || !SvPOK (string))
		1531	string = sv_2mortal (newSVsv (string));
		1532
		1533	SvUPGRADE (string, SVt_PV);
		1534
		1535	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1536	* assertion with -DDEBUGGING, although SvCUR is documented to
		1537	* return the xpv_cur field which certainly exists after upgrading.
		1538	* according to nicholas clark, calling SvPOK fixes this.
		1539	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1540	* and hope for the best.
		1541	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1542	* assertion business is seriously broken, try yet another workaround
		1543	* for the broken -DDEBUGGING.
		1544	*/
		1545	{
		1546	#ifdef DEBUGGING
		1547	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1548	#else
		1549	STRLEN offset = SvCUR (string);
		1550	#endif
		1551
		1552	if (offset > json->max_size && json->max_size)
		1553	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1554	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1555	}
		1556
		1557	if (DECODE_WANTS_OCTETS (json))
937	sv_utf8_downgrade (string, 0);	1558	sv_utf8_downgrade (string, 0);
938	else	1559	else
939	sv_utf8_upgrade (string);	1560	sv_utf8_upgrade (string);
940		1561
941	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1562	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
942		1563
943	dec_t dec;	1564	dec.json = *json;
944	dec.flags = flags;
945	dec.cur = SvPVX (string);	1565	dec.cur = SvPVX (string);
946	dec.end = SvEND (string);	1566	dec.end = SvEND (string);
947	dec.err = 0;	1567	dec.err = 0;
		1568	dec.depth = 0;
948		1569
		1570	if (dec.json.cb_object || dec.json.cb_sk_object)
		1571	dec.json.flags |= F_HOOK;
		1572
		1573	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1574
		1575	decode_ws (&dec);
949	sv = decode_sv (&dec);	1576	sv = decode_sv (&dec);
950		1577
		1578	if (offset_return)
		1579	*offset_return = dec.cur;
		1580
		1581	if (!(offset_return || !sv))
		1582	{
		1583	// check for trailing garbage
		1584	decode_ws (&dec);
		1585
		1586	if (*dec.cur)
		1587	{
		1588	dec.err = "garbage after JSON object";
		1589	SvREFCNT_dec (sv);
		1590	sv = 0;
		1591	}
		1592	}
		1593
951	if (!sv)	1594	if (!sv)
952	{	1595	{
953	IV offset = dec.flags & F_UTF8
954	? dec.cur - SvPVX (string)
955	: utf8_distance (dec.cur, SvPVX (string));
956	SV *uni = sv_newmortal ();	1596	SV *uni = sv_newmortal ();
957		1597
958	// horrible hack to silence warning inside pv_uni_display	1598	// horrible hack to silence warning inside pv_uni_display
959	COP cop = *PL_curcop;	1599	COP cop = *PL_curcop;
960	cop.cop_warnings = pWARN_NONE;	1600	cop.cop_warnings = pWARN_NONE;
…		…
962	SAVEVPTR (PL_curcop);	1602	SAVEVPTR (PL_curcop);
963	PL_curcop = &cop;	1603	PL_curcop = &cop;
964	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1604	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
965	LEAVE;	1605	LEAVE;
966		1606
967	croak ("%s, at character offset %d (%s)",	1607	croak ("%s, at character offset %d (before \"%s\")",
968	dec.err,	1608	dec.err,
969	(int)offset,	1609	ptr_to_index (string, dec.cur),
970	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1610	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
971	}	1611	}
972		1612
973	sv = sv_2mortal (sv);	1613	sv = sv_2mortal (sv);
974		1614
975	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1615	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
976	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1616	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
977		1617
978	return sv;	1618	return sv;
		1619	}
		1620
		1621	/////////////////////////////////////////////////////////////////////////////
		1622	// incremental parser
		1623
		1624	static void
		1625	incr_parse (JSON *self)
		1626	{
		1627	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1628
		1629	// the state machine here is a bit convoluted and could be simplified a lot
		1630	// but this would make it slower, so...
		1631
		1632	for (;;)
		1633	{
		1634	//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
		1635	switch (self->incr_mode)
		1636	{
		1637	// only used for initial whitespace skipping
		1638	case INCR_M_WS:
		1639	for (;;)
		1640	{
		1641	if (*p > 0x20)
		1642	{
		1643	if (*p == '#')
		1644	{
		1645	self->incr_mode = INCR_M_C0;
		1646	goto incr_m_c;
		1647	}
		1648	else
		1649	{
		1650	self->incr_mode = INCR_M_JSON;
		1651	goto incr_m_json;
		1652	}
		1653	}
		1654	else if (!*p)
		1655	goto interrupt;
		1656
		1657	++p;
		1658	}
		1659
		1660	// skip a single char inside a string (for \\-processing)
		1661	case INCR_M_BS:
		1662	if (!*p)
		1663	goto interrupt;
		1664
		1665	++p;
		1666	self->incr_mode = INCR_M_STR;
		1667	goto incr_m_str;
		1668
		1669	// inside #-style comments
		1670	case INCR_M_C0:
		1671	case INCR_M_C1:
		1672	incr_m_c:
		1673	for (;;)
		1674	{
		1675	if (*p == '\n')
		1676	{
		1677	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1678	break;
		1679	}
		1680	else if (!*p)
		1681	goto interrupt;
		1682
		1683	++p;
		1684	}
		1685
		1686	break;
		1687
		1688	// inside a string
		1689	case INCR_M_STR:
		1690	incr_m_str:
		1691	for (;;)
		1692	{
		1693	if (*p == '"')
		1694	{
		1695	++p;
		1696	self->incr_mode = INCR_M_JSON;
		1697
		1698	if (!self->incr_nest)
		1699	goto interrupt;
		1700
		1701	goto incr_m_json;
		1702	}
		1703	else if (*p == '\\')
		1704	{
		1705	++p; // "virtually" consumes character after \
		1706
		1707	if (!*p) // if at end of string we have to switch modes
		1708	{
		1709	self->incr_mode = INCR_M_BS;
		1710	goto interrupt;
		1711	}
		1712	}
		1713	else if (!*p)
		1714	goto interrupt;
		1715
		1716	++p;
		1717	}
		1718
		1719	// after initial ws, outside string
		1720	case INCR_M_JSON:
		1721	incr_m_json:
		1722	for (;;)
		1723	{
		1724	switch (*p++)
		1725	{
		1726	case 0:
		1727	--p;
		1728	goto interrupt;
		1729
		1730	case 0x09:
		1731	case 0x0a:
		1732	case 0x0d:
		1733	case 0x20:
		1734	if (!self->incr_nest)
		1735	{
		1736	--p; // do not eat the whitespace, let the next round do it
		1737	goto interrupt;
		1738	}
		1739	break;
		1740
		1741	case '"':
		1742	self->incr_mode = INCR_M_STR;
		1743	goto incr_m_str;
		1744
		1745	case '[':
		1746	case '{':
		1747	if (++self->incr_nest > self->max_depth)
		1748	croak (ERR_NESTING_EXCEEDED);
		1749	break;
		1750
		1751	case ']':
		1752	case '}':
		1753	if (--self->incr_nest <= 0)
		1754	goto interrupt;
		1755	break;
		1756
		1757	case '#':
		1758	self->incr_mode = INCR_M_C1;
		1759	goto incr_m_c;
		1760	}
		1761	}
		1762	}
		1763
		1764	modechange:
		1765	;
		1766	}
		1767
		1768	interrupt:
		1769	self->incr_pos = p - SvPVX (self->incr_text);
		1770	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1771	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
979	}	1772	}
980		1773
981	/////////////////////////////////////////////////////////////////////////////	1774	/////////////////////////////////////////////////////////////////////////////
982	// XS interface functions	1775	// XS interface functions
983		1776
…		…
985		1778
986	BOOT:	1779	BOOT:
987	{	1780	{
988	int i;	1781	int i;
989		1782
990	memset (decode_hexdigit, 0xff, 256);
991	for (i = 10; i--; )	1783	for (i = 0; i < 256; ++i)
992	decode_hexdigit ['0' + i] = i;	1784	decode_hexdigit [i] =
		1785	i >= '0' && i <= '9' ? i - '0'
		1786	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1787	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1788	: -1;
993		1789
994	for (i = 7; i--; )
995	{
996	decode_hexdigit ['a' + i] = 10 + i;
997	decode_hexdigit ['A' + i] = 10 + i;
998	}
999
1000	json_stash = gv_stashpv ("JSON::XS", 1);	1790	json_stash = gv_stashpv ("JSON::XS" , 1);
		1791	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1792
		1793	json_true = get_bool ("JSON::XS::true");
		1794	json_false = get_bool ("JSON::XS::false");
		1795
		1796	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1001	}	1797	}
1002		1798
1003	PROTOTYPES: DISABLE	1799	PROTOTYPES: DISABLE
1004		1800
1005	SV *new (char *dummy)	1801	void CLONE (...)
1006	CODE:	1802	CODE:
1007	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1803	json_stash = 0;
		1804	json_boolean_stash = 0;
		1805
		1806	void new (char *klass)
		1807	PPCODE:
		1808	{
		1809	SV *pv = NEWSV (0, sizeof (JSON));
		1810	SvPOK_only (pv);
		1811	json_init ((JSON *)SvPVX (pv));
		1812	XPUSHs (sv_2mortal (sv_bless (
		1813	newRV_noinc (pv),
		1814	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1815	)));
		1816	}
		1817
		1818	void ascii (JSON *self, int enable = 1)
		1819	ALIAS:
		1820	ascii = F_ASCII
		1821	latin1 = F_LATIN1
		1822	utf8 = F_UTF8
		1823	indent = F_INDENT
		1824	canonical = F_CANONICAL
		1825	space_before = F_SPACE_BEFORE
		1826	space_after = F_SPACE_AFTER
		1827	pretty = F_PRETTY
		1828	allow_nonref = F_ALLOW_NONREF
		1829	shrink = F_SHRINK
		1830	allow_blessed = F_ALLOW_BLESSED
		1831	convert_blessed = F_CONV_BLESSED
		1832	relaxed = F_RELAXED
		1833	allow_unknown = F_ALLOW_UNKNOWN
		1834	PPCODE:
		1835	{
		1836	if (enable)
		1837	self->flags |= ix;
		1838	else
		1839	self->flags &= ~ix;
		1840
		1841	XPUSHs (ST (0));
		1842	}
		1843
		1844	void get_ascii (JSON *self)
		1845	ALIAS:
		1846	get_ascii = F_ASCII
		1847	get_latin1 = F_LATIN1
		1848	get_utf8 = F_UTF8
		1849	get_indent = F_INDENT
		1850	get_canonical = F_CANONICAL
		1851	get_space_before = F_SPACE_BEFORE
		1852	get_space_after = F_SPACE_AFTER
		1853	get_allow_nonref = F_ALLOW_NONREF
		1854	get_shrink = F_SHRINK
		1855	get_allow_blessed = F_ALLOW_BLESSED
		1856	get_convert_blessed = F_CONV_BLESSED
		1857	get_relaxed = F_RELAXED
		1858	get_allow_unknown = F_ALLOW_UNKNOWN
		1859	PPCODE:
		1860	XPUSHs (boolSV (self->flags & ix));
		1861
		1862	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		1863	PPCODE:
		1864	self->max_depth = max_depth;
		1865	XPUSHs (ST (0));
		1866
		1867	U32 get_max_depth (JSON *self)
		1868	CODE:
		1869	RETVAL = self->max_depth;
1008	OUTPUT:	1870	OUTPUT:
1009	RETVAL	1871	RETVAL
1010		1872
1011	SV *ascii (SV *self, int enable = 1)	1873	void max_size (JSON *self, U32 max_size = 0)
1012	ALIAS:	1874	PPCODE:
1013	ascii = F_ASCII	1875	self->max_size = max_size;
1014	utf8 = F_UTF8	1876	XPUSHs (ST (0));
1015	indent = F_INDENT	1877
1016	canonical = F_CANONICAL	1878	int get_max_size (JSON *self)
1017	space_before = F_SPACE_BEFORE
1018	space_after = F_SPACE_AFTER
1019	pretty = F_PRETTY
1020	allow_nonref = F_ALLOW_NONREF
1021	shrink = F_SHRINK
1022	CODE:	1879	CODE:
1023	{	1880	RETVAL = self->max_size;
1024	UV *uv = SvJSON (self);
1025	if (enable)
1026	*uv |= ix;
1027	else
1028	*uv &= ~ix;
1029
1030	RETVAL = newSVsv (self);
1031	}
1032	OUTPUT:	1881	OUTPUT:
1033	RETVAL	1882	RETVAL
1034		1883
1035	void encode (SV *self, SV *scalar)	1884	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1036	PPCODE:	1885	PPCODE:
1037	XPUSHs (encode_json (scalar, *SvJSON (self)));	1886	{
		1887	SvREFCNT_dec (self->cb_object);
		1888	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
1038		1889
1039	void decode (SV *self, SV *jsonstr)	1890	XPUSHs (ST (0));
		1891	}
		1892
		1893	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1040	PPCODE:	1894	PPCODE:
		1895	{
		1896	if (!self->cb_sk_object)
		1897	self->cb_sk_object = newHV ();
		1898
		1899	if (SvOK (cb))
		1900	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1901	else
		1902	{
		1903	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1904
		1905	if (!HvKEYS (self->cb_sk_object))
		1906	{
		1907	SvREFCNT_dec (self->cb_sk_object);
		1908	self->cb_sk_object = 0;
		1909	}
		1910	}
		1911
		1912	XPUSHs (ST (0));
		1913	}
		1914
		1915	void encode (JSON *self, SV *scalar)
		1916	PPCODE:
		1917	XPUSHs (encode_json (scalar, self));
		1918
		1919	void decode (JSON *self, SV *jsonstr)
		1920	PPCODE:
1041	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1921	XPUSHs (decode_json (jsonstr, self, 0));
		1922
		1923	void decode_prefix (JSON *self, SV *jsonstr)
		1924	PPCODE:
		1925	{
		1926	char *offset;
		1927	EXTEND (SP, 2);
		1928	PUSHs (decode_json (jsonstr, self, &offset));
		1929	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		1930	}
		1931
		1932	void incr_parse (JSON *self, SV *jsonstr = 0)
		1933	PPCODE:
		1934	{
		1935	if (!self->incr_text)
		1936	self->incr_text = newSVpvn ("", 0);
		1937
		1938	/* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
		1939	if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
		1940	if (DECODE_WANTS_OCTETS (self))
		1941	{
		1942	if (self->incr_pos)
		1943	self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
		1944	(U8 *)SvPVX (self->incr_text) + self->incr_pos);
		1945
		1946	sv_utf8_downgrade (self->incr_text, 0);
		1947	}
		1948	else
		1949	{
		1950	sv_utf8_upgrade (self->incr_text);
		1951
		1952	if (self->incr_pos)
		1953	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1954	- (U8 *)SvPVX (self->incr_text);
		1955	}
		1956
		1957	// append data, if any
		1958	if (jsonstr)
		1959	{
		1960	/* make sure both strings have same encoding */
		1961	if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
		1962	if (SvUTF8 (jsonstr))
		1963	sv_utf8_downgrade (jsonstr, 0);
		1964	else
		1965	sv_utf8_upgrade (jsonstr);
		1966
		1967	/* and then just blindly append */
		1968	{
		1969	STRLEN len;
		1970	const char *str = SvPV (jsonstr, len);
		1971	STRLEN cur = SvCUR (self->incr_text);
		1972
		1973	if (SvLEN (self->incr_text) <= cur + len)
		1974	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1975
		1976	Move (str, SvEND (self->incr_text), len, char);
		1977	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1978	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1979	}
		1980	}
		1981
		1982	if (GIMME_V != G_VOID)
		1983	do
		1984	{
		1985	char *offset;
		1986
		1987	if (!INCR_DONE (self))
		1988	{
		1989	incr_parse (self);
		1990
		1991	if (self->incr_pos > self->max_size && self->max_size)
		1992	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1993	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1994
		1995	if (!INCR_DONE (self))
		1996	break;
		1997	}
		1998
		1999	XPUSHs (decode_json (self->incr_text, self, &offset));
		2000
		2001	self->incr_pos -= offset - SvPVX (self->incr_text);
		2002	self->incr_nest = 0;
		2003	self->incr_mode = 0;
		2004
		2005	sv_chop (self->incr_text, offset);
		2006	}
		2007	while (GIMME_V == G_ARRAY);
		2008	}
		2009
		2010	SV *incr_text (JSON *self)
		2011	ATTRS: lvalue
		2012	CODE:
		2013	{
		2014	if (self->incr_pos)
		2015	croak ("incr_text can not be called when the incremental parser already started parsing");
		2016
		2017	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		2018	}
		2019	OUTPUT:
		2020	RETVAL
		2021
		2022	void incr_skip (JSON *self)
		2023	CODE:
		2024	{
		2025	if (self->incr_pos)
		2026	{
		2027	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2028	self->incr_pos = 0;
		2029	self->incr_nest = 0;
		2030	self->incr_mode = 0;
		2031	}
		2032	}
		2033
		2034	void incr_reset (JSON *self)
		2035	CODE:
		2036	{
		2037	SvREFCNT_dec (self->incr_text);
		2038	self->incr_text = 0;
		2039	self->incr_pos = 0;
		2040	self->incr_nest = 0;
		2041	self->incr_mode = 0;
		2042	}
		2043
		2044	void DESTROY (JSON *self)
		2045	CODE:
		2046	SvREFCNT_dec (self->cb_sk_object);
		2047	SvREFCNT_dec (self->cb_object);
		2048	SvREFCNT_dec (self->incr_text);
1042		2049
1043	PROTOTYPES: ENABLE	2050	PROTOTYPES: ENABLE
1044		2051
1045	void to_json (SV *scalar)	2052	void encode_json (SV *scalar)
1046	ALIAS:	2053	ALIAS:
1047	objToJson = 0	2054	to_json_ = 0
		2055	encode_json = F_UTF8
1048	PPCODE:	2056	PPCODE:
		2057	{
		2058	JSON json;
		2059	json_init (&json);
		2060	json.flags |= ix;
1049	XPUSHs (encode_json (scalar, F_UTF8));	2061	XPUSHs (encode_json (scalar, &json));
		2062	}
1050		2063
1051	void from_json (SV *jsonstr)	2064	void decode_json (SV *jsonstr)
1052	ALIAS:	2065	ALIAS:
1053	jsonToObj = 0	2066	from_json_ = 0
		2067	decode_json = F_UTF8
1054	PPCODE:	2068	PPCODE:
		2069	{
		2070	JSON json;
		2071	json_init (&json);
		2072	json.flags |= ix;
1055	XPUSHs (decode_json (jsonstr, F_UTF8));	2073	XPUSHs (decode_json (jsonstr, &json, 0));
		2074	}
1056		2075

Diff Legend

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