ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.11 by root, Sat Mar 24 19:42:14 2007 UTC vs.
Revision 1.66 by root, Sun Nov 25 19:02:42 2007 UTC

1	#include "EXTERN.h"	1	#include "EXTERN.h"
2	#include "perl.h"	2	#include "perl.h"
3	#include "XSUB.h"	3	#include "XSUB.h"
4		4
5	#include "assert.h"	5	#include <assert.h>
6	#include "string.h"	6	#include <string.h>
7	#include "stdlib.h"	7	#include <stdlib.h>
		8	#include <stdio.h>
		9	#include <float.h>
8		10
		11	#if defined(__BORLANDC__) || defined(_MSC_VER)
		12	# define snprintf _snprintf // C compilers have this in stdio.h
		13	#endif
		14
		15	// 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	#ifndef UTF8_MAXBYTES
		18	# define UTF8_MAXBYTES 13
		19	#endif
		20
9	#define F_ASCII 0x00000001	21	#define F_ASCII 0x00000001UL
		22	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	23	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	24	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	25	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	26	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	27	#define F_SPACE_AFTER 0x00000040UL
15	#define F_ALLOW_NONREF 0x00000080	28	#define F_ALLOW_NONREF 0x00000100UL
16	#define F_SHRINK 0x00000100	29	#define F_SHRINK 0x00000200UL
		30	#define F_ALLOW_BLESSED 0x00000400UL
		31	#define F_CONV_BLESSED 0x00000800UL
		32	#define F_RELAXED 0x00001000UL
		33
		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
		39
		40	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
		41	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
17		42
18	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	43	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
19	#define F_DEFAULT 0	44	#define F_DEFAULT (9UL << S_MAXDEPTH)
20		45
21	#define INIT_SIZE 32 // initial scalar size to be allocated	46	#define INIT_SIZE 32 // initial scalar size to be allocated
		47	#define INDENT_STEP 3 // spaces per indentation level
		48
		49	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
22		50
23	#define SB do {	51	#define SB do {
24	#define SE } while (0)	52	#define SE } while (0)
25		53
26	static HV *json_stash;	54	#if __GNUC__ >= 3
		55	# define expect(expr,value) __builtin_expect ((expr),(value))
		56	# define inline inline
		57	#else
		58	# define expect(expr,value) (expr)
		59	# define inline static
		60	#endif
		61
		62	#define expect_false(expr) expect ((expr) != 0, 0)
		63	#define expect_true(expr) expect ((expr) != 0, 1)
		64
		65	#ifdef USE_ITHREADS
		66	# define JSON_SLOW 1
		67	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		68	#else
		69	# define JSON_SLOW 0
		70	# define JSON_STASH json_stash
		71	#endif
		72
		73	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		74	static SV *json_true, *json_false;
		75
		76	typedef struct {
		77	U32 flags;
		78	SV *cb_object;
		79	HV *cb_sk_object;
		80	} JSON;
		81
		82	/////////////////////////////////////////////////////////////////////////////
		83	// utility functions
		84
		85	inline void
		86	shrink (SV *sv)
		87	{
		88	sv_utf8_downgrade (sv, 1);
		89	if (SvLEN (sv) > SvCUR (sv) + 1)
		90	{
		91	#ifdef SvPV_shrink_to_cur
		92	SvPV_shrink_to_cur (sv);
		93	#elif defined (SvPV_renew)
		94	SvPV_renew (sv, SvCUR (sv) + 1);
		95	#endif
		96	}
		97	}
		98
		99	// decode an utf-8 character and return it, or (UV)-1 in
		100	// case of an error.
		101	// we special-case "safe" characters from U+80 .. U+7FF,
		102	// but use the very good perl function to parse anything else.
		103	// note that we never call this function for a ascii codepoints
		104	inline UV
		105	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		106	{
		107	if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
		108	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		109	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
		110	{
		111	*clen = 2;
		112	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		113	}
		114	else
		115	{
		116	*clen = (STRLEN)-1;
		117	return (UV)-1;
		118	}
		119	}
		120
		121	/////////////////////////////////////////////////////////////////////////////
		122	// encoder
27		123
28	// structure used for encoding JSON	124	// structure used for encoding JSON
29	typedef struct	125	typedef struct
30	{	126	{
31	char *cur;	127	char *cur; // SvPVX (sv) + current output position
32	STRLEN len; // SvLEN (sv)
33	char *end; // SvEND (sv)	128	char *end; // SvEND (sv)
34	SV *sv;	129	SV *sv; // result scalar
35	UV flags;	130	JSON json;
36	int max_recurse;	131	U32 indent; // indentation level
37	int indent;	132	U32 maxdepth; // max. indentation/recursion level
38	} enc_t;	133	} enc_t;
39		134
40	// structure used for decoding JSON	135	inline void
41	typedef struct
42	{
43	char *cur;
44	char *end;
45	const char *err;
46	UV flags;
47	} dec_t;
48
49	static UV *
50	SvJSON (SV *sv)
51	{
52	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
53	croak ("object is not of type JSON::XS");
54
55	return &SvUVX (SvRV (sv));
56	}
57
58	static void
59	shrink (SV *sv)
60	{
61	sv_utf8_downgrade (sv, 1);
62	#ifdef SvPV_shrink_to_cur
63	SvPV_shrink_to_cur (sv);
64	#endif
65	}
66
67	/////////////////////////////////////////////////////////////////////////////
68
69	static void
70	need (enc_t *enc, STRLEN len)	136	need (enc_t *enc, STRLEN len)
71	{	137	{
72	if (enc->cur + len >= enc->end)	138	if (expect_false (enc->cur + len >= enc->end))
73	{	139	{
74	STRLEN cur = enc->cur - SvPVX (enc->sv);	140	STRLEN cur = enc->cur - SvPVX (enc->sv);
75	SvGROW (enc->sv, cur + len + 1);	141	SvGROW (enc->sv, cur + len + 1);
76	enc->cur = SvPVX (enc->sv) + cur;	142	enc->cur = SvPVX (enc->sv) + cur;
77	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	143	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
78	}	144	}
79	}	145	}
80		146
81	static void	147	inline void
82	encode_ch (enc_t *enc, char ch)	148	encode_ch (enc_t *enc, char ch)
83	{	149	{
84	need (enc, 1);	150	need (enc, 1);
85	*enc->cur++ = ch;	151	*enc->cur++ = ch;
86	}	152	}
…		…
94		160
95	while (str < end)	161	while (str < end)
96	{	162	{
97	unsigned char ch = *(unsigned char *)str;	163	unsigned char ch = *(unsigned char *)str;
98		164
99	if (ch >= 0x20 && ch < 0x80) // most common case	165	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
100	{	166	{
101	if (ch == '"') // but with slow exceptions	167	if (expect_false (ch == '"')) // but with slow exceptions
102	{	168	{
103	need (enc, len += 1);	169	need (enc, len += 1);
104	*enc->cur++ = '\\';	170	*enc->cur++ = '\\';
105	*enc->cur++ = '"';	171	*enc->cur++ = '"';
106	}	172	}
107	else if (ch == '\\')	173	else if (expect_false (ch == '\\'))
108	{	174	{
109	need (enc, len += 1);	175	need (enc, len += 1);
110	*enc->cur++ = '\\';	176	*enc->cur++ = '\\';
111	*enc->cur++ = '\\';	177	*enc->cur++ = '\\';
112	}	178	}
…		…
130	STRLEN clen;	196	STRLEN clen;
131	UV uch;	197	UV uch;
132		198
133	if (is_utf8)	199	if (is_utf8)
134	{	200	{
135	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	201	uch = decode_utf8 (str, end - str, &clen);
136	if (clen == (STRLEN)-1)	202	if (clen == (STRLEN)-1)
137	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	203	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
138	}	204	}
139	else	205	else
140	{	206	{
…		…
143	}	209	}
144		210
145	if (uch > 0x10FFFFUL)	211	if (uch > 0x10FFFFUL)
146	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);	212	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
147		213
148	if (uch < 0x80 || enc->flags & F_ASCII)	214	if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
149	{	215	{
150	if (uch > 0xFFFFUL)	216	if (uch > 0xFFFFUL)
151	{	217	{
152	need (enc, len += 11);	218	need (enc, len += 11);
153	sprintf (enc->cur, "\\u%04x\\u%04x",	219	sprintf (enc->cur, "\\u%04x\\u%04x",
…		…
167	*enc->cur++ = hexdigit [(uch >> 0) & 15];	233	*enc->cur++ = hexdigit [(uch >> 0) & 15];
168	}	234	}
169		235
170	str += clen;	236	str += clen;
171	}	237	}
		238	else if (enc->json.flags & F_LATIN1)
		239	{
		240	*enc->cur++ = uch;
		241	str += clen;
		242	}
172	else if (is_utf8)	243	else if (is_utf8)
173	{	244	{
174	need (enc, len += clen);	245	need (enc, len += clen);
175	do	246	do
176	{	247	{
…		…
178	}	249	}
179	while (--clen);	250	while (--clen);
180	}	251	}
181	else	252	else
182	{	253	{
183	need (enc, len += 10); // never more than 11 bytes needed	254	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
184	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	255	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
185	++str;	256	++str;
186	}	257	}
187	}	258	}
188	}	259	}
…		…
190		261
191	--len;	262	--len;
192	}	263	}
193	}	264	}
194		265
195	#define INDENT SB \	266	inline void
		267	encode_indent (enc_t *enc)
		268	{
196	if (enc->flags & F_INDENT) \	269	if (enc->json.flags & F_INDENT)
197	{ \	270	{
198	int i_; \	271	int spaces = enc->indent * INDENT_STEP;
199	need (enc, enc->indent); \	272
200	for (i_ = enc->indent * 3; i_--; )\	273	need (enc, spaces);
		274	memset (enc->cur, ' ', spaces);
		275	enc->cur += spaces;
		276	}
		277	}
		278
		279	inline void
		280	encode_space (enc_t *enc)
		281	{
		282	need (enc, 1);
		283	encode_ch (enc, ' ');
		284	}
		285
		286	inline void
		287	encode_nl (enc_t *enc)
		288	{
		289	if (enc->json.flags & F_INDENT)
		290	{
		291	need (enc, 1);
201	encode_ch (enc, ' '); \	292	encode_ch (enc, '\n');
202	} \	293	}
203	SE	294	}
204		295
205	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	296	inline void
206	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	297	encode_comma (enc_t *enc)
207	#define COMMA SB \	298	{
208	encode_ch (enc, ','); \	299	encode_ch (enc, ',');
		300
209	if (enc->flags & F_INDENT) \	301	if (enc->json.flags & F_INDENT)
210	NL; \	302	encode_nl (enc);
211	else if (enc->flags & F_SPACE_AFTER) \	303	else if (enc->json.flags & F_SPACE_AFTER)
212	SPACE; \	304	encode_space (enc);
213	SE	305	}
214		306
215	static void encode_sv (enc_t *enc, SV *sv);	307	static void encode_sv (enc_t *enc, SV *sv);
216		308
217	static void	309	static void
218	encode_av (enc_t *enc, AV *av)	310	encode_av (enc_t *enc, AV *av)
219	{	311	{
220	int i, len = av_len (av);	312	int i, len = av_len (av);
221		313
		314	if (enc->indent >= enc->maxdepth)
		315	croak ("data structure too deep (hit recursion limit)");
		316
222	encode_ch (enc, '['); NL;	317	encode_ch (enc, '[');
223	++enc->indent;	318
		319	if (len >= 0)
		320	{
		321	encode_nl (enc); ++enc->indent;
224		322
225	for (i = 0; i <= len; ++i)	323	for (i = 0; i <= len; ++i)
226	{	324	{
227	INDENT;	325	SV **svp = av_fetch (av, i, 0);
228	encode_sv (enc, *av_fetch (av, i, 0));
229		326
		327	encode_indent (enc);
		328
		329	if (svp)
		330	encode_sv (enc, *svp);
		331	else
		332	encode_str (enc, "null", 4, 0);
		333
230	if (i < len)	334	if (i < len)
231	COMMA;	335	encode_comma (enc);
232	}	336	}
233		337
234	NL;	338	encode_nl (enc); --enc->indent; encode_indent (enc);
235		339	}
236	--enc->indent;	340
237	INDENT; encode_ch (enc, ']');	341	encode_ch (enc, ']');
238	}	342	}
239		343
240	static void	344	static void
241	encode_he (enc_t *enc, HE *he)	345	encode_hk (enc_t *enc, HE *he)
242	{	346	{
243	encode_ch (enc, '"');	347	encode_ch (enc, '"');
244		348
245	if (HeKLEN (he) == HEf_SVKEY)	349	if (HeKLEN (he) == HEf_SVKEY)
246	{	350	{
…		…
256	else	360	else
257	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	361	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
258		362
259	encode_ch (enc, '"');	363	encode_ch (enc, '"');
260		364
261	if (enc->flags & F_SPACE_BEFORE) SPACE;	365	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
262	encode_ch (enc, ':');	366	encode_ch (enc, ':');
263	if (enc->flags & F_SPACE_AFTER ) SPACE;	367	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
264	encode_sv (enc, HeVAL (he));
265	}	368	}
266		369
267	// compare hash entries, used when all keys are bytestrings	370	// compare hash entries, used when all keys are bytestrings
268	static int	371	static int
269	he_cmp_fast (const void *a_, const void *b_)	372	he_cmp_fast (const void *a_, const void *b_)
…		…
274	HE *b = *(HE **)b_;	377	HE *b = *(HE **)b_;
275		378
276	STRLEN la = HeKLEN (a);	379	STRLEN la = HeKLEN (a);
277	STRLEN lb = HeKLEN (b);	380	STRLEN lb = HeKLEN (b);
278		381
279	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	382	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
280	cmp = la - lb;	383	cmp = lb - la;
281		384
282	return cmp;	385	return cmp;
283	}	386	}
284		387
285	// compare hash entries, used when some keys are sv's or utf-x	388	// compare hash entries, used when some keys are sv's or utf-x
286	static int	389	static int
287	he_cmp_slow (const void *a, const void *b)	390	he_cmp_slow (const void *a, const void *b)
288	{	391	{
289	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	392	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
290	}	393	}
291		394
292	static void	395	static void
293	encode_hv (enc_t *enc, HV *hv)	396	encode_hv (enc_t *enc, HV *hv)
294	{	397	{
		398	HE *he;
295	int count, i;	399	int count;
296		400
297	encode_ch (enc, '{'); NL; ++enc->indent;	401	if (enc->indent >= enc->maxdepth)
		402	croak ("data structure too deep (hit recursion limit)");
298		403
299	if ((count = hv_iterinit (hv)))	404	encode_ch (enc, '{');
300	{	405
301	// for canonical output we have to sort by keys first	406	// for canonical output we have to sort by keys first
302	// actually, this is mostly due to the stupid so-called	407	// actually, this is mostly due to the stupid so-called
303	// security workaround added somewhere in 5.8.x.	408	// security workaround added somewhere in 5.8.x.
304	// that randomises hash orderings	409	// that randomises hash orderings
305	if (enc->flags & F_CANONICAL)	410	if (enc->json.flags & F_CANONICAL)
		411	{
		412	int count = hv_iterinit (hv);
		413
		414	if (SvMAGICAL (hv))
306	{	415	{
307	HE *he, *hes [count];	416	// need to count by iterating. could improve by dynamically building the vector below
		417	// but I don't care for the speed of this special case.
		418	// note also that we will run into undefined behaviour when the two iterations
		419	// do not result in the same count, something I might care for in some later release.
		420
		421	count = 0;
		422	while (hv_iternext (hv))
		423	++count;
		424
		425	hv_iterinit (hv);
		426	}
		427
		428	if (count)
		429	{
308	int fast = 1;	430	int i, fast = 1;
		431	#if defined(__BORLANDC__) || defined(_MSC_VER)
		432	HE **hes = _alloca (count * sizeof (HE));
		433	#else
		434	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		435	#endif
309		436
310	i = 0;	437	i = 0;
311	while ((he = hv_iternext (hv)))	438	while ((he = hv_iternext (hv)))
312	{	439	{
313	hes [i++] = he;	440	hes [i++] = he;
…		…
335		462
336	FREETMPS;	463	FREETMPS;
337	LEAVE;	464	LEAVE;
338	}	465	}
339		466
340	for (i = 0; i < count; ++i)	467	encode_nl (enc); ++enc->indent;
		468
		469	while (count--)
341	{	470	{
342	INDENT;	471	encode_indent (enc);
		472	he = hes [count];
343	encode_he (enc, hes [i]);	473	encode_hk (enc, he);
		474	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
344		475
345	if (i < count - 1)	476	if (count)
346	COMMA;	477	encode_comma (enc);
347	}	478	}
348		479
		480	encode_nl (enc); --enc->indent; encode_indent (enc);
		481	}
		482	}
		483	else
		484	{
		485	if (hv_iterinit (hv) || SvMAGICAL (hv))
		486	if ((he = hv_iternext (hv)))
349	NL;	487	{
		488	encode_nl (enc); ++enc->indent;
		489
		490	for (;;)
		491	{
		492	encode_indent (enc);
		493	encode_hk (enc, he);
		494	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		495
		496	if (!(he = hv_iternext (hv)))
		497	break;
		498
		499	encode_comma (enc);
		500	}
		501
		502	encode_nl (enc); --enc->indent; encode_indent (enc);
		503	}
		504	}
		505
		506	encode_ch (enc, '}');
		507	}
		508
		509	// encode objects, arrays and special \0=false and \1=true values.
		510	static void
		511	encode_rv (enc_t *enc, SV *sv)
		512	{
		513	svtype svt;
		514
		515	SvGETMAGIC (sv);
		516	svt = SvTYPE (sv);
		517
		518	if (expect_false (SvOBJECT (sv)))
		519	{
		520	HV *stash = !JSON_SLOW || json_boolean_stash
		521	? json_boolean_stash
		522	: gv_stashpv ("JSON::XS::Boolean", 1);
		523
		524	if (SvSTASH (sv) == stash)
		525	{
		526	if (SvIV (sv))
		527	encode_str (enc, "true", 4, 0);
		528	else
		529	encode_str (enc, "false", 5, 0);
350	}	530	}
351	else	531	else
352	{	532	{
353	SV *sv;	533	#if 0
354	HE *he = hv_iternext (hv);	534	if (0 && sv_derived_from (rv, "JSON::Literal"))
355
356	for (;;)
357	{	535	{
358	INDENT;	536	// not yet
359	encode_he (enc, he);
360
361	if (!(he = hv_iternext (hv)))
362	break;
363
364	COMMA;
365	}	537	}
366		538	#endif
		539	if (enc->json.flags & F_CONV_BLESSED)
367	NL;	540	{
		541	// we re-bless the reference to get overload and other niceties right
		542	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
		543
		544	if (to_json)
		545	{
		546	dSP;
		547
		548	ENTER; SAVETMPS; PUSHMARK (SP);
		549	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		550
		551	// calling with G_SCALAR ensures that we always get a 1 return value
		552	PUTBACK;
		553	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		554	SPAGAIN;
		555
		556	// catch this surprisingly common error
		557	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		558	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		559
		560	sv = POPs;
		561	PUTBACK;
		562
		563	encode_sv (enc, sv);
		564
		565	FREETMPS; LEAVE;
		566	}
		567	else if (enc->json.flags & F_ALLOW_BLESSED)
		568	encode_str (enc, "null", 4, 0);
		569	else
		570	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		571	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		572	}
		573	else if (enc->json.flags & F_ALLOW_BLESSED)
		574	encode_str (enc, "null", 4, 0);
		575	else
		576	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		577	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
368	}	578	}
369	}	579	}
		580	else if (svt == SVt_PVHV)
		581	encode_hv (enc, (HV *)sv);
		582	else if (svt == SVt_PVAV)
		583	encode_av (enc, (AV *)sv);
		584	else if (svt < SVt_PVAV)
		585	{
		586	STRLEN len = 0;
		587	char *pv = svt ? SvPV (sv, len) : 0;
370		588
371	--enc->indent; INDENT; encode_ch (enc, '}');	589	if (len == 1 && *pv == '1')
		590	encode_str (enc, "true", 4, 0);
		591	else if (len == 1 && *pv == '0')
		592	encode_str (enc, "false", 5, 0);
		593	else
		594	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		595	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		596	}
		597	else
		598	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		599	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
372	}	600	}
373		601
374	static void	602	static void
375	encode_sv (enc_t *enc, SV *sv)	603	encode_sv (enc_t *enc, SV *sv)
376	{	604	{
…		…
384	encode_str (enc, str, len, SvUTF8 (sv));	612	encode_str (enc, str, len, SvUTF8 (sv));
385	encode_ch (enc, '"');	613	encode_ch (enc, '"');
386	}	614	}
387	else if (SvNOKp (sv))	615	else if (SvNOKp (sv))
388	{	616	{
		617	// trust that perl will do the right thing w.r.t. JSON syntax.
389	need (enc, NV_DIG + 32);	618	need (enc, NV_DIG + 32);
390	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	619	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
391	enc->cur += strlen (enc->cur);	620	enc->cur += strlen (enc->cur);
392	}	621	}
393	else if (SvIOKp (sv))	622	else if (SvIOKp (sv))
394	{	623	{
395	need (enc, 64);	624	// we assume we can always read an IV as a UV
		625	if (SvUV (sv) & ~(UV)0x7fff)
		626	{
		627	// large integer, use the (rather slow) snprintf way.
		628	need (enc, sizeof (UV) * 3);
396	enc->cur +=	629	enc->cur +=
397	SvIsUV(sv)	630	SvIsUV(sv)
398	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	631	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
399	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	632	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		633	}
		634	else
		635	{
		636	// optimise the "small number case"
		637	// code will likely be branchless and use only a single multiplication
		638	I32 i = SvIV (sv);
		639	U32 u;
		640	char digit, nz = 0;
		641
		642	need (enc, 6);
		643
		644	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		645	u = i < 0 ? -i : i;
		646
		647	// convert to 4.28 fixed-point representation
		648	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		649
		650	// 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,
		652	// resulting in a net multiplication by 10.
		653	// we always write the digit to memory but conditionally increment
		654	// the pointer, to ease the usage of conditional move instructions.
		655	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
		656	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
		657	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
		658	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
		659	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		660	}
400	}	661	}
401	else if (SvROK (sv))	662	else if (SvROK (sv))
402	{	663	encode_rv (enc, SvRV (sv));
403	SV *rv = SvRV (sv);
404
405	if (!--enc->max_recurse)
406	croak ("data structure too deep (hit recursion limit)");
407
408	switch (SvTYPE (rv))
409	{
410	case SVt_PVAV: encode_av (enc, (AV *)rv); break;
411	case SVt_PVHV: encode_hv (enc, (HV *)rv); break;
412
413	default:
414	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
415	SvPV_nolen (sv));
416	}
417	}
418	else if (!SvOK (sv))	664	else if (!SvOK (sv))
419	encode_str (enc, "null", 4, 0);	665	encode_str (enc, "null", 4, 0);
420	else	666	else
421	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	667	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
422	SvPV_nolen (sv), SvFLAGS (sv));	668	SvPV_nolen (sv), SvFLAGS (sv));
423	}	669	}
424		670
425	static SV *	671	static SV *
426	encode_json (SV *scalar, UV flags)	672	encode_json (SV *scalar, JSON *json)
427	{	673	{
		674	enc_t enc;
		675
428	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	676	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
429	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	677	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
430		678
431	enc_t enc;	679	enc.json = *json;
432	enc.flags = flags;
433	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	680	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
434	enc.cur = SvPVX (enc.sv);	681	enc.cur = SvPVX (enc.sv);
435	enc.end = SvEND (enc.sv);	682	enc.end = SvEND (enc.sv);
436	enc.max_recurse = 0;
437	enc.indent = 0;	683	enc.indent = 0;
		684	enc.maxdepth = DEC_DEPTH (enc.json.flags);
438		685
439	SvPOK_only (enc.sv);	686	SvPOK_only (enc.sv);
440	encode_sv (&enc, scalar);	687	encode_sv (&enc, scalar);
441		688
		689	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		690	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		691
442	if (!(flags & (F_ASCII | F_UTF8)))	692	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
443	SvUTF8_on (enc.sv);	693	SvUTF8_on (enc.sv);
444		694
445	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
446
447	if (enc.flags & F_SHRINK)	695	if (enc.json.flags & F_SHRINK)
448	shrink (enc.sv);	696	shrink (enc.sv);
449		697
450	return enc.sv;	698	return enc.sv;
451	}	699	}
452		700
453	/////////////////////////////////////////////////////////////////////////////	701	/////////////////////////////////////////////////////////////////////////////
		702	// decoder
454		703
455	#define WS \	704	// structure used for decoding JSON
		705	typedef struct
		706	{
		707	char *cur; // current parser pointer
		708	char *end; // end of input string
		709	const char *err; // parse error, if != 0
		710	JSON json;
		711	U32 depth; // recursion depth
		712	U32 maxdepth; // recursion depth limit
		713	} dec_t;
		714
		715	inline void
		716	decode_comment (dec_t *dec)
		717	{
		718	// only '#'-style comments allowed a.t.m.
		719
		720	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		721	++dec->cur;
		722	}
		723
		724	inline void
		725	decode_ws (dec_t *dec)
		726	{
456	for (;;) \	727	for (;;)
457	{ \	728	{
458	char ch = *dec->cur; \	729	char ch = *dec->cur;
		730
459	if (ch > 0x20 \	731	if (ch > 0x20)
		732	{
		733	if (expect_false (ch == '#'))
		734	{
		735	if (dec->json.flags & F_RELAXED)
		736	decode_comment (dec);
		737	else
		738	break;
		739	}
		740	else
		741	break;
		742	}
460	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	743	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
461	break; \	744	break; // parse error, but let higher level handle it, gives better error messages
		745
462	++dec->cur; \	746	++dec->cur;
463	}	747	}
		748	}
464		749
465	#define ERR(reason) SB dec->err = reason; goto fail; SE	750	#define ERR(reason) SB dec->err = reason; goto fail; SE
		751
466	#define EXPECT_CH(ch) SB \	752	#define EXPECT_CH(ch) SB \
467	if (*dec->cur != ch) \	753	if (*dec->cur != ch) \
468	ERR (# ch " expected"); \	754	ERR (# ch " expected"); \
469	++dec->cur; \	755	++dec->cur; \
470	SE	756	SE
471		757
		758	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		759	#define DEC_DEC_DEPTH --dec->depth
		760
472	static SV *decode_sv (dec_t *dec);	761	static SV *decode_sv (dec_t *dec);
473		762
474	static signed char decode_hexdigit[256];	763	static signed char decode_hexdigit[256];
475		764
476	static UV	765	static UV
477	decode_4hex (dec_t *dec)	766	decode_4hex (dec_t *dec)
478	{	767	{
479	signed char d1, d2, d3, d4;	768	signed char d1, d2, d3, d4;
		769	unsigned char *cur = (unsigned char *)dec->cur;
480		770
481	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	771	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
482	if (d1 < 0) ERR ("four hexadecimal digits expected");	772	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
483	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	773	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
484	if (d2 < 0) ERR ("four hexadecimal digits expected");	774	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
485	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
486	if (d3 < 0) ERR ("four hexadecimal digits expected");
487	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
488	if (d4 < 0) ERR ("four hexadecimal digits expected");
489		775
490	dec->cur += 4;	776	dec->cur += 4;
491		777
492	return ((UV)d1) << 12	778	return ((UV)d1) << 12
493	| ((UV)d2) << 8	779	| ((UV)d2) << 8
…		…
496		782
497	fail:	783	fail:
498	return (UV)-1;	784	return (UV)-1;
499	}	785	}
500		786
501	#define APPEND_GROW(n) SB \
502	if (cur + (n) >= end) \
503	{ \
504	STRLEN ofs = cur - SvPVX (sv); \
505	SvGROW (sv, ofs + (n) + 1); \
506	cur = SvPVX (sv) + ofs; \
507	end = SvEND (sv); \
508	} \
509	SE
510
511	#define APPEND_CH(ch) SB \
512	APPEND_GROW (1); \
513	*cur++ = (ch); \
514	SE
515
516	static SV *	787	static SV *
517	decode_str (dec_t *dec)	788	decode_str (dec_t *dec)
518	{	789	{
519	SV *sv = NEWSV (0,2);	790	SV *sv = 0;
520	int utf8 = 0;	791	int utf8 = 0;
521	char *cur = SvPVX (sv);	792	char *dec_cur = dec->cur;
522	char *end = SvEND (sv);
523		793
524	for (;;)	794	do
525	{	795	{
526	unsigned char ch = *(unsigned char *)dec->cur;	796	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		797	char *cur = buf;
527		798
528	if (ch == '"')	799	do
529	break;
530	else if (ch == '\\')
531	{	800	{
532	switch (*++dec->cur)	801	unsigned char ch = *(unsigned char *)dec_cur++;
		802
		803	if (expect_false (ch == '"'))
533	{	804	{
534	case '\\':	805	--dec_cur;
535	case '/':	806	break;
536	case '"': APPEND_CH (*dec->cur++); break;	807	}
537		808	else if (expect_false (ch == '\\'))
538	case 'b': APPEND_CH ('\010'); ++dec->cur; break;	809	{
539	case 't': APPEND_CH ('\011'); ++dec->cur; break;	810	switch (*dec_cur)
540	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
541	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
542	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
543
544	case 'u':
545	{	811	{
546	UV lo, hi;	812	case '\\':
547	++dec->cur;	813	case '/':
		814	case '"': *cur++ = *dec_cur++; break;
548		815
549	hi = decode_4hex (dec);	816	case 'b': ++dec_cur; *cur++ = '\010'; break;
550	if (hi == (UV)-1)	817	case 't': ++dec_cur; *cur++ = '\011'; break;
551	goto fail;	818	case 'n': ++dec_cur; *cur++ = '\012'; break;
		819	case 'f': ++dec_cur; *cur++ = '\014'; break;
		820	case 'r': ++dec_cur; *cur++ = '\015'; break;
552		821
553	// possibly a surrogate pair	822	case 'u':
554	if (hi >= 0xd800 && hi < 0xdc00)
555	{	823	{
556	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	824	UV lo, hi;
557	ERR ("missing low surrogate character in surrogate pair");	825	++dec_cur;
558		826
559	dec->cur += 2;	827	dec->cur = dec_cur;
560
561	lo = decode_4hex (dec);	828	hi = decode_4hex (dec);
		829	dec_cur = dec->cur;
562	if (lo == (UV)-1)	830	if (hi == (UV)-1)
563	goto fail;	831	goto fail;
564		832
		833	// possibly a surrogate pair
		834	if (hi >= 0xd800)
		835	if (hi < 0xdc00)
		836	{
		837	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		838	ERR ("missing low surrogate character in surrogate pair");
		839
		840	dec_cur += 2;
		841
		842	dec->cur = dec_cur;
		843	lo = decode_4hex (dec);
		844	dec_cur = dec->cur;
		845	if (lo == (UV)-1)
		846	goto fail;
		847
565	if (lo < 0xdc00 || lo >= 0xe000)	848	if (lo < 0xdc00 || lo >= 0xe000)
566	ERR ("surrogate pair expected");	849	ERR ("surrogate pair expected");
567		850
568	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	851	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		852	}
		853	else if (hi < 0xe000)
		854	ERR ("missing high surrogate character in surrogate pair");
		855
		856	if (hi >= 0x80)
		857	{
		858	utf8 = 1;
		859
		860	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
		861	}
		862	else
		863	*cur++ = hi;
569	}	864	}
570	else if (hi >= 0xdc00 && hi < 0xe000)
571	ERR ("missing high surrogate character in surrogate pair");
572
573	if (hi >= 0x80)
574	{	865	break;
575	utf8 = 1;
576		866
577	APPEND_GROW (4); // at most 4 bytes for 21 bits
578	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
579	}
580	else	867	default:
581	APPEND_CH (hi);	868	--dec_cur;
		869	ERR ("illegal backslash escape sequence in string");
582	}	870	}
583	break;	871	}
		872	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
		873	*cur++ = ch;
		874	else if (ch >= 0x80)
		875	{
		876	STRLEN clen;
		877	UV uch;
584		878
585	default:
586	--dec->cur;	879	--dec_cur;
587	ERR ("illegal backslash escape sequence in string");	880
		881	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		882	if (clen == (STRLEN)-1)
		883	ERR ("malformed UTF-8 character in JSON string");
		884
		885	do
		886	*cur++ = *dec_cur++;
		887	while (--clen);
		888
		889	utf8 = 1;
		890	}
		891	else
		892	{
		893	--dec_cur;
		894
		895	if (!ch)
		896	ERR ("unexpected end of string while parsing JSON string");
		897	else
		898	ERR ("invalid character encountered while parsing JSON string");
588	}	899	}
589	}	900	}
590	else if (ch >= 0x20 && ch <= 0x7f)	901	while (cur < buf + SHORT_STRING_LEN);
591	APPEND_CH (*dec->cur++);	902
592	else if (ch >= 0x80)
593	{	903	{
594	STRLEN clen;	904	STRLEN len = cur - buf;
595	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
596	if (clen == (STRLEN)-1)
597	ERR ("malformed UTF-8 character in JSON string");
598		905
599	APPEND_GROW (clen);	906	if (sv)
600	do
601	{	907	{
602	*cur++ = *dec->cur++;	908	SvGROW (sv, SvCUR (sv) + len + 1);
		909	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		910	SvCUR_set (sv, SvCUR (sv) + len);
603	}	911	}
604	while (--clen);
605
606	utf8 = 1;
607	}
608	else if (dec->cur == dec->end)
609	ERR ("unexpected end of string while parsing json string");
610	else	912	else
611	ERR ("invalid character encountered");	913	sv = newSVpvn (buf, len);
612	}	914	}
		915	}
		916	while (*dec_cur != '"');
613		917
614	++dec->cur;	918	++dec_cur;
615		919
616	SvCUR_set (sv, cur - SvPVX (sv));	920	if (sv)
617		921	{
618	SvPOK_only (sv);	922	SvPOK_only (sv);
619	*SvEND (sv) = 0;	923	*SvEND (sv) = 0;
620		924
621	if (utf8)	925	if (utf8)
622	SvUTF8_on (sv);	926	SvUTF8_on (sv);
		927	}
		928	else
		929	sv = newSVpvn ("", 0);
623		930
624	if (dec->flags & F_SHRINK)	931	dec->cur = dec_cur;
625	shrink (sv);
626
627	return sv;	932	return sv;
628		933
629	fail:	934	fail:
630	SvREFCNT_dec (sv);	935	dec->cur = dec_cur;
631	return 0;	936	return 0;
632	}	937	}
633		938
634	static SV *	939	static SV *
635	decode_num (dec_t *dec)	940	decode_num (dec_t *dec)
…		…
693	is_nv = 1;	998	is_nv = 1;
694	}	999	}
695		1000
696	if (!is_nv)	1001	if (!is_nv)
697	{	1002	{
698	UV uv;	1003	int len = dec->cur - start;
699	int numtype = grok_number (start, dec->cur - start, &uv);	1004
700	if (numtype & IS_NUMBER_IN_UV)	1005	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
701	if (numtype & IS_NUMBER_NEG)	1006	if (*start == '-')
		1007	switch (len)
702	{	1008	{
703	if (uv < (UV)IV_MIN)	1009	case 2: return newSViv (-( start [1] - '0' * 1));
704	return newSViv (-(IV)uv);	1010	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		1011	case 4: return newSViv (-( 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));
705	}	1013	}
		1014	else
		1015	switch (len)
		1016	{
		1017	case 1: return newSViv ( start [0] - '0' * 1);
		1018	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);
		1020	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1021	}
		1022
		1023	{
		1024	UV uv;
		1025	int numtype = grok_number (start, len, &uv);
		1026	if (numtype & IS_NUMBER_IN_UV)
		1027	if (numtype & IS_NUMBER_NEG)
		1028	{
		1029	if (uv < (UV)IV_MIN)
		1030	return newSViv (-(IV)uv);
		1031	}
706	else	1032	else
707	return newSVuv (uv);	1033	return newSVuv (uv);
708	}	1034	}
709		1035
		1036	len -= *start == '-' ? 1 : 0;
		1037
		1038	// does not fit into IV or UV, try NV
		1039	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		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
		1045	return newSVnv (Atof (start));
		1046
		1047	// everything else fails, convert it to a string
		1048	return newSVpvn (start, dec->cur - start);
		1049	}
		1050
		1051	// loss of precision here
710	return newSVnv (Atof (start));	1052	return newSVnv (Atof (start));
711		1053
712	fail:	1054	fail:
713	return 0;	1055	return 0;
714	}	1056	}
…		…
716	static SV *	1058	static SV *
717	decode_av (dec_t *dec)	1059	decode_av (dec_t *dec)
718	{	1060	{
719	AV *av = newAV ();	1061	AV *av = newAV ();
720		1062
721	WS;	1063	DEC_INC_DEPTH;
		1064	decode_ws (dec);
		1065
722	if (*dec->cur == ']')	1066	if (*dec->cur == ']')
723	++dec->cur;	1067	++dec->cur;
724	else	1068	else
725	for (;;)	1069	for (;;)
726	{	1070	{
…		…
730	if (!value)	1074	if (!value)
731	goto fail;	1075	goto fail;
732		1076
733	av_push (av, value);	1077	av_push (av, value);
734		1078
735	WS;	1079	decode_ws (dec);
736		1080
737	if (*dec->cur == ']')	1081	if (*dec->cur == ']')
738	{	1082	{
739	++dec->cur;	1083	++dec->cur;
740	break;	1084	break;
…		…
742		1086
743	if (*dec->cur != ',')	1087	if (*dec->cur != ',')
744	ERR (", or ] expected while parsing array");	1088	ERR (", or ] expected while parsing array");
745		1089
746	++dec->cur;	1090	++dec->cur;
		1091
		1092	decode_ws (dec);
		1093
		1094	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1095	{
		1096	++dec->cur;
		1097	break;
		1098	}
747	}	1099	}
748		1100
		1101	DEC_DEC_DEPTH;
749	return newRV_noinc ((SV *)av);	1102	return newRV_noinc ((SV *)av);
750		1103
751	fail:	1104	fail:
752	SvREFCNT_dec (av);	1105	SvREFCNT_dec (av);
		1106	DEC_DEC_DEPTH;
753	return 0;	1107	return 0;
754	}	1108	}
755		1109
756	static SV *	1110	static SV *
757	decode_hv (dec_t *dec)	1111	decode_hv (dec_t *dec)
758	{	1112	{
		1113	SV *sv;
759	HV *hv = newHV ();	1114	HV *hv = newHV ();
760		1115
761	WS;	1116	DEC_INC_DEPTH;
		1117	decode_ws (dec);
		1118
762	if (*dec->cur == '}')	1119	if (*dec->cur == '}')
763	++dec->cur;	1120	++dec->cur;
764	else	1121	else
765	for (;;)	1122	for (;;)
766	{	1123	{
767	SV *key, *value;
768
769	WS; EXPECT_CH ('"');	1124	EXPECT_CH ('"');
770		1125
771	key = decode_str (dec);	1126	// heuristic: assume that
772	if (!key)	1127	// a) decode_str + hv_store_ent are abysmally slow.
773	goto fail;	1128	// b) most hash keys are short, simple ascii text.
		1129	// => try to "fast-match" such strings to avoid
		1130	// the overhead of decode_str + hv_store_ent.
		1131	{
		1132	SV *value;
		1133	char *p = dec->cur;
		1134	char *e = p + 24; // only try up to 24 bytes
774		1135
775	WS; EXPECT_CH (':');	1136	for (;;)
776
777	value = decode_sv (dec);
778	if (!value)
779	{	1137	{
		1138	// the >= 0x80 is true on most architectures
		1139	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1140	{
		1141	// slow path, back up and use decode_str
		1142	SV *key = decode_str (dec);
		1143	if (!key)
		1144	goto fail;
		1145
		1146	decode_ws (dec); EXPECT_CH (':');
		1147
		1148	decode_ws (dec);
		1149	value = decode_sv (dec);
		1150	if (!value)
		1151	{
		1152	SvREFCNT_dec (key);
		1153	goto fail;
		1154	}
		1155
		1156	hv_store_ent (hv, key, value, 0);
780	SvREFCNT_dec (key);	1157	SvREFCNT_dec (key);
		1158
		1159	break;
		1160	}
		1161	else if (*p == '"')
		1162	{
		1163	// fast path, got a simple key
		1164	char *key = dec->cur;
		1165	int len = p - key;
		1166	dec->cur = p + 1;
		1167
		1168	decode_ws (dec); EXPECT_CH (':');
		1169
		1170	decode_ws (dec);
		1171	value = decode_sv (dec);
		1172	if (!value)
781	goto fail;	1173	goto fail;
		1174
		1175	hv_store (hv, key, len, value, 0);
		1176
		1177	break;
		1178	}
		1179
		1180	++p;
782	}	1181	}
783
784	//TODO: optimise
785	hv_store_ent (hv, key, value, 0);
786
787	WS;	1182	}
		1183
		1184	decode_ws (dec);
788		1185
789	if (*dec->cur == '}')	1186	if (*dec->cur == '}')
790	{	1187	{
791	++dec->cur;	1188	++dec->cur;
792	break;	1189	break;
…		…
794		1191
795	if (*dec->cur != ',')	1192	if (*dec->cur != ',')
796	ERR (", or } expected while parsing object/hash");	1193	ERR (", or } expected while parsing object/hash");
797		1194
798	++dec->cur;	1195	++dec->cur;
		1196
		1197	decode_ws (dec);
		1198
		1199	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1200	{
		1201	++dec->cur;
		1202	break;
		1203	}
799	}	1204	}
800		1205
		1206	DEC_DEC_DEPTH;
801	return newRV_noinc ((SV *)hv);	1207	sv = newRV_noinc ((SV *)hv);
		1208
		1209	// check filter callbacks
		1210	if (dec->json.flags & F_HOOK)
		1211	{
		1212	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1213	{
		1214	HE *cb, *he;
		1215
		1216	hv_iterinit (hv);
		1217	he = hv_iternext (hv);
		1218	hv_iterinit (hv);
		1219
		1220	// the next line creates a mortal sv each time its called.
		1221	// might want to optimise this for common cases.
		1222	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1223
		1224	if (cb)
		1225	{
		1226	dSP;
		1227	int count;
		1228
		1229	ENTER; SAVETMPS; PUSHMARK (SP);
		1230	XPUSHs (HeVAL (he));
		1231
		1232	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1233
		1234	if (count == 1)
		1235	{
		1236	sv = newSVsv (POPs);
		1237	FREETMPS; LEAVE;
		1238	return sv;
		1239	}
		1240
		1241	FREETMPS; LEAVE;
		1242	}
		1243	}
		1244
		1245	if (dec->json.cb_object)
		1246	{
		1247	dSP;
		1248	int count;
		1249
		1250	ENTER; SAVETMPS; PUSHMARK (SP);
		1251	XPUSHs (sv_2mortal (sv));
		1252
		1253	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1254
		1255	if (count == 1)
		1256	{
		1257	sv = newSVsv (POPs);
		1258	FREETMPS; LEAVE;
		1259	return sv;
		1260	}
		1261
		1262	SvREFCNT_inc (sv);
		1263	FREETMPS; LEAVE;
		1264	}
		1265	}
		1266
		1267	return sv;
802		1268
803	fail:	1269	fail:
804	SvREFCNT_dec (hv);	1270	SvREFCNT_dec (hv);
		1271	DEC_DEC_DEPTH;
805	return 0;	1272	return 0;
806	}	1273	}
807		1274
808	static SV *	1275	static SV *
809	decode_sv (dec_t *dec)	1276	decode_sv (dec_t *dec)
810	{	1277	{
811	WS;	1278	// the beauty of JSON: you need exactly one character lookahead
		1279	// to parse anything.
812	switch (*dec->cur)	1280	switch (*dec->cur)
813	{	1281	{
814	case '"': ++dec->cur; return decode_str (dec);	1282	case '"': ++dec->cur; return decode_str (dec);
815	case '[': ++dec->cur; return decode_av (dec);	1283	case '[': ++dec->cur; return decode_av (dec);
816	case '{': ++dec->cur; return decode_hv (dec);	1284	case '{': ++dec->cur; return decode_hv (dec);
…		…
822		1290
823	case 't':	1291	case 't':
824	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1292	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
825	{	1293	{
826	dec->cur += 4;	1294	dec->cur += 4;
827	return newSViv (1);	1295	#if JSON_SLOW
		1296	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);
		1297	#endif
		1298	return SvREFCNT_inc (json_true);
828	}	1299	}
829	else	1300	else
830	ERR ("'true' expected");	1301	ERR ("'true' expected");
831		1302
832	break;	1303	break;
833		1304
834	case 'f':	1305	case 'f':
835	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1306	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
836	{	1307	{
837	dec->cur += 5;	1308	dec->cur += 5;
838	return newSViv (0);	1309	#if JSON_SLOW
		1310	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1311	#endif
		1312	return SvREFCNT_inc (json_false);
839	}	1313	}
840	else	1314	else
841	ERR ("'false' expected");	1315	ERR ("'false' expected");
842		1316
843	break;	1317	break;
…		…
852	ERR ("'null' expected");	1326	ERR ("'null' expected");
853		1327
854	break;	1328	break;
855		1329
856	default:	1330	default:
857	ERR ("malformed json string, neither array, object, number, string or atom");	1331	ERR ("malformed JSON string, neither array, object, number, string or atom");
858	break;	1332	break;
859	}	1333	}
860		1334
861	fail:	1335	fail:
862	return 0;	1336	return 0;
863	}	1337	}
864		1338
865	static SV *	1339	static SV *
866	decode_json (SV *string, UV flags)	1340	decode_json (SV *string, JSON *json, UV *offset_return)
867	{	1341	{
		1342	dec_t dec;
		1343	UV offset;
868	SV *sv;	1344	SV *sv;
869		1345
		1346	SvGETMAGIC (string);
		1347	SvUPGRADE (string, SVt_PV);
		1348
		1349	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1350	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));
		1352
870	if (flags & F_UTF8)	1353	if (json->flags & F_UTF8)
871	sv_utf8_downgrade (string, 0);	1354	sv_utf8_downgrade (string, 0);
872	else	1355	else
873	sv_utf8_upgrade (string);	1356	sv_utf8_upgrade (string);
874		1357
875	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1358	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
876		1359
877	dec_t dec;	1360	dec.json = *json;
878	dec.flags = flags;
879	dec.cur = SvPVX (string);	1361	dec.cur = SvPVX (string);
880	dec.end = SvEND (string);	1362	dec.end = SvEND (string);
881	dec.err = 0;	1363	dec.err = 0;
		1364	dec.depth = 0;
		1365	dec.maxdepth = DEC_DEPTH (dec.json.flags);
882		1366
		1367	if (dec.json.cb_object || dec.json.cb_sk_object)
		1368	dec.json.flags |= F_HOOK;
		1369
		1370	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1371
		1372	decode_ws (&dec);
883	sv = decode_sv (&dec);	1373	sv = decode_sv (&dec);
884		1374
		1375	if (!(offset_return || !sv))
		1376	{
		1377	// check for trailing garbage
		1378	decode_ws (&dec);
		1379
		1380	if (*dec.cur)
		1381	{
		1382	dec.err = "garbage after JSON object";
		1383	SvREFCNT_dec (sv);
		1384	sv = 0;
		1385	}
		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	}
		1397
885	if (!sv)	1398	if (!sv)
886	{	1399	{
887	IV offset = dec.flags & F_UTF8
888	? dec.cur - SvPVX (string)
889	: utf8_distance (dec.cur, SvPVX (string));
890	SV *uni = sv_newmortal ();	1400	SV *uni = sv_newmortal ();
891		1401
892	// horrible hack to silence warning inside pv_uni_display	1402	// horrible hack to silence warning inside pv_uni_display
893	COP cop = *PL_curcop;	1403	COP cop = *PL_curcop;
894	cop.cop_warnings = pWARN_NONE;	1404	cop.cop_warnings = pWARN_NONE;
…		…
896	SAVEVPTR (PL_curcop);	1406	SAVEVPTR (PL_curcop);
897	PL_curcop = &cop;	1407	PL_curcop = &cop;
898	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1408	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
899	LEAVE;	1409	LEAVE;
900		1410
901	croak ("%s, at character offset %d (%s)",	1411	croak ("%s, at character offset %d [\"%s\"]",
902	dec.err,	1412	dec.err,
903	(int)offset,	1413	(int)offset,
904	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1414	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
905	}	1415	}
906		1416
907	sv = sv_2mortal (sv);	1417	sv = sv_2mortal (sv);
908		1418
909	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1419	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
910	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1420	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
911		1421
912	return sv;	1422	return sv;
913	}	1423	}
914		1424
		1425	/////////////////////////////////////////////////////////////////////////////
		1426	// XS interface functions
		1427
915	MODULE = JSON::XS PACKAGE = JSON::XS	1428	MODULE = JSON::XS PACKAGE = JSON::XS
916		1429
917	BOOT:	1430	BOOT:
918	{	1431	{
919	int i;	1432	int i;
920		1433
921	memset (decode_hexdigit, 0xff, 256);
922	for (i = 10; i--; )	1434	for (i = 0; i < 256; ++i)
923	decode_hexdigit ['0' + i] = i;	1435	decode_hexdigit [i] =
		1436	i >= '0' && i <= '9' ? i - '0'
		1437	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1438	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1439	: -1;
924		1440
925	for (i = 7; i--; )
926	{
927	decode_hexdigit ['a' + i] = 10 + i;
928	decode_hexdigit ['A' + i] = 10 + i;
929	}
930
931	json_stash = gv_stashpv ("JSON::XS", 1);	1441	json_stash = gv_stashpv ("JSON::XS" , 1);
		1442	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1443
		1444	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1445	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
932	}	1446	}
933		1447
934	PROTOTYPES: DISABLE	1448	PROTOTYPES: DISABLE
935		1449
936	SV *new (char *dummy)	1450	void CLONE (...)
937	CODE:	1451	CODE:
938	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1452	json_stash = 0;
		1453	json_boolean_stash = 0;
		1454
		1455	void new (char *klass)
		1456	PPCODE:
		1457	{
		1458	SV *pv = NEWSV (0, sizeof (JSON));
		1459	SvPOK_only (pv);
		1460	Zero (SvPVX (pv), 1, JSON);
		1461	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1462	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH)));
		1463	}
		1464
		1465	void ascii (JSON *self, int enable = 1)
		1466	ALIAS:
		1467	ascii = F_ASCII
		1468	latin1 = F_LATIN1
		1469	utf8 = F_UTF8
		1470	indent = F_INDENT
		1471	canonical = F_CANONICAL
		1472	space_before = F_SPACE_BEFORE
		1473	space_after = F_SPACE_AFTER
		1474	pretty = F_PRETTY
		1475	allow_nonref = F_ALLOW_NONREF
		1476	shrink = F_SHRINK
		1477	allow_blessed = F_ALLOW_BLESSED
		1478	convert_blessed = F_CONV_BLESSED
		1479	relaxed = F_RELAXED
		1480	PPCODE:
		1481	{
		1482	if (enable)
		1483	self->flags |= ix;
		1484	else
		1485	self->flags &= ~ix;
		1486
		1487	XPUSHs (ST (0));
		1488	}
		1489
		1490	void get_ascii (JSON *self)
		1491	ALIAS:
		1492	get_ascii = F_ASCII
		1493	get_latin1 = F_LATIN1
		1494	get_utf8 = F_UTF8
		1495	get_indent = F_INDENT
		1496	get_canonical = F_CANONICAL
		1497	get_space_before = F_SPACE_BEFORE
		1498	get_space_after = F_SPACE_AFTER
		1499	get_pretty = F_PRETTY
		1500	get_allow_nonref = F_ALLOW_NONREF
		1501	get_shrink = F_SHRINK
		1502	get_allow_blessed = F_ALLOW_BLESSED
		1503	get_convert_blessed = F_CONV_BLESSED
		1504	get_relaxed = F_RELAXED
		1505	PPCODE:
		1506	XPUSHs (boolSV (self->flags & ix));
		1507
		1508	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1509	PPCODE:
		1510	{
		1511	UV log2 = 0;
		1512
		1513	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1514
		1515	while ((1UL << log2) < max_depth)
		1516	++log2;
		1517
		1518	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1519
		1520	XPUSHs (ST (0));
		1521	}
		1522
		1523	int get_max_depth (JSON *self)
		1524	CODE:
		1525	RETVAL = DEC_DEPTH (self->flags);
939	OUTPUT:	1526	OUTPUT:
940	RETVAL	1527	RETVAL
941		1528
942	SV *ascii (SV *self, int enable = 1)	1529	void max_size (JSON *self, UV max_size = 0)
943	ALIAS:	1530	PPCODE:
944	ascii = F_ASCII	1531	{
945	utf8 = F_UTF8	1532	UV log2 = 0;
946	indent = F_INDENT	1533
947	canonical = F_CANONICAL	1534	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
948	space_before = F_SPACE_BEFORE	1535	if (max_size == 1) max_size = 2;
949	space_after = F_SPACE_AFTER	1536
950	pretty = F_PRETTY	1537	while ((1UL << log2) < max_size)
951	allow_nonref = F_ALLOW_NONREF	1538	++log2;
952	shrink = F_SHRINK	1539
		1540	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1541
		1542	XPUSHs (ST (0));
		1543	}
		1544
		1545	int get_max_size (JSON *self)
953	CODE:	1546	CODE:
954	{	1547	RETVAL = DEC_SIZE (self->flags);
955	UV *uv = SvJSON (self);
956	if (enable)
957	*uv |= ix;
958	else
959	*uv &= ~ix;
960
961	RETVAL = newSVsv (self);
962	}
963	OUTPUT:	1548	OUTPUT:
964	RETVAL	1549	RETVAL
965		1550
966	void encode (SV *self, SV *scalar)	1551	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
967	PPCODE:	1552	PPCODE:
968	XPUSHs (encode_json (scalar, *SvJSON (self)));	1553	{
		1554	SvREFCNT_dec (self->cb_object);
		1555	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
969		1556
970	void decode (SV *self, SV *jsonstr)	1557	XPUSHs (ST (0));
		1558	}
		1559
		1560	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
971	PPCODE:	1561	PPCODE:
		1562	{
		1563	if (!self->cb_sk_object)
		1564	self->cb_sk_object = newHV ();
		1565
		1566	if (SvOK (cb))
		1567	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1568	else
		1569	{
		1570	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1571
		1572	if (!HvKEYS (self->cb_sk_object))
		1573	{
		1574	SvREFCNT_dec (self->cb_sk_object);
		1575	self->cb_sk_object = 0;
		1576	}
		1577	}
		1578
		1579	XPUSHs (ST (0));
		1580	}
		1581
		1582	void encode (JSON *self, SV *scalar)
		1583	PPCODE:
		1584	XPUSHs (encode_json (scalar, self));
		1585
		1586	void decode (JSON *self, SV *jsonstr)
		1587	PPCODE:
972	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1588	XPUSHs (decode_json (jsonstr, self, 0));
		1589
		1590	void decode_prefix (JSON *self, SV *jsonstr)
		1591	PPCODE:
		1592	{
		1593	UV offset;
		1594	EXTEND (SP, 2);
		1595	PUSHs (decode_json (jsonstr, self, &offset));
		1596	PUSHs (sv_2mortal (newSVuv (offset)));
		1597	}
		1598
		1599	void DESTROY (JSON *self)
		1600	CODE:
		1601	SvREFCNT_dec (self->cb_sk_object);
		1602	SvREFCNT_dec (self->cb_object);
973		1603
974	PROTOTYPES: ENABLE	1604	PROTOTYPES: ENABLE
975		1605
976	void to_json (SV *scalar)	1606	void to_json (SV *scalar)
977	PPCODE:	1607	PPCODE:
		1608	{
		1609	JSON json = { F_DEFAULT | F_UTF8 };
978	XPUSHs (encode_json (scalar, F_UTF8));	1610	XPUSHs (encode_json (scalar, &json));
		1611	}
979		1612
980	void from_json (SV *jsonstr)	1613	void from_json (SV *jsonstr)
981	PPCODE:	1614	PPCODE:
		1615	{
		1616	JSON json = { F_DEFAULT | F_UTF8 };
982	XPUSHs (decode_json (jsonstr, F_UTF8));	1617	XPUSHs (decode_json (jsonstr, &json, 0));
		1618	}
983		1619

Diff Legend

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