[ViewVC] Diff of: cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.69 by root, Wed Dec 5 10:59:28 2007 UTC vs.
Revision 1.115 by root, Fri Dec 21 05:16:19 2012 UTC

…		…
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>
9	#include <float.h>	10	#include <float.h>
10		11
11	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	12	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
12	# define snprintf _snprintf // C compilers have this in stdio.h	13	# define snprintf _snprintf // C compilers have this in stdio.h
13	#endif	14	#endif
14		15
15	// some old perls do not have this, try to make it work, no	16	// some old perls do not have this, try to make it work, no
16	// guarentees, though. if it breaks, you get to keep the pieces.	17	// guarantees, though. if it breaks, you get to keep the pieces.
17	#ifndef UTF8_MAXBYTES	18	#ifndef UTF8_MAXBYTES
18	# define UTF8_MAXBYTES 13	19	# define UTF8_MAXBYTES 13
19	#endif	20	#endif
		21
		22	// three extra for rounding, sign, and end of string
		23	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
20		24
21	#define F_ASCII 0x00000001UL	25	#define F_ASCII 0x00000001UL
22	#define F_LATIN1 0x00000002UL	26	#define F_LATIN1 0x00000002UL
23	#define F_UTF8 0x00000004UL	27	#define F_UTF8 0x00000004UL
24	#define F_INDENT 0x00000008UL	28	#define F_INDENT 0x00000008UL
…		…
28	#define F_ALLOW_NONREF 0x00000100UL	32	#define F_ALLOW_NONREF 0x00000100UL
29	#define F_SHRINK 0x00000200UL	33	#define F_SHRINK 0x00000200UL
30	#define F_ALLOW_BLESSED 0x00000400UL	34	#define F_ALLOW_BLESSED 0x00000400UL
31	#define F_CONV_BLESSED 0x00000800UL	35	#define F_CONV_BLESSED 0x00000800UL
32	#define F_RELAXED 0x00001000UL	36	#define F_RELAXED 0x00001000UL
33		37	#define F_ALLOW_UNKNOWN 0x00002000UL
34	#define F_MAXDEPTH 0xf8000000UL
35	#define S_MAXDEPTH 27
36	#define F_MAXSIZE 0x01f00000UL
37	#define S_MAXSIZE 20
38	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing	38	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
39		39
40	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
41	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
42
43	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER	40	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER
44	#define F_DEFAULT (9UL << S_MAXDEPTH)
45		41
46	#define INIT_SIZE 32 // initial scalar size to be allocated	42	#define INIT_SIZE 32 // initial scalar size to be allocated
47	#define INDENT_STEP 3 // spaces per indentation level	43	#define INDENT_STEP 3 // spaces per indentation level
48		44
49	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size	45	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
50		46
		47	#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
		48
51	#define SB do {	49	#define SB do {
52	#define SE } while (0)	50	#define SE } while (0)
53		51
54	#if __GNUC__ >= 3	52	#if __GNUC__ >= 3
55	# define expect(expr,value) __builtin_expect ((expr),(value))	53	# define expect(expr,value) __builtin_expect ((expr), (value))
56	# define inline inline	54	# define INLINE static inline
57	#else	55	#else
58	# define expect(expr,value) (expr)	56	# define expect(expr,value) (expr)
59	# define inline static	57	# define INLINE static
60	#endif	58	#endif
61		59
62	#define expect_false(expr) expect ((expr) != 0, 0)	60	#define expect_false(expr) expect ((expr) != 0, 0)
63	#define expect_true(expr) expect ((expr) != 0, 1)	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?)"
64		68
65	#ifdef USE_ITHREADS	69	#ifdef USE_ITHREADS
66	# define JSON_SLOW 1	70	# define JSON_SLOW 1
67	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))	71	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
68	#else	72	#else
69	# define JSON_SLOW 0	73	# define JSON_SLOW 0
70	# define JSON_STASH json_stash	74	# define JSON_STASH json_stash
71	#endif	75	#endif
72		76
		77	// the amount of HEs to allocate on the stack, when sorting keys
		78	#define STACK_HES 64
		79
73	static HV json_stash, json_boolean_stash; // JSON::XS::	80	static HV json_stash, json_boolean_stash; // JSON::XS::
74	static SV json_true, json_false;	81	static SV json_true, json_false;
75		82
		83	enum {
		84	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		85	INCR_M_STR, // inside string
		86	INCR_M_BS, // inside backslash
		87	INCR_M_C0, // inside comment in initial whitespace sequence
		88	INCR_M_C1, // inside comment in other places
		89	INCR_M_JSON // outside anything, count nesting
		90	};
		91
		92	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		93
76	typedef struct {	94	typedef struct {
77	U32 flags;	95	U32 flags;
		96	U32 max_depth;
		97	STRLEN max_size;
		98
78	SV *cb_object;	99	SV *cb_object;
79	HV *cb_sk_object;	100	HV *cb_sk_object;
		101
		102	// for the incremental parser
		103	SV *incr_text; // the source text so far
		104	STRLEN incr_pos; // the current offset into the text
		105	int incr_nest; // {[]}-nesting level
		106	unsigned char incr_mode;
80	} JSON;	107	} JSON;
		108
		109	INLINE void
		110	json_init (JSON *json)
		111	{
		112	Zero (json, 1, JSON);
		113	json->max_depth = 512;
		114	}
81		115
82	/////////////////////////////////////////////////////////////////////////////	116	/////////////////////////////////////////////////////////////////////////////
83	// utility functions	117	// utility functions
84		118
85	inline void	119	INLINE SV *
		120	get_bool (const char *name)
		121	{
		122	SV *sv = get_sv (name, 1);
		123
		124	SvREADONLY_on (sv);
		125	SvREADONLY_on (SvRV (sv));
		126
		127	return sv;
		128	}
		129
		130	INLINE void
86	shrink (SV *sv)	131	shrink (SV *sv)
87	{	132	{
88	sv_utf8_downgrade (sv, 1);	133	sv_utf8_downgrade (sv, 1);
		134
89	if (SvLEN (sv) > SvCUR (sv) + 1)	135	if (SvLEN (sv) > SvCUR (sv) + 1)
90	{	136	{
91	#ifdef SvPV_shrink_to_cur	137	#ifdef SvPV_shrink_to_cur
92	SvPV_shrink_to_cur (sv);	138	SvPV_shrink_to_cur (sv);
93	#elif defined (SvPV_renew)	139	#elif defined (SvPV_renew)
…		…
99	// decode an utf-8 character and return it, or (UV)-1 in	145	// decode an utf-8 character and return it, or (UV)-1 in
100	// case of an error.	146	// case of an error.
101	// we special-case "safe" characters from U+80 .. U+7FF,	147	// we special-case "safe" characters from U+80 .. U+7FF,
102	// but use the very good perl function to parse anything else.	148	// but use the very good perl function to parse anything else.
103	// note that we never call this function for a ascii codepoints	149	// note that we never call this function for a ascii codepoints
104	inline UV	150	INLINE UV
105	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	151	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
106	{	152	{
107	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))	153	if (expect_true (len >= 2
108	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	154	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
109	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	155	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
110	{	156	{
111	*clen = 2;	157	*clen = 2;
112	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);	158	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
113	}	159	}
114	else	160	else
115	{	161	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
116	*clen = (STRLEN)-1;	162	}
117	return (UV)-1;	163
		164	// likewise for encoding, also never called for ascii codepoints
		165	// this function takes advantage of this fact, although current gccs
		166	// seem to optimise the check for >= 0x80 away anyways
		167	INLINE unsigned char *
		168	encode_utf8 (unsigned char *s, UV ch)
		169	{
		170	if (expect_false (ch < 0x000080))
		171	*s++ = ch;
		172	else if (expect_true (ch < 0x000800))
		173	*s++ = 0xc0 \| ( ch >> 6),
		174	*s++ = 0x80 \| ( ch & 0x3f);
		175	else if ( ch < 0x010000)
		176	*s++ = 0xe0 \| ( ch >> 12),
		177	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		178	*s++ = 0x80 \| ( ch & 0x3f);
		179	else if ( ch < 0x110000)
		180	*s++ = 0xf0 \| ( ch >> 18),
		181	*s++ = 0x80 \| ((ch >> 12) & 0x3f),
		182	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		183	*s++ = 0x80 \| ( ch & 0x3f);
		184
		185	return s;
		186	}
		187
		188	// convert offset pointer to character index, sv must be string
		189	static STRLEN
		190	ptr_to_index (SV sv, char offset)
		191	{
		192	return SvUTF8 (sv)
		193	? utf8_distance (offset, SvPVX (sv))
		194	: offset - SvPVX (sv);
		195	}
		196
		197	/////////////////////////////////////////////////////////////////////////////
		198	// fp hell
		199
		200	// scan a group of digits, and a trailing exponent
		201	static void
		202	json_atof_scan1 (const char s, NV accum, int *expo, int postdp, int maxdepth)
		203	{
		204	UV uaccum = 0;
		205	int eaccum = 0;
		206
		207	// if we recurse too deep, skip all remaining digits
		208	// to avoid a stack overflow attack
		209	if (expect_false (--maxdepth <= 0))
		210	while (((U8)*s - '0') < 10)
		211	++s;
		212
		213	for (;;)
118	}	214	{
119	}	215	U8 dig = (U8)*s - '0';
120		216
		217	if (expect_false (dig >= 10))
		218	{
		219	if (dig == (U8)((U8)'.' - (U8)'0'))
		220	{
		221	++s;
		222	json_atof_scan1 (s, accum, expo, 1, maxdepth);
		223	}
		224	else if ((dig \| ' ') == 'e' - '0')
		225	{
		226	int exp2 = 0;
		227	int neg = 0;
		228
		229	++s;
		230
		231	if (*s == '-')
		232	{
		233	++s;
		234	neg = 1;
		235	}
		236	else if (*s == '+')
		237	++s;
		238
		239	while ((dig = (U8)*s - '0') < 10)
		240	exp2 = exp2 * 10 + *s++ - '0';
		241
		242	*expo += neg ? -exp2 : exp2;
		243	}
		244
		245	break;
		246	}
		247
		248	++s;
		249
		250	uaccum = uaccum * 10 + dig;
		251	++eaccum;
		252
		253	// if we have too many digits, then recurse for more
		254	// we actually do this for rather few digits
		255	if (uaccum >= (UV_MAX - 9) / 10)
		256	{
		257	if (postdp) *expo -= eaccum;
		258	json_atof_scan1 (s, accum, expo, postdp, maxdepth);
		259	if (postdp) *expo += eaccum;
		260
		261	break;
		262	}
		263	}
		264
		265	// this relies greatly on the quality of the pow ()
		266	// implementation of the platform, but a good
		267	// implementation is hard to beat.
		268	// (IEEE 754 conformant ones are required to be exact)
		269	if (postdp) *expo -= eaccum;
		270	accum += uaccum Perl_pow (10., *expo);
		271	*expo += eaccum;
		272	}
		273
		274	static NV
		275	json_atof (const char *s)
		276	{
		277	NV accum = 0.;
		278	int expo = 0;
		279	int neg = 0;
		280
		281	if (*s == '-')
		282	{
		283	++s;
		284	neg = 1;
		285	}
		286
		287	// a recursion depth of ten gives us >>500 bits
		288	json_atof_scan1 (s, &accum, &expo, 0, 10);
		289
		290	return neg ? -accum : accum;
		291	}
121	/////////////////////////////////////////////////////////////////////////////	292	/////////////////////////////////////////////////////////////////////////////
122	// encoder	293	// encoder
123		294
124	// structure used for encoding JSON	295	// structure used for encoding JSON
125	typedef struct	296	typedef struct
…		…
127	char *cur; // SvPVX (sv) + current output position	298	char *cur; // SvPVX (sv) + current output position
128	char *end; // SvEND (sv)	299	char *end; // SvEND (sv)
129	SV *sv; // result scalar	300	SV *sv; // result scalar
130	JSON json;	301	JSON json;
131	U32 indent; // indentation level	302	U32 indent; // indentation level
132	U32 maxdepth; // max. indentation/recursion level	303	UV limit; // escape character values >= this value when encoding
133	} enc_t;	304	} enc_t;
134		305
135	inline void	306	INLINE void
136	need (enc_t *enc, STRLEN len)	307	need (enc_t *enc, STRLEN len)
137	{	308	{
138	if (expect_false (enc->cur + len >= enc->end))	309	if (expect_false (enc->cur + len >= enc->end))
139	{	310	{
140	STRLEN cur = enc->cur - SvPVX (enc->sv);	311	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
141	SvGROW (enc->sv, cur + len + 1);	312	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
142	enc->cur = SvPVX (enc->sv) + cur;	313	enc->cur = SvPVX (enc->sv) + cur;
143	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	314	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
144	}	315	}
145	}	316	}
146		317
147	inline void	318	INLINE void
148	encode_ch (enc_t *enc, char ch)	319	encode_ch (enc_t *enc, char ch)
149	{	320	{
150	need (enc, 1);	321	need (enc, 1);
151	*enc->cur++ = ch;	322	*enc->cur++ = ch;
152	}	323	}
…		…
206	{	377	{
207	uch = ch;	378	uch = ch;
208	clen = 1;	379	clen = 1;
209	}	380	}
210		381
211	if (uch > 0x10FFFFUL)	382	if (uch < 0x80/0x20/ \|\| uch >= enc->limit)
212	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
213
214	if (uch < 0x80 \|\| enc->json.flags & F_ASCII \|\| (enc->json.flags & F_LATIN1 && uch > 0xFF))
215	{	383	{
216	if (uch > 0xFFFFUL)	384	if (uch >= 0x10000UL)
217	{	385	{
		386	if (uch >= 0x110000UL)
		387	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		388
218	need (enc, len += 11);	389	need (enc, len += 11);
219	sprintf (enc->cur, "\\u%04x\\u%04x",	390	sprintf (enc->cur, "\\u%04x\\u%04x",
220	(int)((uch - 0x10000) / 0x400 + 0xD800),	391	(int)((uch - 0x10000) / 0x400 + 0xD800),
221	(int)((uch - 0x10000) % 0x400 + 0xDC00));	392	(int)((uch - 0x10000) % 0x400 + 0xDC00));
222	enc->cur += 12;	393	enc->cur += 12;
223	}	394	}
224	else	395	else
225	{	396	{
226	static char hexdigit [16] = "0123456789abcdef";
227	need (enc, len += 5);	397	need (enc, len += 5);
228	*enc->cur++ = '\\';	398	*enc->cur++ = '\\';
229	*enc->cur++ = 'u';	399	*enc->cur++ = 'u';
230	*enc->cur++ = hexdigit [ uch >> 12 ];	400	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
231	*enc->cur++ = hexdigit [(uch >> 8) & 15];	401	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
232	*enc->cur++ = hexdigit [(uch >> 4) & 15];	402	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
233	*enc->cur++ = hexdigit [(uch >> 0) & 15];	403	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
234	}	404	}
235		405
236	str += clen;	406	str += clen;
237	}	407	}
238	else if (enc->json.flags & F_LATIN1)	408	else if (enc->json.flags & F_LATIN1)
…		…
250	while (--clen);	420	while (--clen);
251	}	421	}
252	else	422	else
253	{	423	{
254	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	424	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
255	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	425	enc->cur = encode_utf8 (enc->cur, uch);
256	++str;	426	++str;
257	}	427	}
258	}	428	}
259	}	429	}
260	}	430	}
261		431
262	--len;	432	--len;
263	}	433	}
264	}	434	}
265		435
266	inline void	436	INLINE void
267	encode_indent (enc_t *enc)	437	encode_indent (enc_t *enc)
268	{	438	{
269	if (enc->json.flags & F_INDENT)	439	if (enc->json.flags & F_INDENT)
270	{	440	{
271	int spaces = enc->indent * INDENT_STEP;	441	int spaces = enc->indent * INDENT_STEP;
…		…
274	memset (enc->cur, ' ', spaces);	444	memset (enc->cur, ' ', spaces);
275	enc->cur += spaces;	445	enc->cur += spaces;
276	}	446	}
277	}	447	}
278		448
279	inline void	449	INLINE void
280	encode_space (enc_t *enc)	450	encode_space (enc_t *enc)
281	{	451	{
282	need (enc, 1);	452	need (enc, 1);
283	encode_ch (enc, ' ');	453	encode_ch (enc, ' ');
284	}	454	}
285		455
286	inline void	456	INLINE void
287	encode_nl (enc_t *enc)	457	encode_nl (enc_t *enc)
288	{	458	{
289	if (enc->json.flags & F_INDENT)	459	if (enc->json.flags & F_INDENT)
290	{	460	{
291	need (enc, 1);	461	need (enc, 1);
292	encode_ch (enc, '\n');	462	encode_ch (enc, '\n');
293	}	463	}
294	}	464	}
295		465
296	inline void	466	INLINE void
297	encode_comma (enc_t *enc)	467	encode_comma (enc_t *enc)
298	{	468	{
299	encode_ch (enc, ',');	469	encode_ch (enc, ',');
300		470
301	if (enc->json.flags & F_INDENT)	471	if (enc->json.flags & F_INDENT)
…		…
309	static void	479	static void
310	encode_av (enc_t enc, AV av)	480	encode_av (enc_t enc, AV av)
311	{	481	{
312	int i, len = av_len (av);	482	int i, len = av_len (av);
313		483
314	if (enc->indent >= enc->maxdepth)	484	if (enc->indent >= enc->json.max_depth)
315	croak ("data structure too deep (hit recursion limit)");	485	croak (ERR_NESTING_EXCEEDED);
316		486
317	encode_ch (enc, '[');	487	encode_ch (enc, '[');
318		488
319	if (len >= 0)	489	if (len >= 0)
320	{	490	{
…		…
394		564
395	static void	565	static void
396	encode_hv (enc_t enc, HV hv)	566	encode_hv (enc_t enc, HV hv)
397	{	567	{
398	HE *he;	568	HE *he;
399	int count;
400		569
401	if (enc->indent >= enc->maxdepth)	570	if (enc->indent >= enc->json.max_depth)
402	croak ("data structure too deep (hit recursion limit)");	571	croak (ERR_NESTING_EXCEEDED);
403		572
404	encode_ch (enc, '{');	573	encode_ch (enc, '{');
405		574
406	// for canonical output we have to sort by keys first	575	// for canonical output we have to sort by keys first
407	// actually, this is mostly due to the stupid so-called	576	// actually, this is mostly due to the stupid so-called
408	// security workaround added somewhere in 5.8.x.	577	// security workaround added somewhere in 5.8.x
409	// that randomises hash orderings	578	// that randomises hash orderings
410	if (enc->json.flags & F_CANONICAL)	579	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
411	{	580	{
412	int count = hv_iterinit (hv);	581	int count = hv_iterinit (hv);
413		582
414	if (SvMAGICAL (hv))	583	if (SvMAGICAL (hv))
415	{	584	{
…		…
426	}	595	}
427		596
428	if (count)	597	if (count)
429	{	598	{
430	int i, fast = 1;	599	int i, fast = 1;
431	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	600	HE *hes_stack [STACK_HES];
432	HE *hes = _alloca (count sizeof (HE));	601	HE **hes = hes_stack;
433	#else	602
434	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode	603	// allocate larger arrays on the heap
435	#endif	604	if (count > STACK_HES)
		605	{
		606	SV sv = sv_2mortal (NEWSV (0, count sizeof (*hes)));
		607	hes = (HE **)SvPVX (sv);
		608	}
436		609
437	i = 0;	610	i = 0;
438	while ((he = hv_iternext (hv)))	611	while ((he = hv_iternext (hv)))
439	{	612	{
440	hes [i++] = he;	613	hes [i++] = he;
…		…
588		761
589	if (len == 1 && *pv == '1')	762	if (len == 1 && *pv == '1')
590	encode_str (enc, "true", 4, 0);	763	encode_str (enc, "true", 4, 0);
591	else if (len == 1 && *pv == '0')	764	else if (len == 1 && *pv == '0')
592	encode_str (enc, "false", 5, 0);	765	encode_str (enc, "false", 5, 0);
		766	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		767	encode_str (enc, "null", 4, 0);
593	else	768	else
594	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	769	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
595	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	770	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
596	}	771	}
		772	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		773	encode_str (enc, "null", 4, 0);
597	else	774	else
598	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	775	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
599	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	776	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
600	}	777	}
601		778
…		…
619	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	796	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
620	enc->cur += strlen (enc->cur);	797	enc->cur += strlen (enc->cur);
621	}	798	}
622	else if (SvIOKp (sv))	799	else if (SvIOKp (sv))
623	{	800	{
624	// we assume we can always read an IV as a UV	801	// we assume we can always read an IV as a UV and vice versa
625	if (SvUV (sv) & ~(UV)0x7fff)	802	// we assume two's complement
626	{	803	// we assume no aliasing issues in the union
627	// large integer, use the (rather slow) snprintf way.	804	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
628	need (enc, sizeof (UV) * 3);	805	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
629	enc->cur +=
630	SvIsUV(sv)
631	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
632	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
633	}
634	else
635	{	806	{
636	// optimise the "small number case"	807	// optimise the "small number case"
637	// code will likely be branchless and use only a single multiplication	808	// code will likely be branchless and use only a single multiplication
		809	// works for numbers up to 59074
638	I32 i = SvIV (sv);	810	I32 i = SvIVX (sv);
639	U32 u;	811	U32 u;
640	char digit, nz = 0;	812	char digit, nz = 0;
641		813
642	need (enc, 6);	814	need (enc, 6);
643		815
…		…
649		821
650	// now output digit by digit, each time masking out the integer part	822	// now output digit by digit, each time masking out the integer part
651	// and multiplying by 5 while moving the decimal point one to the right,	823	// and multiplying by 5 while moving the decimal point one to the right,
652	// resulting in a net multiplication by 10.	824	// resulting in a net multiplication by 10.
653	// we always write the digit to memory but conditionally increment	825	// we always write the digit to memory but conditionally increment
654	// the pointer, to ease the usage of conditional move instructions.	826	// the pointer, to enable the use of conditional move instructions.
655	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffff) 5;	827	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffffUL) 5;
656	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffff) 5;	828	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffffUL) 5;
657	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffff) 5;	829	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffffUL) 5;
658	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffff) 5;	830	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffffUL) 5;
659	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'	831	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
660	}	832	}
		833	else
		834	{
		835	// large integer, use the (rather slow) snprintf way.
		836	need (enc, IVUV_MAXCHARS);
		837	enc->cur +=
		838	SvIsUV(sv)
		839	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
		840	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		841	}
661	}	842	}
662	else if (SvROK (sv))	843	else if (SvROK (sv))
663	encode_rv (enc, SvRV (sv));	844	encode_rv (enc, SvRV (sv));
664	else if (!SvOK (sv))	845	else if (!SvOK (sv) \|\| enc->json.flags & F_ALLOW_UNKNOWN)
665	encode_str (enc, "null", 4, 0);	846	encode_str (enc, "null", 4, 0);
666	else	847	else
667	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	848	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
668	SvPV_nolen (sv), SvFLAGS (sv));	849	SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
669	}	850	}
670		851
671	static SV *	852	static SV *
672	encode_json (SV scalar, JSON json)	853	encode_json (SV scalar, JSON json)
673	{	854	{
…		…
679	enc.json = *json;	860	enc.json = *json;
680	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	861	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
681	enc.cur = SvPVX (enc.sv);	862	enc.cur = SvPVX (enc.sv);
682	enc.end = SvEND (enc.sv);	863	enc.end = SvEND (enc.sv);
683	enc.indent = 0;	864	enc.indent = 0;
684	enc.maxdepth = DEC_DEPTH (enc.json.flags);	865	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		866	: enc.json.flags & F_LATIN1 ? 0x000100UL
		867	: 0x110000UL;
685		868
686	SvPOK_only (enc.sv);	869	SvPOK_only (enc.sv);
687	encode_sv (&enc, scalar);	870	encode_sv (&enc, scalar);
		871	encode_nl (&enc);
688		872
689	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	873	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
690	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	874	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
691		875
692	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))	876	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))
…		…
710	JSON json;	894	JSON json;
711	U32 depth; // recursion depth	895	U32 depth; // recursion depth
712	U32 maxdepth; // recursion depth limit	896	U32 maxdepth; // recursion depth limit
713	} dec_t;	897	} dec_t;
714		898
715	inline void	899	INLINE void
716	decode_comment (dec_t *dec)	900	decode_comment (dec_t *dec)
717	{	901	{
718	// only '#'-style comments allowed a.t.m.	902	// only '#'-style comments allowed a.t.m.
719		903
720	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)	904	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)
721	++dec->cur;	905	++dec->cur;
722	}	906	}
723		907
724	inline void	908	INLINE void
725	decode_ws (dec_t *dec)	909	decode_ws (dec_t *dec)
726	{	910	{
727	for (;;)	911	for (;;)
728	{	912	{
729	char ch = *dec->cur;	913	char ch = *dec->cur;
…		…
753	if (*dec->cur != ch) \	937	if (*dec->cur != ch) \
754	ERR (# ch " expected"); \	938	ERR (# ch " expected"); \
755	++dec->cur; \	939	++dec->cur; \
756	SE	940	SE
757		941
758	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	942	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
759	#define DEC_DEC_DEPTH --dec->depth	943	#define DEC_DEC_DEPTH --dec->depth
760		944
761	static SV decode_sv (dec_t dec);	945	static SV decode_sv (dec_t dec);
762		946
763	static signed char decode_hexdigit[256];	947	static signed char decode_hexdigit[256];
…		…
855		1039
856	if (hi >= 0x80)	1040	if (hi >= 0x80)
857	{	1041	{
858	utf8 = 1;	1042	utf8 = 1;
859		1043
860	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	1044	cur = encode_utf8 (cur, hi);
861	}	1045	}
862	else	1046	else
863	*cur++ = hi;	1047	*cur++ = hi;
864	}	1048	}
865	break;	1049	break;
…		…
867	default:	1051	default:
868	--dec_cur;	1052	--dec_cur;
869	ERR ("illegal backslash escape sequence in string");	1053	ERR ("illegal backslash escape sequence in string");
870	}	1054	}
871	}	1055	}
872	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	1056	else if (expect_true (ch >= 0x20 && ch < 0x80))
873	*cur++ = ch;	1057	*cur++ = ch;
874	else if (ch >= 0x80)	1058	else if (ch >= 0x80)
875	{	1059	{
876	STRLEN clen;	1060	STRLEN clen;
877	UV uch;
878		1061
879	--dec_cur;	1062	--dec_cur;
880		1063
881	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);	1064	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
882	if (clen == (STRLEN)-1)	1065	if (clen == (STRLEN)-1)
883	ERR ("malformed UTF-8 character in JSON string");	1066	ERR ("malformed UTF-8 character in JSON string");
884		1067
885	do	1068	do
886	cur++ = dec_cur++;	1069	cur++ = dec_cur++;
…		…
903	{	1086	{
904	STRLEN len = cur - buf;	1087	STRLEN len = cur - buf;
905		1088
906	if (sv)	1089	if (sv)
907	{	1090	{
908	SvGROW (sv, SvCUR (sv) + len + 1);	1091	STRLEN cur = SvCUR (sv);
		1092
		1093	if (SvLEN (sv) <= cur + len)
		1094	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1095
909	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	1096	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
910	SvCUR_set (sv, SvCUR (sv) + len);	1097	SvCUR_set (sv, SvCUR (sv) + len);
911	}	1098	}
912	else	1099	else
913	sv = newSVpvn (buf, len);	1100	sv = newSVpvn (buf, len);
…		…
1000		1187
1001	if (!is_nv)	1188	if (!is_nv)
1002	{	1189	{
1003	int len = dec->cur - start;	1190	int len = dec->cur - start;
1004		1191
1005	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1192	// special case the rather common 1..5-digit-int case
1006	if (*start == '-')	1193	if (*start == '-')
1007	switch (len)	1194	switch (len)
1008	{	1195	{
1009	case 2: return newSViv (-( start [1] - '0' * 1));	1196	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1010	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1197	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1011	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));	1198	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1012	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));	1199	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1200	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1013	}	1201	}
1014	else	1202	else
1015	switch (len)	1203	switch (len)
1016	{	1204	{
1017	case 1: return newSViv ( start [0] - '0' * 1);	1205	case 1: return newSViv ( start [0] - '0' * 1);
1018	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1206	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1019	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);	1207	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1020	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);	1208	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1209	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1021	}	1210	}
1022		1211
1023	{	1212	{
1024	UV uv;	1213	UV uv;
1025	int numtype = grok_number (start, len, &uv);	1214	int numtype = grok_number (start, len, &uv);
…		…
1034	}	1223	}
1035		1224
1036	len -= *start == '-' ? 1 : 0;	1225	len -= *start == '-' ? 1 : 0;
1037		1226
1038	// does not fit into IV or UV, try NV	1227	// does not fit into IV or UV, try NV
1039	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)	1228	if (len <= NV_DIG)
1040	#if defined (LDBL_DIG)
1041	\|\| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1042	#endif
1043	)
1044	// fits into NV without loss of precision	1229	// fits into NV without loss of precision
1045	return newSVnv (Atof (start));	1230	return newSVnv (json_atof (start));
1046		1231
1047	// everything else fails, convert it to a string	1232	// everything else fails, convert it to a string
1048	return newSVpvn (start, dec->cur - start);	1233	return newSVpvn (start, dec->cur - start);
1049	}	1234	}
1050		1235
1051	// loss of precision here	1236	// loss of precision here
1052	return newSVnv (Atof (start));	1237	return newSVnv (json_atof (start));
1053		1238
1054	fail:	1239	fail:
1055	return 0;	1240	return 0;
1056	}	1241	}
1057		1242
…		…
1133	char *p = dec->cur;	1318	char *p = dec->cur;
1134	char *e = p + 24; // only try up to 24 bytes	1319	char *e = p + 24; // only try up to 24 bytes
1135		1320
1136	for (;;)	1321	for (;;)
1137	{	1322	{
1138	// the >= 0x80 is true on most architectures	1323	// the >= 0x80 is false on most architectures
1139	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')	1324	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
1140	{	1325	{
1141	// slow path, back up and use decode_str	1326	// slow path, back up and use decode_str
1142	SV *key = decode_str (dec);	1327	SV *key = decode_str (dec);
1143	if (!key)	1328	if (!key)
…		…
1226	dSP;	1411	dSP;
1227	int count;	1412	int count;
1228		1413
1229	ENTER; SAVETMPS; PUSHMARK (SP);	1414	ENTER; SAVETMPS; PUSHMARK (SP);
1230	XPUSHs (HeVAL (he));	1415	XPUSHs (HeVAL (he));
		1416	sv_2mortal (sv);
1231		1417
1232	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;	1418	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1233		1419
1234	if (count == 1)	1420	if (count == 1)
1235	{	1421	{
1236	sv = newSVsv (POPs);	1422	sv = newSVsv (POPs);
1237	FREETMPS; LEAVE;	1423	FREETMPS; LEAVE;
1238	return sv;	1424	return sv;
1239	}	1425	}
1240		1426
		1427	SvREFCNT_inc (sv);
1241	FREETMPS; LEAVE;	1428	FREETMPS; LEAVE;
1242	}	1429	}
1243	}	1430	}
1244		1431
1245	if (dec->json.cb_object)	1432	if (dec->json.cb_object)
…		…
1274		1461
1275	static SV *	1462	static SV *
1276	decode_sv (dec_t *dec)	1463	decode_sv (dec_t *dec)
1277	{	1464	{
1278	// the beauty of JSON: you need exactly one character lookahead	1465	// the beauty of JSON: you need exactly one character lookahead
1279	// to parse anything.	1466	// to parse everything.
1280	switch (*dec->cur)	1467	switch (*dec->cur)
1281	{	1468	{
1282	case '"': ++dec->cur; return decode_str (dec);	1469	case '"': ++dec->cur; return decode_str (dec);
1283	case '[': ++dec->cur; return decode_av (dec);	1470	case '[': ++dec->cur; return decode_av (dec);
1284	case '{': ++dec->cur; return decode_hv (dec);	1471	case '{': ++dec->cur; return decode_hv (dec);
1285		1472
1286	case '-':	1473	case '-':
1287	case '0': case '1': case '2': case '3': case '4':	1474	case '0': case '1': case '2': case '3': case '4':
1288	case '5': case '6': case '7': case '8': case '9':	1475	case '5': case '6': case '7': case '8': case '9':
1289	return decode_num (dec);	1476	return decode_num (dec);
…		…
1291	case 't':	1478	case 't':
1292	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1479	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1293	{	1480	{
1294	dec->cur += 4;	1481	dec->cur += 4;
1295	#if JSON_SLOW	1482	#if JSON_SLOW
1296	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);	1483	json_true = get_bool ("JSON::XS::true");
1297	#endif	1484	#endif
1298	return SvREFCNT_inc (json_true);	1485	return newSVsv (json_true);
1299	}	1486	}
1300	else	1487	else
1301	ERR ("'true' expected");	1488	ERR ("'true' expected");
1302		1489
1303	break;	1490	break;
…		…
1305	case 'f':	1492	case 'f':
1306	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1493	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1307	{	1494	{
1308	dec->cur += 5;	1495	dec->cur += 5;
1309	#if JSON_SLOW	1496	#if JSON_SLOW
1310	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1497	json_false = get_bool ("JSON::XS::false");
1311	#endif	1498	#endif
1312	return SvREFCNT_inc (json_false);	1499	return newSVsv (json_false);
1313	}	1500	}
1314	else	1501	else
1315	ERR ("'false' expected");	1502	ERR ("'false' expected");
1316		1503
1317	break;	1504	break;
…		…
1335	fail:	1522	fail:
1336	return 0;	1523	return 0;
1337	}	1524	}
1338		1525
1339	static SV *	1526	static SV *
1340	decode_json (SV string, JSON json, UV *offset_return)	1527	decode_json (SV string, JSON json, char **offset_return)
1341	{	1528	{
1342	dec_t dec;	1529	dec_t dec;
1343	UV offset;
1344	SV *sv;	1530	SV *sv;
1345		1531
		1532	/* work around bugs in 5.10 where manipulating magic values
		1533	* will perl ignore the magic in subsequent accesses.
		1534	* also make a copy of non-PV values, to get them into a clean
		1535	* state (SvPV should do that, but it's buggy, see below).
		1536	*/
1346	SvGETMAGIC (string);	1537	/SvGETMAGIC (string);/
		1538	if (SvMAGICAL (string) \|\| !SvPOK (string))
		1539	string = sv_2mortal (newSVsv (string));
		1540
1347	SvUPGRADE (string, SVt_PV);	1541	SvUPGRADE (string, SVt_PV);
1348		1542
1349	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))	1543	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1544	* assertion with -DDEBUGGING, although SvCUR is documented to
		1545	* return the xpv_cur field which certainly exists after upgrading.
		1546	* according to nicholas clark, calling SvPOK fixes this.
		1547	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1548	* and hope for the best.
		1549	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1550	* assertion business is seriously broken, try yet another workaround
		1551	* for the broken -DDEBUGGING.
		1552	*/
		1553	{
		1554	#ifdef DEBUGGING
		1555	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1556	#else
		1557	STRLEN offset = SvCUR (string);
		1558	#endif
		1559
		1560	if (offset > json->max_size && json->max_size)
1350	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",	1561	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1351	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));	1562	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1563	}
1352		1564
1353	if (json->flags & F_UTF8)	1565	if (DECODE_WANTS_OCTETS (json))
1354	sv_utf8_downgrade (string, 0);	1566	sv_utf8_downgrade (string, 0);
1355	else	1567	else
1356	sv_utf8_upgrade (string);	1568	sv_utf8_upgrade (string);
1357		1569
1358	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1570	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1359		1571
1360	dec.json = *json;	1572	dec.json = *json;
1361	dec.cur = SvPVX (string);	1573	dec.cur = SvPVX (string);
1362	dec.end = SvEND (string);	1574	dec.end = SvEND (string);
1363	dec.err = 0;	1575	dec.err = 0;
1364	dec.depth = 0;	1576	dec.depth = 0;
1365	dec.maxdepth = DEC_DEPTH (dec.json.flags);
1366		1577
1367	if (dec.json.cb_object \|\| dec.json.cb_sk_object)	1578	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
1368	dec.json.flags \|= F_HOOK;	1579	dec.json.flags \|= F_HOOK;
1369		1580
1370	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1581	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
1371		1582
1372	decode_ws (&dec);	1583	decode_ws (&dec);
1373	sv = decode_sv (&dec);	1584	sv = decode_sv (&dec);
		1585
		1586	if (offset_return)
		1587	*offset_return = dec.cur;
1374		1588
1375	if (!(offset_return \|\| !sv))	1589	if (!(offset_return \|\| !sv))
1376	{	1590	{
1377	// check for trailing garbage	1591	// check for trailing garbage
1378	decode_ws (&dec);	1592	decode_ws (&dec);
…		…
1381	{	1595	{
1382	dec.err = "garbage after JSON object";	1596	dec.err = "garbage after JSON object";
1383	SvREFCNT_dec (sv);	1597	SvREFCNT_dec (sv);
1384	sv = 0;	1598	sv = 0;
1385	}	1599	}
1386	}
1387
1388	if (offset_return \|\| !sv)
1389	{
1390	offset = dec.json.flags & F_UTF8
1391	? dec.cur - SvPVX (string)
1392	: utf8_distance (dec.cur, SvPVX (string));
1393
1394	if (offset_return)
1395	*offset_return = offset;
1396	}	1600	}
1397		1601
1398	if (!sv)	1602	if (!sv)
1399	{	1603	{
1400	SV *uni = sv_newmortal ();	1604	SV *uni = sv_newmortal ();
…		…
1406	SAVEVPTR (PL_curcop);	1610	SAVEVPTR (PL_curcop);
1407	PL_curcop = &cop;	1611	PL_curcop = &cop;
1408	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1612	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1409	LEAVE;	1613	LEAVE;
1410		1614
1411	croak ("%s, at character offset %d [\"%s\"]",	1615	croak ("%s, at character offset %d (before \"%s\")",
1412	dec.err,	1616	dec.err,
1413	(int)offset,	1617	(int)ptr_to_index (string, dec.cur),
1414	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1618	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1415	}	1619	}
1416		1620
1417	sv = sv_2mortal (sv);	1621	sv = sv_2mortal (sv);
1418		1622
1419	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))	1623	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1420	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1624	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1421		1625
1422	return sv;	1626	return sv;
		1627	}
		1628
		1629	/////////////////////////////////////////////////////////////////////////////
		1630	// incremental parser
		1631
		1632	static void
		1633	incr_parse (JSON *self)
		1634	{
		1635	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1636
		1637	// the state machine here is a bit convoluted and could be simplified a lot
		1638	// but this would make it slower, so...
		1639
		1640	for (;;)
		1641	{
		1642	//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
		1643	switch (self->incr_mode)
		1644	{
		1645	// only used for initial whitespace skipping
		1646	case INCR_M_WS:
		1647	for (;;)
		1648	{
		1649	if (*p > 0x20)
		1650	{
		1651	if (*p == '#')
		1652	{
		1653	self->incr_mode = INCR_M_C0;
		1654	goto incr_m_c;
		1655	}
		1656	else
		1657	{
		1658	self->incr_mode = INCR_M_JSON;
		1659	goto incr_m_json;
		1660	}
		1661	}
		1662	else if (!*p)
		1663	goto interrupt;
		1664
		1665	++p;
		1666	}
		1667
		1668	// skip a single char inside a string (for \\-processing)
		1669	case INCR_M_BS:
		1670	if (!*p)
		1671	goto interrupt;
		1672
		1673	++p;
		1674	self->incr_mode = INCR_M_STR;
		1675	goto incr_m_str;
		1676
		1677	// inside #-style comments
		1678	case INCR_M_C0:
		1679	case INCR_M_C1:
		1680	incr_m_c:
		1681	for (;;)
		1682	{
		1683	if (*p == '\n')
		1684	{
		1685	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1686	break;
		1687	}
		1688	else if (!*p)
		1689	goto interrupt;
		1690
		1691	++p;
		1692	}
		1693
		1694	break;
		1695
		1696	// inside a string
		1697	case INCR_M_STR:
		1698	incr_m_str:
		1699	for (;;)
		1700	{
		1701	if (*p == '"')
		1702	{
		1703	++p;
		1704	self->incr_mode = INCR_M_JSON;
		1705
		1706	if (!self->incr_nest)
		1707	goto interrupt;
		1708
		1709	goto incr_m_json;
		1710	}
		1711	else if (*p == '\\')
		1712	{
		1713	++p; // "virtually" consumes character after \
		1714
		1715	if (!*p) // if at end of string we have to switch modes
		1716	{
		1717	self->incr_mode = INCR_M_BS;
		1718	goto interrupt;
		1719	}
		1720	}
		1721	else if (!*p)
		1722	goto interrupt;
		1723
		1724	++p;
		1725	}
		1726
		1727	// after initial ws, outside string
		1728	case INCR_M_JSON:
		1729	incr_m_json:
		1730	for (;;)
		1731	{
		1732	switch (*p++)
		1733	{
		1734	case 0:
		1735	--p;
		1736	goto interrupt;
		1737
		1738	case 0x09:
		1739	case 0x0a:
		1740	case 0x0d:
		1741	case 0x20:
		1742	if (!self->incr_nest)
		1743	{
		1744	--p; // do not eat the whitespace, let the next round do it
		1745	goto interrupt;
		1746	}
		1747	break;
		1748
		1749	case '"':
		1750	self->incr_mode = INCR_M_STR;
		1751	goto incr_m_str;
		1752
		1753	case '[':
		1754	case '{':
		1755	if (++self->incr_nest > self->max_depth)
		1756	croak (ERR_NESTING_EXCEEDED);
		1757	break;
		1758
		1759	case ']':
		1760	case '}':
		1761	if (--self->incr_nest <= 0)
		1762	goto interrupt;
		1763	break;
		1764
		1765	case '#':
		1766	self->incr_mode = INCR_M_C1;
		1767	goto incr_m_c;
		1768	}
		1769	}
		1770	}
		1771
		1772	modechange:
		1773	;
		1774	}
		1775
		1776	interrupt:
		1777	self->incr_pos = p - SvPVX (self->incr_text);
		1778	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1779	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1423	}	1780	}
1424		1781
1425	/////////////////////////////////////////////////////////////////////////////	1782	/////////////////////////////////////////////////////////////////////////////
1426	// XS interface functions	1783	// XS interface functions
1427		1784
…		…
1439	: -1;	1796	: -1;
1440		1797
1441	json_stash = gv_stashpv ("JSON::XS" , 1);	1798	json_stash = gv_stashpv ("JSON::XS" , 1);
1442	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);	1799	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1443		1800
1444	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );	1801	json_true = get_bool ("JSON::XS::true");
1445	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1802	json_false = get_bool ("JSON::XS::false");
		1803
		1804	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1446	}	1805	}
1447		1806
1448	PROTOTYPES: DISABLE	1807	PROTOTYPES: DISABLE
1449		1808
1450	void CLONE (...)	1809	void CLONE (...)
…		…
1453	json_boolean_stash = 0;	1812	json_boolean_stash = 0;
1454		1813
1455	void new (char *klass)	1814	void new (char *klass)
1456	PPCODE:	1815	PPCODE:
1457	{	1816	{
1458	SV *pv = NEWSV (0, sizeof (JSON));	1817	SV *pv = NEWSV (0, sizeof (JSON));
1459	SvPOK_only (pv);	1818	SvPOK_only (pv);
1460	Zero (SvPVX (pv), 1, JSON);	1819	json_init ((JSON *)SvPVX (pv));
1461	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1462	XPUSHs (sv_2mortal (sv_bless (	1820	XPUSHs (sv_2mortal (sv_bless (
1463	newRV_noinc (pv),	1821	newRV_noinc (pv),
1464	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)	1822	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1465	)));	1823	)));
1466	}	1824	}
…		…
1478	allow_nonref = F_ALLOW_NONREF	1836	allow_nonref = F_ALLOW_NONREF
1479	shrink = F_SHRINK	1837	shrink = F_SHRINK
1480	allow_blessed = F_ALLOW_BLESSED	1838	allow_blessed = F_ALLOW_BLESSED
1481	convert_blessed = F_CONV_BLESSED	1839	convert_blessed = F_CONV_BLESSED
1482	relaxed = F_RELAXED	1840	relaxed = F_RELAXED
		1841	allow_unknown = F_ALLOW_UNKNOWN
1483	PPCODE:	1842	PPCODE:
1484	{	1843	{
1485	if (enable)	1844	if (enable)
1486	self->flags \|= ix;	1845	self->flags \|= ix;
1487	else	1846	else
…		…
1502	get_allow_nonref = F_ALLOW_NONREF	1861	get_allow_nonref = F_ALLOW_NONREF
1503	get_shrink = F_SHRINK	1862	get_shrink = F_SHRINK
1504	get_allow_blessed = F_ALLOW_BLESSED	1863	get_allow_blessed = F_ALLOW_BLESSED
1505	get_convert_blessed = F_CONV_BLESSED	1864	get_convert_blessed = F_CONV_BLESSED
1506	get_relaxed = F_RELAXED	1865	get_relaxed = F_RELAXED
		1866	get_allow_unknown = F_ALLOW_UNKNOWN
1507	PPCODE:	1867	PPCODE:
1508	XPUSHs (boolSV (self->flags & ix));	1868	XPUSHs (boolSV (self->flags & ix));
1509		1869
1510	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1870	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1511	PPCODE:	1871	PPCODE:
1512	{	1872	self->max_depth = max_depth;
1513	UV log2 = 0;
1514
1515	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1516
1517	while ((1UL << log2) < max_depth)
1518	++log2;
1519
1520	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1521
1522	XPUSHs (ST (0));	1873	XPUSHs (ST (0));
1523	}
1524		1874
1525	U32 get_max_depth (JSON *self)	1875	U32 get_max_depth (JSON *self)
1526	CODE:	1876	CODE:
1527	RETVAL = DEC_DEPTH (self->flags);	1877	RETVAL = self->max_depth;
1528	OUTPUT:	1878	OUTPUT:
1529	RETVAL	1879	RETVAL
1530		1880
1531	void max_size (JSON *self, UV max_size = 0)	1881	void max_size (JSON *self, U32 max_size = 0)
1532	PPCODE:	1882	PPCODE:
1533	{	1883	self->max_size = max_size;
1534	UV log2 = 0;
1535
1536	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1537	if (max_size == 1) max_size = 2;
1538
1539	while ((1UL << log2) < max_size)
1540	++log2;
1541
1542	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
1543
1544	XPUSHs (ST (0));	1884	XPUSHs (ST (0));
1545	}
1546		1885
1547	int get_max_size (JSON *self)	1886	int get_max_size (JSON *self)
1548	CODE:	1887	CODE:
1549	RETVAL = DEC_SIZE (self->flags);	1888	RETVAL = self->max_size;
1550	OUTPUT:	1889	OUTPUT:
1551	RETVAL	1890	RETVAL
1552		1891
1553	void filter_json_object (JSON self, SV cb = &PL_sv_undef)	1892	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1554	PPCODE:	1893	PPCODE:
…		…
1560	}	1899	}
1561		1900
1562	void filter_json_single_key_object (JSON self, SV key, SV *cb = &PL_sv_undef)	1901	void filter_json_single_key_object (JSON self, SV key, SV *cb = &PL_sv_undef)
1563	PPCODE:	1902	PPCODE:
1564	{	1903	{
1565	if (!self->cb_sk_object)	1904	if (!self->cb_sk_object)
1566	self->cb_sk_object = newHV ();	1905	self->cb_sk_object = newHV ();
1567		1906
1568	if (SvOK (cb))	1907	if (SvOK (cb))
1569	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);	1908	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1570	else	1909	else
…		…
1581	XPUSHs (ST (0));	1920	XPUSHs (ST (0));
1582	}	1921	}
1583		1922
1584	void encode (JSON self, SV scalar)	1923	void encode (JSON self, SV scalar)
1585	PPCODE:	1924	PPCODE:
1586	XPUSHs (encode_json (scalar, self));	1925	PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
		1926	XPUSHs (scalar);
1587		1927
1588	void decode (JSON self, SV jsonstr)	1928	void decode (JSON self, SV jsonstr)
1589	PPCODE:	1929	PPCODE:
1590	XPUSHs (decode_json (jsonstr, self, 0));	1930	PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
		1931	XPUSHs (jsonstr);
1591		1932
1592	void decode_prefix (JSON self, SV jsonstr)	1933	void decode_prefix (JSON self, SV jsonstr)
1593	PPCODE:	1934	PPCODE:
1594	{	1935	{
1595	UV offset;	1936	SV *sv;
		1937	char *offset;
		1938	PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1596	EXTEND (SP, 2);	1939	EXTEND (SP, 2);
1597	PUSHs (decode_json (jsonstr, self, &offset));	1940	PUSHs (sv);
1598	PUSHs (sv_2mortal (newSVuv (offset)));	1941	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		1942	}
		1943
		1944	void incr_parse (JSON self, SV jsonstr = 0)
		1945	PPCODE:
		1946	{
		1947	if (!self->incr_text)
		1948	self->incr_text = newSVpvn ("", 0);
		1949
		1950	/* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
		1951	if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
		1952	if (DECODE_WANTS_OCTETS (self))
		1953	{
		1954	if (self->incr_pos)
		1955	self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
		1956	(U8 *)SvPVX (self->incr_text) + self->incr_pos);
		1957
		1958	sv_utf8_downgrade (self->incr_text, 0);
		1959	}
		1960	else
		1961	{
		1962	sv_utf8_upgrade (self->incr_text);
		1963
		1964	if (self->incr_pos)
		1965	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1966	- (U8 *)SvPVX (self->incr_text);
		1967	}
		1968
		1969	// append data, if any
		1970	if (jsonstr)
		1971	{
		1972	/* make sure both strings have same encoding */
		1973	if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
		1974	if (SvUTF8 (jsonstr))
		1975	sv_utf8_downgrade (jsonstr, 0);
		1976	else
		1977	sv_utf8_upgrade (jsonstr);
		1978
		1979	/* and then just blindly append */
		1980	{
		1981	STRLEN len;
		1982	const char *str = SvPV (jsonstr, len);
		1983	STRLEN cur = SvCUR (self->incr_text);
		1984
		1985	if (SvLEN (self->incr_text) <= cur + len)
		1986	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1987
		1988	Move (str, SvEND (self->incr_text), len, char);
		1989	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1990	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1991	}
		1992	}
		1993
		1994	if (GIMME_V != G_VOID)
		1995	do
		1996	{
		1997	SV *sv;
		1998	char *offset;
		1999
		2000	if (!INCR_DONE (self))
		2001	{
		2002	incr_parse (self);
		2003
		2004	if (self->incr_pos > self->max_size && self->max_size)
		2005	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		2006	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		2007
		2008	if (!INCR_DONE (self))
		2009	{
		2010	// as an optimisation, do not accumulate white space in the incr buffer
		2011	if (self->incr_mode == INCR_M_WS && self->incr_pos)
		2012	{
		2013	self->incr_pos = 0;
		2014	SvCUR_set (self->incr_text, 0);
		2015	}
		2016
		2017	break;
		2018	}
		2019	}
		2020
		2021	PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
		2022	XPUSHs (sv);
		2023
		2024	self->incr_pos -= offset - SvPVX (self->incr_text);
		2025	self->incr_nest = 0;
		2026	self->incr_mode = 0;
		2027
		2028	sv_chop (self->incr_text, offset);
		2029	}
		2030	while (GIMME_V == G_ARRAY);
		2031	}
		2032
		2033	SV incr_text (JSON self)
		2034	ATTRS: lvalue
		2035	CODE:
		2036	{
		2037	if (self->incr_pos)
		2038	croak ("incr_text can not be called when the incremental parser already started parsing");
		2039
		2040	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		2041	}
		2042	OUTPUT:
		2043	RETVAL
		2044
		2045	void incr_skip (JSON *self)
		2046	CODE:
		2047	{
		2048	if (self->incr_pos)
		2049	{
		2050	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2051	self->incr_pos = 0;
		2052	self->incr_nest = 0;
		2053	self->incr_mode = 0;
		2054	}
		2055	}
		2056
		2057	void incr_reset (JSON *self)
		2058	CODE:
		2059	{
		2060	SvREFCNT_dec (self->incr_text);
		2061	self->incr_text = 0;
		2062	self->incr_pos = 0;
		2063	self->incr_nest = 0;
		2064	self->incr_mode = 0;
1599	}	2065	}
1600		2066
1601	void DESTROY (JSON *self)	2067	void DESTROY (JSON *self)
1602	CODE:	2068	CODE:
1603	SvREFCNT_dec (self->cb_sk_object);	2069	SvREFCNT_dec (self->cb_sk_object);
1604	SvREFCNT_dec (self->cb_object);	2070	SvREFCNT_dec (self->cb_object);
		2071	SvREFCNT_dec (self->incr_text);
1605		2072
1606	PROTOTYPES: ENABLE	2073	PROTOTYPES: ENABLE
1607		2074
1608	void encode_json (SV *scalar)	2075	void encode_json (SV *scalar)
		2076	ALIAS:
		2077	to_json_ = 0
		2078	encode_json = F_UTF8
1609	PPCODE:	2079	PPCODE:
1610	{	2080	{
1611	JSON json = { F_DEFAULT \| F_UTF8 };	2081	JSON json;
1612	XPUSHs (encode_json (scalar, &json));	2082	json_init (&json);
		2083	json.flags \|= ix;
		2084	PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
		2085	XPUSHs (scalar);
1613	}	2086	}
1614		2087
1615	void decode_json (SV *jsonstr)	2088	void decode_json (SV *jsonstr)
		2089	ALIAS:
		2090	from_json_ = 0
		2091	decode_json = F_UTF8
1616	PPCODE:	2092	PPCODE:
1617	{	2093	{
1618	JSON json = { F_DEFAULT \| F_UTF8 };	2094	JSON json;
1619	XPUSHs (decode_json (jsonstr, &json, 0));	2095	json_init (&json);
		2096	json.flags \|= ix;
		2097	PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
		2098	XPUSHs (jsonstr);
1620	}	2099	}
1621		2100

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.69 by root, Wed Dec 5 10:59:28 2007 UTC vs. Revision 1.115 by root, Fri Dec 21 05:16:19 2012 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.69 by root, Wed Dec 5 10:59:28 2007 UTC vs.
Revision 1.115 by root, Fri Dec 21 05:16:19 2012 UTC