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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.55 by root, Mon Jul 23 22:57:40 2007 UTC vs.
Revision 1.99 by root, Sat Aug 8 10:06:02 2009 UTC

…		…
1	#include "EXTERN.h"	1	#include "EXTERN.h"
2	#include "perl.h"	2	#include "perl.h"
3	#include "XSUB.h"	3	#include "XSUB.h"
4		4
5	#include "assert.h"	5	#include <assert.h>
6	#include "string.h"	6	#include <string.h>
7	#include "stdlib.h"	7	#include <stdlib.h>
8	#include "stdio.h"	8	#include <stdio.h>
		9	#include <limits.h>
		10	#include <float.h>
9		11
10	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	12	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
11	# define snprintf _snprintf // C compilers have this in stdio.h	13	# define snprintf _snprintf // C compilers have this in stdio.h
12	#endif	14	#endif
13		15
14	// 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
15	// guarentees, though. if it breaks, you get to keep the pieces.	17	// guarantees, though. if it breaks, you get to keep the pieces.
16	#ifndef UTF8_MAXBYTES	18	#ifndef UTF8_MAXBYTES
17	# define UTF8_MAXBYTES 13	19	# define UTF8_MAXBYTES 13
18	#endif	20	#endif
		21
		22	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2)
19		23
20	#define F_ASCII 0x00000001UL	24	#define F_ASCII 0x00000001UL
21	#define F_LATIN1 0x00000002UL	25	#define F_LATIN1 0x00000002UL
22	#define F_UTF8 0x00000004UL	26	#define F_UTF8 0x00000004UL
23	#define F_INDENT 0x00000008UL	27	#define F_INDENT 0x00000008UL
26	#define F_SPACE_AFTER 0x00000040UL	30	#define F_SPACE_AFTER 0x00000040UL
27	#define F_ALLOW_NONREF 0x00000100UL	31	#define F_ALLOW_NONREF 0x00000100UL
28	#define F_SHRINK 0x00000200UL	32	#define F_SHRINK 0x00000200UL
29	#define F_ALLOW_BLESSED 0x00000400UL	33	#define F_ALLOW_BLESSED 0x00000400UL
30	#define F_CONV_BLESSED 0x00000800UL	34	#define F_CONV_BLESSED 0x00000800UL
31	#define F_MAXDEPTH 0xf8000000UL
32	#define S_MAXDEPTH 27
33	#define F_MAXSIZE 0x01f00000UL	35	#define F_RELAXED 0x00001000UL
34	#define S_MAXSIZE 20	36	#define F_ALLOW_UNKNOWN 0x00002000UL
35	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing	37	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
36		38
37	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
38	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
39
40	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER	39	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER
41	#define F_DEFAULT (9UL << S_MAXDEPTH)
42		40
43	#define INIT_SIZE 32 // initial scalar size to be allocated	41	#define INIT_SIZE 32 // initial scalar size to be allocated
44	#define INDENT_STEP 3 // spaces per indentation level	42	#define INDENT_STEP 3 // spaces per indentation level
45		43
46	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size	44	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
47		45
48	#define SB do {	46	#define SB do {
49	#define SE } while (0)	47	#define SE } while (0)
50		48
51	#if __GNUC__ >= 3	49	#if __GNUC__ >= 3
52	# define expect(expr,value) __builtin_expect ((expr),(value))	50	# define expect(expr,value) __builtin_expect ((expr), (value))
53	# define inline inline	51	# define INLINE static inline
54	#else	52	#else
55	# define expect(expr,value) (expr)	53	# define expect(expr,value) (expr)
56	# define inline static	54	# define INLINE static
57	#endif	55	#endif
58		56
59	#define expect_false(expr) expect ((expr) != 0, 0)	57	#define expect_false(expr) expect ((expr) != 0, 0)
60	#define expect_true(expr) expect ((expr) != 0, 1)	58	#define expect_true(expr) expect ((expr) != 0, 1)
61		59
		60	#define IN_RANGE_INC(type,val,beg,end) \
		61	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		62	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		63
		64	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
		65
		66	#ifdef USE_ITHREADS
		67	# define JSON_SLOW 1
		68	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		69	#else
		70	# define JSON_SLOW 0
		71	# define JSON_STASH json_stash
		72	#endif
		73
62	static HV json_stash, json_boolean_stash; // JSON::XS::	74	static HV json_stash, json_boolean_stash; // JSON::XS::
63	static SV json_true, json_false;	75	static SV json_true, json_false;
64		76
		77	enum {
		78	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		79	INCR_M_STR, // inside string
		80	INCR_M_BS, // inside backslash
		81	INCR_M_JSON // outside anything, count nesting
		82	};
		83
		84	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		85
65	typedef struct {	86	typedef struct {
66	U32 flags;	87	U32 flags;
		88	U32 max_depth;
		89	STRLEN max_size;
		90
67	SV *cb_object;	91	SV *cb_object;
68	HV *cb_sk_object;	92	HV *cb_sk_object;
		93
		94	// for the incremental parser
		95	SV *incr_text; // the source text so far
		96	STRLEN incr_pos; // the current offset into the text
		97	int incr_nest; // {[]}-nesting level
		98	unsigned char incr_mode;
69	} JSON;	99	} JSON;
		100
		101	INLINE void
		102	json_init (JSON *json)
		103	{
		104	Zero (json, 1, JSON);
		105	json->max_depth = 512;
		106	}
70		107
71	/////////////////////////////////////////////////////////////////////////////	108	/////////////////////////////////////////////////////////////////////////////
72	// utility functions	109	// utility functions
73		110
74	inline void	111	INLINE SV *
		112	get_bool (const char *name)
		113	{
		114	SV *sv = get_sv (name, 1);
		115
		116	SvREADONLY_on (sv);
		117	SvREADONLY_on (SvRV (sv));
		118
		119	return sv;
		120	}
		121
		122	INLINE void
75	shrink (SV *sv)	123	shrink (SV *sv)
76	{	124	{
77	sv_utf8_downgrade (sv, 1);	125	sv_utf8_downgrade (sv, 1);
		126
78	if (SvLEN (sv) > SvCUR (sv) + 1)	127	if (SvLEN (sv) > SvCUR (sv) + 1)
79	{	128	{
80	#ifdef SvPV_shrink_to_cur	129	#ifdef SvPV_shrink_to_cur
81	SvPV_shrink_to_cur (sv);	130	SvPV_shrink_to_cur (sv);
82	#elif defined (SvPV_renew)	131	#elif defined (SvPV_renew)
…		…
88	// decode an utf-8 character and return it, or (UV)-1 in	137	// decode an utf-8 character and return it, or (UV)-1 in
89	// case of an error.	138	// case of an error.
90	// we special-case "safe" characters from U+80 .. U+7FF,	139	// we special-case "safe" characters from U+80 .. U+7FF,
91	// but use the very good perl function to parse anything else.	140	// but use the very good perl function to parse anything else.
92	// note that we never call this function for a ascii codepoints	141	// note that we never call this function for a ascii codepoints
93	inline UV	142	INLINE UV
94	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	143	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
95	{	144	{
96	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))	145	if (expect_true (len >= 2
97	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	146	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
98	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	147	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
99	{	148	{
100	*clen = 2;	149	*clen = 2;
101	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);	150	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
102	}	151	}
103	else	152	else
104	{	153	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
105	*clen = (STRLEN)-1;	154	}
106	return (UV)-1;	155
107	}	156	// likewise for encoding, also never called for ascii codepoints
		157	// this function takes advantage of this fact, although current gccs
		158	// seem to optimise the check for >= 0x80 away anyways
		159	INLINE unsigned char *
		160	encode_utf8 (unsigned char *s, UV ch)
		161	{
		162	if (expect_false (ch < 0x000080))
		163	*s++ = ch;
		164	else if (expect_true (ch < 0x000800))
		165	*s++ = 0xc0 \| ( ch >> 6),
		166	*s++ = 0x80 \| ( ch & 0x3f);
		167	else if ( ch < 0x010000)
		168	*s++ = 0xe0 \| ( ch >> 12),
		169	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		170	*s++ = 0x80 \| ( ch & 0x3f);
		171	else if ( ch < 0x110000)
		172	*s++ = 0xf0 \| ( ch >> 18),
		173	*s++ = 0x80 \| ((ch >> 12) & 0x3f),
		174	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		175	*s++ = 0x80 \| ( ch & 0x3f);
		176
		177	return s;
		178	}
		179
		180	// convert offset pointer to character index, sv must be string
		181	static STRLEN
		182	ptr_to_index (SV sv, char offset)
		183	{
		184	return SvUTF8 (sv)
		185	? utf8_distance (offset, SvPVX (sv))
		186	: offset - SvPVX (sv);
108	}	187	}
109		188
110	/////////////////////////////////////////////////////////////////////////////	189	/////////////////////////////////////////////////////////////////////////////
111	// encoder	190	// encoder
112		191
…		…
116	char *cur; // SvPVX (sv) + current output position	195	char *cur; // SvPVX (sv) + current output position
117	char *end; // SvEND (sv)	196	char *end; // SvEND (sv)
118	SV *sv; // result scalar	197	SV *sv; // result scalar
119	JSON json;	198	JSON json;
120	U32 indent; // indentation level	199	U32 indent; // indentation level
121	U32 maxdepth; // max. indentation/recursion level	200	UV limit; // escape character values >= this value when encoding
122	} enc_t;	201	} enc_t;
123		202
124	inline void	203	INLINE void
125	need (enc_t *enc, STRLEN len)	204	need (enc_t *enc, STRLEN len)
126	{	205	{
127	if (expect_false (enc->cur + len >= enc->end))	206	if (expect_false (enc->cur + len >= enc->end))
128	{	207	{
129	STRLEN cur = enc->cur - SvPVX (enc->sv);	208	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
130	SvGROW (enc->sv, cur + len + 1);	209	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
131	enc->cur = SvPVX (enc->sv) + cur;	210	enc->cur = SvPVX (enc->sv) + cur;
132	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	211	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
133	}	212	}
134	}	213	}
135		214
136	inline void	215	INLINE void
137	encode_ch (enc_t *enc, char ch)	216	encode_ch (enc_t *enc, char ch)
138	{	217	{
139	need (enc, 1);	218	need (enc, 1);
140	*enc->cur++ = ch;	219	*enc->cur++ = ch;
141	}	220	}
…		…
195	{	274	{
196	uch = ch;	275	uch = ch;
197	clen = 1;	276	clen = 1;
198	}	277	}
199		278
200	if (uch > 0x10FFFFUL)	279	if (uch < 0x80/0x20/ \|\| uch >= enc->limit)
201	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
202
203	if (uch < 0x80 \|\| enc->json.flags & F_ASCII \|\| (enc->json.flags & F_LATIN1 && uch > 0xFF))
204	{	280	{
205	if (uch > 0xFFFFUL)	281	if (uch >= 0x10000UL)
206	{	282	{
		283	if (uch >= 0x110000UL)
		284	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		285
207	need (enc, len += 11);	286	need (enc, len += 11);
208	sprintf (enc->cur, "\\u%04x\\u%04x",	287	sprintf (enc->cur, "\\u%04x\\u%04x",
209	(int)((uch - 0x10000) / 0x400 + 0xD800),	288	(int)((uch - 0x10000) / 0x400 + 0xD800),
210	(int)((uch - 0x10000) % 0x400 + 0xDC00));	289	(int)((uch - 0x10000) % 0x400 + 0xDC00));
211	enc->cur += 12;	290	enc->cur += 12;
212	}	291	}
213	else	292	else
214	{	293	{
215	static char hexdigit [16] = "0123456789abcdef";
216	need (enc, len += 5);	294	need (enc, len += 5);
217	*enc->cur++ = '\\';	295	*enc->cur++ = '\\';
218	*enc->cur++ = 'u';	296	*enc->cur++ = 'u';
219	*enc->cur++ = hexdigit [ uch >> 12 ];	297	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
220	*enc->cur++ = hexdigit [(uch >> 8) & 15];	298	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
221	*enc->cur++ = hexdigit [(uch >> 4) & 15];	299	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
222	*enc->cur++ = hexdigit [(uch >> 0) & 15];	300	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
223	}	301	}
224		302
225	str += clen;	303	str += clen;
226	}	304	}
227	else if (enc->json.flags & F_LATIN1)	305	else if (enc->json.flags & F_LATIN1)
…		…
239	while (--clen);	317	while (--clen);
240	}	318	}
241	else	319	else
242	{	320	{
243	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	321	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
244	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	322	enc->cur = encode_utf8 (enc->cur, uch);
245	++str;	323	++str;
246	}	324	}
247	}	325	}
248	}	326	}
249	}	327	}
250		328
251	--len;	329	--len;
252	}	330	}
253	}	331	}
254		332
255	inline void	333	INLINE void
256	encode_indent (enc_t *enc)	334	encode_indent (enc_t *enc)
257	{	335	{
258	if (enc->json.flags & F_INDENT)	336	if (enc->json.flags & F_INDENT)
259	{	337	{
260	int spaces = enc->indent * INDENT_STEP;	338	int spaces = enc->indent * INDENT_STEP;
…		…
263	memset (enc->cur, ' ', spaces);	341	memset (enc->cur, ' ', spaces);
264	enc->cur += spaces;	342	enc->cur += spaces;
265	}	343	}
266	}	344	}
267		345
268	inline void	346	INLINE void
269	encode_space (enc_t *enc)	347	encode_space (enc_t *enc)
270	{	348	{
271	need (enc, 1);	349	need (enc, 1);
272	encode_ch (enc, ' ');	350	encode_ch (enc, ' ');
273	}	351	}
274		352
275	inline void	353	INLINE void
276	encode_nl (enc_t *enc)	354	encode_nl (enc_t *enc)
277	{	355	{
278	if (enc->json.flags & F_INDENT)	356	if (enc->json.flags & F_INDENT)
279	{	357	{
280	need (enc, 1);	358	need (enc, 1);
281	encode_ch (enc, '\n');	359	encode_ch (enc, '\n');
282	}	360	}
283	}	361	}
284		362
285	inline void	363	INLINE void
286	encode_comma (enc_t *enc)	364	encode_comma (enc_t *enc)
287	{	365	{
288	encode_ch (enc, ',');	366	encode_ch (enc, ',');
289		367
290	if (enc->json.flags & F_INDENT)	368	if (enc->json.flags & F_INDENT)
…		…
298	static void	376	static void
299	encode_av (enc_t enc, AV av)	377	encode_av (enc_t enc, AV av)
300	{	378	{
301	int i, len = av_len (av);	379	int i, len = av_len (av);
302		380
303	if (enc->indent >= enc->maxdepth)	381	if (enc->indent >= enc->json.max_depth)
304	croak ("data structure too deep (hit recursion limit)");	382	croak (ERR_NESTING_EXCEEDED);
305		383
306	encode_ch (enc, '['); encode_nl (enc);	384	encode_ch (enc, '[');
307	++enc->indent;	385
		386	if (len >= 0)
		387	{
		388	encode_nl (enc); ++enc->indent;
308		389
309	for (i = 0; i <= len; ++i)	390	for (i = 0; i <= len; ++i)
310	{	391	{
311	SV **svp = av_fetch (av, i, 0);	392	SV **svp = av_fetch (av, i, 0);
312		393
313	encode_indent (enc);	394	encode_indent (enc);
314		395
315	if (svp)	396	if (svp)
316	encode_sv (enc, *svp);	397	encode_sv (enc, *svp);
317	else	398	else
318	encode_str (enc, "null", 4, 0);	399	encode_str (enc, "null", 4, 0);
319		400
320	if (i < len)	401	if (i < len)
321	encode_comma (enc);	402	encode_comma (enc);
322	}	403	}
323		404
		405	encode_nl (enc); --enc->indent; encode_indent (enc);
		406	}
		407
324	encode_nl (enc);	408	encode_ch (enc, ']');
325
326	--enc->indent;
327	encode_indent (enc); encode_ch (enc, ']');
328	}	409	}
329		410
330	static void	411	static void
331	encode_he (enc_t enc, HE he)	412	encode_hk (enc_t enc, HE he)
332	{	413	{
333	encode_ch (enc, '"');	414	encode_ch (enc, '"');
334		415
335	if (HeKLEN (he) == HEf_SVKEY)	416	if (HeKLEN (he) == HEf_SVKEY)
336	{	417	{
…		…
349	encode_ch (enc, '"');	430	encode_ch (enc, '"');
350		431
351	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);	432	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
352	encode_ch (enc, ':');	433	encode_ch (enc, ':');
353	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);	434	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
354	encode_sv (enc, HeVAL (he));
355	}	435	}
356		436
357	// compare hash entries, used when all keys are bytestrings	437	// compare hash entries, used when all keys are bytestrings
358	static int	438	static int
359	he_cmp_fast (const void a_, const void b_)	439	he_cmp_fast (const void a_, const void b_)
…		…
364	HE b = (HE **)b_;	444	HE b = (HE **)b_;
365		445
366	STRLEN la = HeKLEN (a);	446	STRLEN la = HeKLEN (a);
367	STRLEN lb = HeKLEN (b);	447	STRLEN lb = HeKLEN (b);
368		448
369	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	449	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
370	cmp = la - lb;	450	cmp = lb - la;
371		451
372	return cmp;	452	return cmp;
373	}	453	}
374		454
375	// compare hash entries, used when some keys are sv's or utf-x	455	// compare hash entries, used when some keys are sv's or utf-x
376	static int	456	static int
377	he_cmp_slow (const void a, const void b)	457	he_cmp_slow (const void a, const void b)
378	{	458	{
379	return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));	459	return sv_cmp (HeSVKEY_force ((HE )b), HeSVKEY_force ((HE **)a));
380	}	460	}
381		461
382	static void	462	static void
383	encode_hv (enc_t enc, HV hv)	463	encode_hv (enc_t enc, HV hv)
384	{	464	{
385	int count, i;	465	HE *he;
386		466
387	if (enc->indent >= enc->maxdepth)	467	if (enc->indent >= enc->json.max_depth)
388	croak ("data structure too deep (hit recursion limit)");	468	croak (ERR_NESTING_EXCEEDED);
389		469
390	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	470	encode_ch (enc, '{');
391		471
392	if ((count = hv_iterinit (hv)))
393	{
394	// for canonical output we have to sort by keys first	472	// for canonical output we have to sort by keys first
395	// actually, this is mostly due to the stupid so-called	473	// actually, this is mostly due to the stupid so-called
396	// security workaround added somewhere in 5.8.x.	474	// security workaround added somewhere in 5.8.x
397	// that randomises hash orderings	475	// that randomises hash orderings
398	if (enc->json.flags & F_CANONICAL)	476	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
		477	{
		478	int count = hv_iterinit (hv);
		479
		480	if (SvMAGICAL (hv))
399	{	481	{
		482	// need to count by iterating. could improve by dynamically building the vector below
		483	// but I don't care for the speed of this special case.
		484	// note also that we will run into undefined behaviour when the two iterations
		485	// do not result in the same count, something I might care for in some later release.
		486
		487	count = 0;
		488	while (hv_iternext (hv))
		489	++count;
		490
		491	hv_iterinit (hv);
		492	}
		493
		494	if (count)
		495	{
400	int fast = 1;	496	int i, fast = 1;
401	HE *he;
402	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	497	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
403	HE *hes = _alloca (count sizeof (HE));	498	HE *hes = _alloca (count sizeof (HE));
404	#else	499	#else
405	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode	500	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
406	#endif	501	#endif
…		…
433		528
434	FREETMPS;	529	FREETMPS;
435	LEAVE;	530	LEAVE;
436	}	531	}
437		532
438	for (i = 0; i < count; ++i)	533	encode_nl (enc); ++enc->indent;
		534
		535	while (count--)
439	{	536	{
440	encode_indent (enc);	537	encode_indent (enc);
		538	he = hes [count];
441	encode_he (enc, hes [i]);	539	encode_hk (enc, he);
		540	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
442		541
443	if (i < count - 1)	542	if (count)
444	encode_comma (enc);	543	encode_comma (enc);
445	}	544	}
446		545
447	encode_nl (enc);	546	encode_nl (enc); --enc->indent; encode_indent (enc);
448	}	547	}
		548	}
449	else	549	else
		550	{
		551	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))
		552	if ((he = hv_iternext (hv)))
450	{	553	{
451	HE *he = hv_iternext (hv);	554	encode_nl (enc); ++enc->indent;
452		555
453	for (;;)	556	for (;;)
454	{	557	{
455	encode_indent (enc);	558	encode_indent (enc);
456	encode_he (enc, he);	559	encode_hk (enc, he);
		560	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
457		561
458	if (!(he = hv_iternext (hv)))	562	if (!(he = hv_iternext (hv)))
459	break;	563	break;
460		564
461	encode_comma (enc);	565	encode_comma (enc);
462	}	566	}
463		567
464	encode_nl (enc);	568	encode_nl (enc); --enc->indent; encode_indent (enc);
465	}	569	}
466	}	570	}
467		571
468	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	572	encode_ch (enc, '}');
469	}	573	}
470		574
471	// encode objects, arrays and special \0=false and \1=true values.	575	// encode objects, arrays and special \0=false and \1=true values.
472	static void	576	static void
473	encode_rv (enc_t enc, SV sv)	577	encode_rv (enc_t enc, SV sv)
…		…
477	SvGETMAGIC (sv);	581	SvGETMAGIC (sv);
478	svt = SvTYPE (sv);	582	svt = SvTYPE (sv);
479		583
480	if (expect_false (SvOBJECT (sv)))	584	if (expect_false (SvOBJECT (sv)))
481	{	585	{
		586	HV *stash = !JSON_SLOW \|\| json_boolean_stash
		587	? json_boolean_stash
		588	: gv_stashpv ("JSON::XS::Boolean", 1);
		589
482	if (SvSTASH (sv) == json_boolean_stash)	590	if (SvSTASH (sv) == stash)
483	{	591	{
484	if (SvIV (sv))	592	if (SvIV (sv))
485	encode_str (enc, "true", 4, 0);	593	encode_str (enc, "true", 4, 0);
486	else	594	else
487	encode_str (enc, "false", 5, 0);	595	encode_str (enc, "false", 5, 0);
…		…
499	// we re-bless the reference to get overload and other niceties right	607	// we re-bless the reference to get overload and other niceties right
500	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);	608	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
501		609
502	if (to_json)	610	if (to_json)
503	{	611	{
504	int count;
505	dSP;	612	dSP;
506		613
507	ENTER; SAVETMPS; PUSHMARK (SP);	614	ENTER; SAVETMPS; PUSHMARK (SP);
508	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));	615	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
509		616
…		…
547		654
548	if (len == 1 && *pv == '1')	655	if (len == 1 && *pv == '1')
549	encode_str (enc, "true", 4, 0);	656	encode_str (enc, "true", 4, 0);
550	else if (len == 1 && *pv == '0')	657	else if (len == 1 && *pv == '0')
551	encode_str (enc, "false", 5, 0);	658	encode_str (enc, "false", 5, 0);
		659	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		660	encode_str (enc, "null", 4, 0);
552	else	661	else
553	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	662	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
554	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	663	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
555	}	664	}
		665	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		666	encode_str (enc, "null", 4, 0);
556	else	667	else
557	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	668	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
558	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	669	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
559	}	670	}
560		671
…		…
578	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	689	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
579	enc->cur += strlen (enc->cur);	690	enc->cur += strlen (enc->cur);
580	}	691	}
581	else if (SvIOKp (sv))	692	else if (SvIOKp (sv))
582	{	693	{
583	// we assume we can always read an IV as a UV	694	// we assume we can always read an IV as a UV and vice versa
584	if (SvUV (sv) & ~(UV)0x7fff)	695	// we assume two's complement
585	{	696	// we assume no aliasing issues in the union
586	// large integer, use the (rather slow) snprintf way.	697	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
587	need (enc, sizeof (UV) * 3);	698	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
588	enc->cur +=
589	SvIsUV(sv)
590	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
591	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
592	}
593	else
594	{	699	{
595	// optimise the "small number case"	700	// optimise the "small number case"
596	// code will likely be branchless and use only a single multiplication	701	// code will likely be branchless and use only a single multiplication
		702	// works for numbers up to 59074
597	I32 i = SvIV (sv);	703	I32 i = SvIVX (sv);
598	U32 u;	704	U32 u;
599	char digit, nz = 0;	705	char digit, nz = 0;
600		706
601	need (enc, 6);	707	need (enc, 6);
602		708
…		…
608		714
609	// now output digit by digit, each time masking out the integer part	715	// now output digit by digit, each time masking out the integer part
610	// and multiplying by 5 while moving the decimal point one to the right,	716	// and multiplying by 5 while moving the decimal point one to the right,
611	// resulting in a net multiplication by 10.	717	// resulting in a net multiplication by 10.
612	// we always write the digit to memory but conditionally increment	718	// we always write the digit to memory but conditionally increment
613	// the pointer, to ease the usage of conditional move instructions.	719	// the pointer, to enable the use of conditional move instructions.
614	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffff) 5;	720	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffffUL) 5;
615	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffff) 5;	721	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffffUL) 5;
616	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffff) 5;	722	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffffUL) 5;
617	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffff) 5;	723	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffffUL) 5;
618	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'	724	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
619	}	725	}
		726	else
		727	{
		728	// large integer, use the (rather slow) snprintf way.
		729	need (enc, IVUV_MAXCHARS);
		730	enc->cur +=
		731	SvIsUV(sv)
		732	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
		733	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		734	}
620	}	735	}
621	else if (SvROK (sv))	736	else if (SvROK (sv))
622	encode_rv (enc, SvRV (sv));	737	encode_rv (enc, SvRV (sv));
623	else if (!SvOK (sv))	738	else if (!SvOK (sv) \|\| enc->json.flags & F_ALLOW_UNKNOWN)
624	encode_str (enc, "null", 4, 0);	739	encode_str (enc, "null", 4, 0);
625	else	740	else
626	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	741	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
627	SvPV_nolen (sv), SvFLAGS (sv));	742	SvPV_nolen (sv), SvFLAGS (sv));
628	}	743	}
…		…
638	enc.json = *json;	753	enc.json = *json;
639	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	754	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
640	enc.cur = SvPVX (enc.sv);	755	enc.cur = SvPVX (enc.sv);
641	enc.end = SvEND (enc.sv);	756	enc.end = SvEND (enc.sv);
642	enc.indent = 0;	757	enc.indent = 0;
643	enc.maxdepth = DEC_DEPTH (enc.json.flags);	758	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		759	: enc.json.flags & F_LATIN1 ? 0x000100UL
		760	: 0x110000UL;
644		761
645	SvPOK_only (enc.sv);	762	SvPOK_only (enc.sv);
646	encode_sv (&enc, scalar);	763	encode_sv (&enc, scalar);
647		764
648	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	765	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
…		…
669	JSON json;	786	JSON json;
670	U32 depth; // recursion depth	787	U32 depth; // recursion depth
671	U32 maxdepth; // recursion depth limit	788	U32 maxdepth; // recursion depth limit
672	} dec_t;	789	} dec_t;
673		790
674	inline void	791	INLINE void
		792	decode_comment (dec_t *dec)
		793	{
		794	// only '#'-style comments allowed a.t.m.
		795
		796	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)
		797	++dec->cur;
		798	}
		799
		800	INLINE void
675	decode_ws (dec_t *dec)	801	decode_ws (dec_t *dec)
676	{	802	{
677	for (;;)	803	for (;;)
678	{	804	{
679	char ch = *dec->cur;	805	char ch = *dec->cur;
680		806
681	if (ch > 0x20	807	if (ch > 0x20)
		808	{
		809	if (expect_false (ch == '#'))
		810	{
		811	if (dec->json.flags & F_RELAXED)
		812	decode_comment (dec);
		813	else
		814	break;
		815	}
		816	else
		817	break;
		818	}
682	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	819	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
683	break;	820	break; // parse error, but let higher level handle it, gives better error messages
684		821
685	++dec->cur;	822	++dec->cur;
686	}	823	}
687	}	824	}
688		825
…		…
692	if (*dec->cur != ch) \	829	if (*dec->cur != ch) \
693	ERR (# ch " expected"); \	830	ERR (# ch " expected"); \
694	++dec->cur; \	831	++dec->cur; \
695	SE	832	SE
696		833
697	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	834	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
698	#define DEC_DEC_DEPTH --dec->depth	835	#define DEC_DEC_DEPTH --dec->depth
699		836
700	static SV decode_sv (dec_t dec);	837	static SV decode_sv (dec_t dec);
701		838
702	static signed char decode_hexdigit[256];	839	static signed char decode_hexdigit[256];
…		…
794		931
795	if (hi >= 0x80)	932	if (hi >= 0x80)
796	{	933	{
797	utf8 = 1;	934	utf8 = 1;
798		935
799	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	936	cur = encode_utf8 (cur, hi);
800	}	937	}
801	else	938	else
802	*cur++ = hi;	939	*cur++ = hi;
803	}	940	}
804	break;	941	break;
…		…
806	default:	943	default:
807	--dec_cur;	944	--dec_cur;
808	ERR ("illegal backslash escape sequence in string");	945	ERR ("illegal backslash escape sequence in string");
809	}	946	}
810	}	947	}
811	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	948	else if (expect_true (ch >= 0x20 && ch < 0x80))
812	*cur++ = ch;	949	*cur++ = ch;
813	else if (ch >= 0x80)	950	else if (ch >= 0x80)
814	{	951	{
815	STRLEN clen;	952	STRLEN clen;
816	UV uch;	953	UV uch;
…		…
842	{	979	{
843	STRLEN len = cur - buf;	980	STRLEN len = cur - buf;
844		981
845	if (sv)	982	if (sv)
846	{	983	{
847	SvGROW (sv, SvCUR (sv) + len + 1);	984	STRLEN cur = SvCUR (sv);
		985
		986	if (SvLEN (sv) <= cur + len)
		987	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		988
848	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	989	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
849	SvCUR_set (sv, SvCUR (sv) + len);	990	SvCUR_set (sv, SvCUR (sv) + len);
850	}	991	}
851	else	992	else
852	sv = newSVpvn (buf, len);	993	sv = newSVpvn (buf, len);
…		…
937	is_nv = 1;	1078	is_nv = 1;
938	}	1079	}
939		1080
940	if (!is_nv)	1081	if (!is_nv)
941	{	1082	{
		1083	int len = dec->cur - start;
		1084
942	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1085	// special case the rather common 1..5-digit-int case
943	if (*start == '-')	1086	if (*start == '-')
944	switch (dec->cur - start)	1087	switch (len)
945	{	1088	{
946	case 2: return newSViv (-( start [1] - '0' * 1));	1089	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
947	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1090	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
948	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));	1091	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
949	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));	1092	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1093	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
950	}	1094	}
951	else	1095	else
952	switch (dec->cur - start)	1096	switch (len)
953	{	1097	{
954	case 1: return newSViv ( start [0] - '0' * 1);	1098	case 1: return newSViv ( start [0] - '0' * 1);
955	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1099	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
956	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);	1100	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
957	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);	1101	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1102	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
958	}	1103	}
959		1104
960	{	1105	{
961	UV uv;	1106	UV uv;
962	int numtype = grok_number (start, dec->cur - start, &uv);	1107	int numtype = grok_number (start, len, &uv);
963	if (numtype & IS_NUMBER_IN_UV)	1108	if (numtype & IS_NUMBER_IN_UV)
964	if (numtype & IS_NUMBER_NEG)	1109	if (numtype & IS_NUMBER_NEG)
965	{	1110	{
966	if (uv < (UV)IV_MIN)	1111	if (uv < (UV)IV_MIN)
967	return newSViv (-(IV)uv);	1112	return newSViv (-(IV)uv);
968	}	1113	}
969	else	1114	else
970	return newSVuv (uv);	1115	return newSVuv (uv);
971
972	// here would likely be the place for bigint support
973	}	1116	}
974	}
975		1117
976	// if we ever support bigint or bigfloat, this is the place for bigfloat	1118	len -= *start == '-' ? 1 : 0;
		1119
		1120	// does not fit into IV or UV, try NV
		1121	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1122	#if defined (LDBL_DIG)
		1123	\|\| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1124	#endif
		1125	)
		1126	// fits into NV without loss of precision
		1127	return newSVnv (Atof (start));
		1128
		1129	// everything else fails, convert it to a string
		1130	return newSVpvn (start, dec->cur - start);
		1131	}
		1132
		1133	// loss of precision here
977	return newSVnv (Atof (start));	1134	return newSVnv (Atof (start));
978		1135
979	fail:	1136	fail:
980	return 0;	1137	return 0;
981	}	1138	}
…		…
1011		1168
1012	if (*dec->cur != ',')	1169	if (*dec->cur != ',')
1013	ERR (", or ] expected while parsing array");	1170	ERR (", or ] expected while parsing array");
1014		1171
1015	++dec->cur;	1172	++dec->cur;
		1173
		1174	decode_ws (dec);
		1175
		1176	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1177	{
		1178	++dec->cur;
		1179	break;
		1180	}
1016	}	1181	}
1017		1182
1018	DEC_DEC_DEPTH;	1183	DEC_DEC_DEPTH;
1019	return newRV_noinc ((SV *)av);	1184	return newRV_noinc ((SV *)av);
1020		1185
…		…
1036	if (*dec->cur == '}')	1201	if (*dec->cur == '}')
1037	++dec->cur;	1202	++dec->cur;
1038	else	1203	else
1039	for (;;)	1204	for (;;)
1040	{	1205	{
1041	decode_ws (dec); EXPECT_CH ('"');	1206	EXPECT_CH ('"');
1042		1207
1043	// heuristic: assume that	1208	// heuristic: assume that
1044	// a) decode_str + hv_store_ent are abysmally slow.	1209	// a) decode_str + hv_store_ent are abysmally slow.
1045	// b) most hash keys are short, simple ascii text.	1210	// b) most hash keys are short, simple ascii text.
1046	// => try to "fast-match" such strings to avoid	1211	// => try to "fast-match" such strings to avoid
…		…
1050	char *p = dec->cur;	1215	char *p = dec->cur;
1051	char *e = p + 24; // only try up to 24 bytes	1216	char *e = p + 24; // only try up to 24 bytes
1052		1217
1053	for (;;)	1218	for (;;)
1054	{	1219	{
1055	// the >= 0x80 is true on most architectures	1220	// the >= 0x80 is false on most architectures
1056	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')	1221	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
1057	{	1222	{
1058	// slow path, back up and use decode_str	1223	// slow path, back up and use decode_str
1059	SV *key = decode_str (dec);	1224	SV *key = decode_str (dec);
1060	if (!key)	1225	if (!key)
1061	goto fail;	1226	goto fail;
1062		1227
1063	decode_ws (dec); EXPECT_CH (':');	1228	decode_ws (dec); EXPECT_CH (':');
1064		1229
		1230	decode_ws (dec);
1065	value = decode_sv (dec);	1231	value = decode_sv (dec);
1066	if (!value)	1232	if (!value)
1067	{	1233	{
1068	SvREFCNT_dec (key);	1234	SvREFCNT_dec (key);
1069	goto fail;	1235	goto fail;
…		…
1081	int len = p - key;	1247	int len = p - key;
1082	dec->cur = p + 1;	1248	dec->cur = p + 1;
1083		1249
1084	decode_ws (dec); EXPECT_CH (':');	1250	decode_ws (dec); EXPECT_CH (':');
1085		1251
		1252	decode_ws (dec);
1086	value = decode_sv (dec);	1253	value = decode_sv (dec);
1087	if (!value)	1254	if (!value)
1088	goto fail;	1255	goto fail;
1089		1256
1090	hv_store (hv, key, len, value, 0);	1257	hv_store (hv, key, len, value, 0);
…		…
1106		1273
1107	if (*dec->cur != ',')	1274	if (*dec->cur != ',')
1108	ERR (", or } expected while parsing object/hash");	1275	ERR (", or } expected while parsing object/hash");
1109		1276
1110	++dec->cur;	1277	++dec->cur;
		1278
		1279	decode_ws (dec);
		1280
		1281	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1282	{
		1283	++dec->cur;
		1284	break;
		1285	}
1111	}	1286	}
1112		1287
1113	DEC_DEC_DEPTH;	1288	DEC_DEC_DEPTH;
1114	sv = newRV_noinc ((SV *)hv);	1289	sv = newRV_noinc ((SV *)hv);
1115		1290
…		…
1180	}	1355	}
1181		1356
1182	static SV *	1357	static SV *
1183	decode_sv (dec_t *dec)	1358	decode_sv (dec_t *dec)
1184	{	1359	{
1185	decode_ws (dec);
1186
1187	// the beauty of JSON: you need exactly one character lookahead	1360	// the beauty of JSON: you need exactly one character lookahead
1188	// to parse anything.	1361	// to parse everything.
1189	switch (*dec->cur)	1362	switch (*dec->cur)
1190	{	1363	{
1191	case '"': ++dec->cur; return decode_str (dec);	1364	case '"': ++dec->cur; return decode_str (dec);
1192	case '[': ++dec->cur; return decode_av (dec);	1365	case '[': ++dec->cur; return decode_av (dec);
1193	case '{': ++dec->cur; return decode_hv (dec);	1366	case '{': ++dec->cur; return decode_hv (dec);
1194		1367
1195	case '-':	1368	case '-':
1196	case '0': case '1': case '2': case '3': case '4':	1369	case '0': case '1': case '2': case '3': case '4':
1197	case '5': case '6': case '7': case '8': case '9':	1370	case '5': case '6': case '7': case '8': case '9':
1198	return decode_num (dec);	1371	return decode_num (dec);
1199		1372
1200	case 't':	1373	case 't':
1201	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1374	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1202	{	1375	{
1203	dec->cur += 4;	1376	dec->cur += 4;
		1377	#if JSON_SLOW
		1378	json_true = get_bool ("JSON::XS::true");
		1379	#endif
1204	return SvREFCNT_inc (json_true);	1380	return newSVsv (json_true);
1205	}	1381	}
1206	else	1382	else
1207	ERR ("'true' expected");	1383	ERR ("'true' expected");
1208		1384
1209	break;	1385	break;
1210		1386
1211	case 'f':	1387	case 'f':
1212	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1388	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1213	{	1389	{
1214	dec->cur += 5;	1390	dec->cur += 5;
		1391	#if JSON_SLOW
		1392	json_false = get_bool ("JSON::XS::false");
		1393	#endif
1215	return SvREFCNT_inc (json_false);	1394	return newSVsv (json_false);
1216	}	1395	}
1217	else	1396	else
1218	ERR ("'false' expected");	1397	ERR ("'false' expected");
1219		1398
1220	break;	1399	break;
…		…
1238	fail:	1417	fail:
1239	return 0;	1418	return 0;
1240	}	1419	}
1241		1420
1242	static SV *	1421	static SV *
1243	decode_json (SV string, JSON json, UV *offset_return)	1422	decode_json (SV string, JSON json, char **offset_return)
1244	{	1423	{
1245	dec_t dec;	1424	dec_t dec;
1246	UV offset;
1247	SV *sv;	1425	SV *sv;
1248		1426
		1427	/* work around bugs in 5.10 where manipulating magic values
		1428	* will perl ignore the magic in subsequent accesses
		1429	*/
1249	SvGETMAGIC (string);	1430	/SvGETMAGIC (string);/
		1431	if (SvMAGICAL (string))
		1432	string = sv_2mortal (newSVsv (string));
		1433
1250	SvUPGRADE (string, SVt_PV);	1434	SvUPGRADE (string, SVt_PV);
1251		1435
1252	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))	1436	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1437	* assertion with -DDEBUGGING, although SvCUR is documented to
		1438	* return the xpv_cur field which certainly exists after upgrading.
		1439	* according to nicholas clark, calling SvPOK fixes this.
		1440	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1441	* and hope for the best.
		1442	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1443	* assertion business is seriously broken, try yet another workaround
		1444	* for the broken -DDEBUGGING.
		1445	*/
		1446	{
		1447	#ifdef DEBUGGING
		1448	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1449	#else
		1450	STRLEN offset = SvCUR (string);
		1451	#endif
		1452
		1453	if (offset > json->max_size && json->max_size)
1253	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",	1454	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1254	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));	1455	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1456	}
1255		1457
1256	if (json->flags & F_UTF8)	1458	if (json->flags & F_UTF8)
1257	sv_utf8_downgrade (string, 0);	1459	sv_utf8_downgrade (string, 0);
1258	else	1460	else
1259	sv_utf8_upgrade (string);	1461	sv_utf8_upgrade (string);
1260		1462
1261	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1463	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1262		1464
1263	dec.json = *json;	1465	dec.json = *json;
1264	dec.cur = SvPVX (string);	1466	dec.cur = SvPVX (string);
1265	dec.end = SvEND (string);	1467	dec.end = SvEND (string);
1266	dec.err = 0;	1468	dec.err = 0;
1267	dec.depth = 0;	1469	dec.depth = 0;
1268	dec.maxdepth = DEC_DEPTH (dec.json.flags);
1269		1470
1270	if (dec.json.cb_object \|\| dec.json.cb_sk_object)	1471	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
1271	dec.json.flags \|= F_HOOK;	1472	dec.json.flags \|= F_HOOK;
1272		1473
1273	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1474	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1475
		1476	decode_ws (&dec);
1274	sv = decode_sv (&dec);	1477	sv = decode_sv (&dec);
		1478
		1479	if (offset_return)
		1480	*offset_return = dec.cur;
1275		1481
1276	if (!(offset_return \|\| !sv))	1482	if (!(offset_return \|\| !sv))
1277	{	1483	{
1278	// check for trailing garbage	1484	// check for trailing garbage
1279	decode_ws (&dec);	1485	decode_ws (&dec);
…		…
1282	{	1488	{
1283	dec.err = "garbage after JSON object";	1489	dec.err = "garbage after JSON object";
1284	SvREFCNT_dec (sv);	1490	SvREFCNT_dec (sv);
1285	sv = 0;	1491	sv = 0;
1286	}	1492	}
1287	}
1288
1289	if (offset_return \|\| !sv)
1290	{
1291	offset = dec.json.flags & F_UTF8
1292	? dec.cur - SvPVX (string)
1293	: utf8_distance (dec.cur, SvPVX (string));
1294
1295	if (offset_return)
1296	*offset_return = offset;
1297	}	1493	}
1298		1494
1299	if (!sv)	1495	if (!sv)
1300	{	1496	{
1301	SV *uni = sv_newmortal ();	1497	SV *uni = sv_newmortal ();
…		…
1307	SAVEVPTR (PL_curcop);	1503	SAVEVPTR (PL_curcop);
1308	PL_curcop = &cop;	1504	PL_curcop = &cop;
1309	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1505	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1310	LEAVE;	1506	LEAVE;
1311		1507
1312	croak ("%s, at character offset %d [\"%s\"]",	1508	croak ("%s, at character offset %d (before \"%s\")",
1313	dec.err,	1509	dec.err,
1314	(int)offset,	1510	ptr_to_index (string, dec.cur),
1315	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1511	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1316	}	1512	}
1317		1513
1318	sv = sv_2mortal (sv);	1514	sv = sv_2mortal (sv);
1319		1515
1320	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))	1516	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1321	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1517	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1322		1518
1323	return sv;	1519	return sv;
		1520	}
		1521
		1522	/////////////////////////////////////////////////////////////////////////////
		1523	// incremental parser
		1524
		1525	static void
		1526	incr_parse (JSON *self)
		1527	{
		1528	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1529
		1530	for (;;)
		1531	{
		1532	//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
		1533	switch (self->incr_mode)
		1534	{
		1535	// only used for intiial whitespace skipping
		1536	case INCR_M_WS:
		1537	for (;;)
		1538	{
		1539	if (*p > 0x20)
		1540	{
		1541	self->incr_mode = INCR_M_JSON;
		1542	goto incr_m_json;
		1543	}
		1544	else if (!*p)
		1545	goto interrupt;
		1546
		1547	++p;
		1548	}
		1549
		1550	// skip a single char inside a string (for \\-processing)
		1551	case INCR_M_BS:
		1552	if (!*p)
		1553	goto interrupt;
		1554
		1555	++p;
		1556	self->incr_mode = INCR_M_STR;
		1557	goto incr_m_str;
		1558
		1559	// inside a string
		1560	case INCR_M_STR:
		1561	incr_m_str:
		1562	for (;;)
		1563	{
		1564	if (*p == '"')
		1565	{
		1566	++p;
		1567	self->incr_mode = INCR_M_JSON;
		1568
		1569	if (!self->incr_nest)
		1570	goto interrupt;
		1571
		1572	goto incr_m_json;
		1573	}
		1574	else if (*p == '\\')
		1575	{
		1576	++p; // "virtually" consumes character after \
		1577
		1578	if (!*p) // if at end of string we have to switch modes
		1579	{
		1580	self->incr_mode = INCR_M_BS;
		1581	goto interrupt;
		1582	}
		1583	}
		1584	else if (!*p)
		1585	goto interrupt;
		1586
		1587	++p;
		1588	}
		1589
		1590	// after initial ws, outside string
		1591	case INCR_M_JSON:
		1592	incr_m_json:
		1593	for (;;)
		1594	{
		1595	switch (*p++)
		1596	{
		1597	case 0:
		1598	--p;
		1599	goto interrupt;
		1600
		1601	case 0x09:
		1602	case 0x0a:
		1603	case 0x0d:
		1604	case 0x20:
		1605	if (!self->incr_nest)
		1606	{
		1607	--p; // do not eat the whitespace, let the next round do it
		1608	goto interrupt;
		1609	}
		1610	break;
		1611
		1612	case '"':
		1613	self->incr_mode = INCR_M_STR;
		1614	goto incr_m_str;
		1615
		1616	case '[':
		1617	case '{':
		1618	if (++self->incr_nest > self->max_depth)
		1619	croak (ERR_NESTING_EXCEEDED);
		1620	break;
		1621
		1622	case ']':
		1623	case '}':
		1624	if (--self->incr_nest <= 0)
		1625	goto interrupt;
		1626	}
		1627	}
		1628	}
		1629
		1630	modechange:
		1631	;
		1632	}
		1633
		1634	interrupt:
		1635	self->incr_pos = p - SvPVX (self->incr_text);
		1636	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1324	}	1637	}
1325		1638
1326	/////////////////////////////////////////////////////////////////////////////	1639	/////////////////////////////////////////////////////////////////////////////
1327	// XS interface functions	1640	// XS interface functions
1328		1641
…		…
1340	: -1;	1653	: -1;
1341		1654
1342	json_stash = gv_stashpv ("JSON::XS" , 1);	1655	json_stash = gv_stashpv ("JSON::XS" , 1);
1343	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);	1656	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1344		1657
1345	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );	1658	json_true = get_bool ("JSON::XS::true");
1346	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1659	json_false = get_bool ("JSON::XS::false");
		1660
		1661	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1347	}	1662	}
1348		1663
1349	PROTOTYPES: DISABLE	1664	PROTOTYPES: DISABLE
		1665
		1666	void CLONE (...)
		1667	CODE:
		1668	json_stash = 0;
		1669	json_boolean_stash = 0;
1350		1670
1351	void new (char *klass)	1671	void new (char *klass)
1352	PPCODE:	1672	PPCODE:
1353	{	1673	{
1354	SV *pv = NEWSV (0, sizeof (JSON));	1674	SV *pv = NEWSV (0, sizeof (JSON));
1355	SvPOK_only (pv);	1675	SvPOK_only (pv);
1356	Zero (SvPVX (pv), 1, JSON);	1676	json_init ((JSON *)SvPVX (pv));
1357	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	1677	XPUSHs (sv_2mortal (sv_bless (
1358	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash)));	1678	newRV_noinc (pv),
		1679	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1680	)));
1359	}	1681	}
1360		1682
1361	void ascii (JSON *self, int enable = 1)	1683	void ascii (JSON *self, int enable = 1)
1362	ALIAS:	1684	ALIAS:
1363	ascii = F_ASCII	1685	ascii = F_ASCII
…		…
1370	pretty = F_PRETTY	1692	pretty = F_PRETTY
1371	allow_nonref = F_ALLOW_NONREF	1693	allow_nonref = F_ALLOW_NONREF
1372	shrink = F_SHRINK	1694	shrink = F_SHRINK
1373	allow_blessed = F_ALLOW_BLESSED	1695	allow_blessed = F_ALLOW_BLESSED
1374	convert_blessed = F_CONV_BLESSED	1696	convert_blessed = F_CONV_BLESSED
		1697	relaxed = F_RELAXED
		1698	allow_unknown = F_ALLOW_UNKNOWN
1375	PPCODE:	1699	PPCODE:
1376	{	1700	{
1377	if (enable)	1701	if (enable)
1378	self->flags \|= ix;	1702	self->flags \|= ix;
1379	else	1703	else
1380	self->flags &= ~ix;	1704	self->flags &= ~ix;
1381		1705
1382	XPUSHs (ST (0));	1706	XPUSHs (ST (0));
1383	}	1707	}
1384		1708
1385	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1709	void get_ascii (JSON *self)
		1710	ALIAS:
		1711	get_ascii = F_ASCII
		1712	get_latin1 = F_LATIN1
		1713	get_utf8 = F_UTF8
		1714	get_indent = F_INDENT
		1715	get_canonical = F_CANONICAL
		1716	get_space_before = F_SPACE_BEFORE
		1717	get_space_after = F_SPACE_AFTER
		1718	get_allow_nonref = F_ALLOW_NONREF
		1719	get_shrink = F_SHRINK
		1720	get_allow_blessed = F_ALLOW_BLESSED
		1721	get_convert_blessed = F_CONV_BLESSED
		1722	get_relaxed = F_RELAXED
		1723	get_allow_unknown = F_ALLOW_UNKNOWN
1386	PPCODE:	1724	PPCODE:
1387	{	1725	XPUSHs (boolSV (self->flags & ix));
1388	UV log2 = 0;
1389		1726
1390	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	1727	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1391		1728	PPCODE:
1392	while ((1UL << log2) < max_depth)	1729	self->max_depth = max_depth;
1393	++log2;
1394
1395	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1396
1397	XPUSHs (ST (0));	1730	XPUSHs (ST (0));
1398	}
1399		1731
		1732	U32 get_max_depth (JSON *self)
		1733	CODE:
		1734	RETVAL = self->max_depth;
		1735	OUTPUT:
		1736	RETVAL
		1737
1400	void max_size (JSON *self, UV max_size = 0)	1738	void max_size (JSON *self, U32 max_size = 0)
1401	PPCODE:	1739	PPCODE:
1402	{	1740	self->max_size = max_size;
1403	UV log2 = 0;
1404
1405	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1406	if (max_size == 1) max_size = 2;
1407
1408	while ((1UL << log2) < max_size)
1409	++log2;
1410
1411	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
1412
1413	XPUSHs (ST (0));	1741	XPUSHs (ST (0));
1414	}	1742
		1743	int get_max_size (JSON *self)
		1744	CODE:
		1745	RETVAL = self->max_size;
		1746	OUTPUT:
		1747	RETVAL
1415		1748
1416	void filter_json_object (JSON self, SV cb = &PL_sv_undef)	1749	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1417	PPCODE:	1750	PPCODE:
1418	{	1751	{
1419	SvREFCNT_dec (self->cb_object);	1752	SvREFCNT_dec (self->cb_object);
…		…
1453	XPUSHs (decode_json (jsonstr, self, 0));	1786	XPUSHs (decode_json (jsonstr, self, 0));
1454		1787
1455	void decode_prefix (JSON self, SV jsonstr)	1788	void decode_prefix (JSON self, SV jsonstr)
1456	PPCODE:	1789	PPCODE:
1457	{	1790	{
1458	UV offset;	1791	char *offset;
1459	EXTEND (SP, 2);	1792	EXTEND (SP, 2);
1460	PUSHs (decode_json (jsonstr, self, &offset));	1793	PUSHs (decode_json (jsonstr, self, &offset));
1461	PUSHs (sv_2mortal (newSVuv (offset)));	1794	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		1795	}
		1796
		1797	void incr_parse (JSON self, SV jsonstr = 0)
		1798	PPCODE:
		1799	{
		1800	if (!self->incr_text)
		1801	self->incr_text = newSVpvn ("", 0);
		1802
		1803	// append data, if any
		1804	if (jsonstr)
		1805	{
		1806	if (SvUTF8 (jsonstr))
		1807	{
		1808	if (!SvUTF8 (self->incr_text))
		1809	{
		1810	/* utf-8-ness differs, need to upgrade */
		1811	sv_utf8_upgrade (self->incr_text);
		1812
		1813	if (self->incr_pos)
		1814	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1815	- (U8 *)SvPVX (self->incr_text);
		1816	}
		1817	}
		1818	else if (SvUTF8 (self->incr_text))
		1819	sv_utf8_upgrade (jsonstr);
		1820
		1821	{
		1822	STRLEN len;
		1823	const char *str = SvPV (jsonstr, len);
		1824	STRLEN cur = SvCUR (self->incr_text);
		1825
		1826	if (SvLEN (self->incr_text) <= cur + len)
		1827	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1828
		1829	Move (str, SvEND (self->incr_text), len, char);
		1830	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1831	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1832	}
		1833	}
		1834
		1835	if (GIMME_V != G_VOID)
		1836	do
		1837	{
		1838	char *offset;
		1839
		1840	if (!INCR_DONE (self))
		1841	{
		1842	incr_parse (self);
		1843
		1844	if (self->incr_pos > self->max_size && self->max_size)
		1845	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1846	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1847
		1848	if (!INCR_DONE (self))
		1849	break;
		1850	}
		1851
		1852	XPUSHs (decode_json (self->incr_text, self, &offset));
		1853
		1854	self->incr_pos -= offset - SvPVX (self->incr_text);
		1855	self->incr_nest = 0;
		1856	self->incr_mode = 0;
		1857
		1858	sv_chop (self->incr_text, offset);
		1859	}
		1860	while (GIMME_V == G_ARRAY);
		1861	}
		1862
		1863	SV incr_text (JSON self)
		1864	ATTRS: lvalue
		1865	CODE:
		1866	{
		1867	if (self->incr_pos)
		1868	croak ("incr_text can not be called when the incremental parser already started parsing");
		1869
		1870	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		1871	}
		1872	OUTPUT:
		1873	RETVAL
		1874
		1875	void incr_skip (JSON *self)
		1876	CODE:
		1877	{
		1878	if (self->incr_pos)
		1879	{
		1880	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1881	self->incr_pos = 0;
		1882	self->incr_nest = 0;
		1883	self->incr_mode = 0;
		1884	}
		1885	}
		1886
		1887	void incr_reset (JSON *self)
		1888	CODE:
		1889	{
		1890	SvREFCNT_dec (self->incr_text);
		1891	self->incr_text = 0;
		1892	self->incr_pos = 0;
		1893	self->incr_nest = 0;
		1894	self->incr_mode = 0;
1462	}	1895	}
1463		1896
1464	void DESTROY (JSON *self)	1897	void DESTROY (JSON *self)
1465	CODE:	1898	CODE:
1466	SvREFCNT_dec (self->cb_sk_object);	1899	SvREFCNT_dec (self->cb_sk_object);
1467	SvREFCNT_dec (self->cb_object);	1900	SvREFCNT_dec (self->cb_object);
		1901	SvREFCNT_dec (self->incr_text);
1468		1902
1469	PROTOTYPES: ENABLE	1903	PROTOTYPES: ENABLE
1470		1904
1471	void to_json (SV *scalar)	1905	void encode_json (SV *scalar)
		1906	ALIAS:
		1907	to_json_ = 0
		1908	encode_json = F_UTF8
1472	PPCODE:	1909	PPCODE:
1473	{	1910	{
1474	JSON json = { F_DEFAULT \| F_UTF8 };	1911	JSON json;
		1912	json_init (&json);
		1913	json.flags \|= ix;
1475	XPUSHs (encode_json (scalar, &json));	1914	XPUSHs (encode_json (scalar, &json));
1476	}	1915	}
1477		1916
1478	void from_json (SV *jsonstr)	1917	void decode_json (SV *jsonstr)
		1918	ALIAS:
		1919	from_json_ = 0
		1920	decode_json = F_UTF8
1479	PPCODE:	1921	PPCODE:
1480	{	1922	{
1481	JSON json = { F_DEFAULT \| F_UTF8 };	1923	JSON json;
		1924	json_init (&json);
		1925	json.flags \|= ix;
1482	XPUSHs (decode_json (jsonstr, &json, 0));	1926	XPUSHs (decode_json (jsonstr, &json, 0));
1483	}	1927	}
1484		1928
		1929

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.55 by root, Mon Jul 23 22:57:40 2007 UTC vs. Revision 1.99 by root, Sat Aug 8 10:06:02 2009 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.55 by root, Mon Jul 23 22:57:40 2007 UTC vs.
Revision 1.99 by root, Sat Aug 8 10:06:02 2009 UTC