ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.17 by root, Sun Mar 25 02:46:41 2007 UTC vs.
Revision 1.114 by root, Wed Aug 1 19:04:41 2012 UTC

1	#include "EXTERN.h"	1	#include "EXTERN.h"
2	#include "perl.h"	2	#include "perl.h"
3	#include "XSUB.h"	3	#include "XSUB.h"
4		4
5	#include "assert.h"	5	#include <assert.h>
6	#include "string.h"	6	#include <string.h>
7	#include "stdlib.h"	7	#include <stdlib.h>
		8	#include <stdio.h>
		9	#include <limits.h>
		10	#include <float.h>
8		11
		12	#if defined(__BORLANDC__) || defined(_MSC_VER)
		13	# define snprintf _snprintf // C compilers have this in stdio.h
		14	#endif
		15
		16	// some old perls do not have this, try to make it work, no
		17	// guarantees, though. if it breaks, you get to keep the pieces.
		18	#ifndef UTF8_MAXBYTES
		19	# define UTF8_MAXBYTES 13
		20	#endif
		21
		22	// three extra for rounding, sign, and end of string
		23	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
		24
9	#define F_ASCII 0x00000001	25	#define F_ASCII 0x00000001UL
		26	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	27	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	28	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	29	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	30	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	31	#define F_SPACE_AFTER 0x00000040UL
15	#define F_ALLOW_NONREF 0x00000080	32	#define F_ALLOW_NONREF 0x00000100UL
16	#define F_SHRINK 0x00000100	33	#define F_SHRINK 0x00000200UL
17		34	#define F_ALLOW_BLESSED 0x00000400UL
18	// F_SKIPINVALID?	35	#define F_CONV_BLESSED 0x00000800UL
19	// F_EXECCODEREF?	36	#define F_RELAXED 0x00001000UL
20	// F_SELFCONVERT? <=> { &__class__ => }	37	#define F_ALLOW_UNKNOWN 0x00002000UL
		38	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
21		39
22	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	40	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
23	#define F_DEFAULT 0
24		41
25	#define INIT_SIZE 32 // initial scalar size to be allocated	42	#define INIT_SIZE 32 // initial scalar size to be allocated
26	#define INDENT_STEP 3 // spaces per indentation level	43	#define INDENT_STEP 3 // spaces per indentation level
27		44
28	#define UTF8_MAX_LEN 11 // for perls UTF-X: max. number of octets per character
29	#define SHORT_STRING_LEN 512 // special-case strings of up to this size	45	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
		46
		47	#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
30		48
31	#define SB do {	49	#define SB do {
32	#define SE } while (0)	50	#define SE } while (0)
33		51
		52	#if __GNUC__ >= 3
		53	# define expect(expr,value) __builtin_expect ((expr), (value))
		54	# define INLINE static inline
		55	#else
		56	# define expect(expr,value) (expr)
		57	# define INLINE static
		58	#endif
		59
		60	#define expect_false(expr) expect ((expr) != 0, 0)
		61	#define expect_true(expr) expect ((expr) != 0, 1)
		62
		63	#define IN_RANGE_INC(type,val,beg,end) \
		64	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		65	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		66
		67	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
		68
		69	#ifdef USE_ITHREADS
		70	# define JSON_SLOW 1
		71	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		72	#else
		73	# define JSON_SLOW 0
		74	# define JSON_STASH json_stash
		75	#endif
		76
34	static HV *json_stash; // JSON::XS::	77	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		78	static SV *json_true, *json_false;
		79
		80	enum {
		81	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		82	INCR_M_STR, // inside string
		83	INCR_M_BS, // inside backslash
		84	INCR_M_C0, // inside comment in initial whitespace sequence
		85	INCR_M_C1, // inside comment in other places
		86	INCR_M_JSON // outside anything, count nesting
		87	};
		88
		89	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		90
		91	typedef struct {
		92	U32 flags;
		93	U32 max_depth;
		94	STRLEN max_size;
		95
		96	SV *cb_object;
		97	HV *cb_sk_object;
		98
		99	// for the incremental parser
		100	SV *incr_text; // the source text so far
		101	STRLEN incr_pos; // the current offset into the text
		102	int incr_nest; // {[]}-nesting level
		103	unsigned char incr_mode;
		104	} JSON;
		105
		106	INLINE void
		107	json_init (JSON *json)
		108	{
		109	Zero (json, 1, JSON);
		110	json->max_depth = 512;
		111	}
35		112
36	/////////////////////////////////////////////////////////////////////////////	113	/////////////////////////////////////////////////////////////////////////////
37	// utility functions	114	// utility functions
38		115
39	static UV *	116	INLINE SV *
40	SvJSON (SV *sv)	117	get_bool (const char *name)
41	{	118	{
42	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))	119	SV *sv = get_sv (name, 1);
43	croak ("object is not of type JSON::XS");
44		120
45	return &SvUVX (SvRV (sv));	121	SvREADONLY_on (sv);
46	}	122	SvREADONLY_on (SvRV (sv));
47		123
48	static void	124	return sv;
		125	}
		126
		127	INLINE void
49	shrink (SV *sv)	128	shrink (SV *sv)
50	{	129	{
51	sv_utf8_downgrade (sv, 1);	130	sv_utf8_downgrade (sv, 1);
		131
52	if (SvLEN (sv) > SvCUR (sv) + 1)	132	if (SvLEN (sv) > SvCUR (sv) + 1)
53	{	133	{
54	#ifdef SvPV_shrink_to_cur	134	#ifdef SvPV_shrink_to_cur
55	SvPV_shrink_to_cur (sv);	135	SvPV_shrink_to_cur (sv);
56	#elif defined (SvPV_renew)	136	#elif defined (SvPV_renew)
…		…
62	// decode an utf-8 character and return it, or (UV)-1 in	142	// decode an utf-8 character and return it, or (UV)-1 in
63	// case of an error.	143	// case of an error.
64	// we special-case "safe" characters from U+80 .. U+7FF,	144	// we special-case "safe" characters from U+80 .. U+7FF,
65	// but use the very good perl function to parse anything else.	145	// but use the very good perl function to parse anything else.
66	// note that we never call this function for a ascii codepoints	146	// note that we never call this function for a ascii codepoints
67	static UV	147	INLINE UV
68	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)	148	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
69	{	149	{
70	if (s[0] > 0xdf || s[0] < 0xc2)	150	if (expect_true (len >= 2
71	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	151	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
72	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	152	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
73	{	153	{
74	*clen = 2;	154	*clen = 2;
75	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);	155	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
76	}	156	}
77	else	157	else
78	return (UV)-1;	158	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
79	}	159	}
80		160
		161	// likewise for encoding, also never called for ascii codepoints
		162	// this function takes advantage of this fact, although current gccs
		163	// seem to optimise the check for >= 0x80 away anyways
		164	INLINE unsigned char *
		165	encode_utf8 (unsigned char *s, UV ch)
		166	{
		167	if (expect_false (ch < 0x000080))
		168	*s++ = ch;
		169	else if (expect_true (ch < 0x000800))
		170	*s++ = 0xc0 | ( ch >> 6),
		171	*s++ = 0x80 | ( ch & 0x3f);
		172	else if ( ch < 0x010000)
		173	*s++ = 0xe0 | ( ch >> 12),
		174	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		175	*s++ = 0x80 | ( ch & 0x3f);
		176	else if ( ch < 0x110000)
		177	*s++ = 0xf0 | ( ch >> 18),
		178	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		179	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		180	*s++ = 0x80 | ( ch & 0x3f);
		181
		182	return s;
		183	}
		184
		185	// convert offset pointer to character index, sv must be string
		186	static STRLEN
		187	ptr_to_index (SV *sv, char *offset)
		188	{
		189	return SvUTF8 (sv)
		190	? utf8_distance (offset, SvPVX (sv))
		191	: offset - SvPVX (sv);
		192	}
		193
		194	/////////////////////////////////////////////////////////////////////////////
		195	// fp hell
		196
		197	// scan a group of digits, and a trailing exponent
		198	static void
		199	json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
		200	{
		201	UV uaccum = 0;
		202	int eaccum = 0;
		203
		204	// if we recurse too deep, skip all remaining digits
		205	// to avoid a stack overflow attack
		206	if (expect_false (--maxdepth <= 0))
		207	while (((U8)*s - '0') < 10)
		208	++s;
		209
		210	for (;;)
		211	{
		212	U8 dig = (U8)*s - '0';
		213
		214	if (expect_false (dig >= 10))
		215	{
		216	if (dig == (U8)((U8)'.' - (U8)'0'))
		217	{
		218	++s;
		219	json_atof_scan1 (s, accum, expo, 1, maxdepth);
		220	}
		221	else if ((dig | ' ') == 'e' - '0')
		222	{
		223	int exp2 = 0;
		224	int neg = 0;
		225
		226	++s;
		227
		228	if (*s == '-')
		229	{
		230	++s;
		231	neg = 1;
		232	}
		233	else if (*s == '+')
		234	++s;
		235
		236	while ((dig = (U8)*s - '0') < 10)
		237	exp2 = exp2 * 10 + *s++ - '0';
		238
		239	*expo += neg ? -exp2 : exp2;
		240	}
		241
		242	break;
		243	}
		244
		245	++s;
		246
		247	uaccum = uaccum * 10 + dig;
		248	++eaccum;
		249
		250	// if we have too many digits, then recurse for more
		251	// we actually do this for rather few digits
		252	if (uaccum >= (UV_MAX - 9) / 10)
		253	{
		254	if (postdp) *expo -= eaccum;
		255	json_atof_scan1 (s, accum, expo, postdp, maxdepth);
		256	if (postdp) *expo += eaccum;
		257
		258	break;
		259	}
		260	}
		261
		262	// this relies greatly on the quality of the pow ()
		263	// implementation of the platform, but a good
		264	// implementation is hard to beat.
		265	// (IEEE 754 conformant ones are required to be exact)
		266	if (postdp) *expo -= eaccum;
		267	*accum += uaccum * Perl_pow (10., *expo);
		268	*expo += eaccum;
		269	}
		270
		271	static NV
		272	json_atof (const char *s)
		273	{
		274	NV accum = 0.;
		275	int expo = 0;
		276	int neg = 0;
		277
		278	if (*s == '-')
		279	{
		280	++s;
		281	neg = 1;
		282	}
		283
		284	// a recursion depth of ten gives us >>500 bits
		285	json_atof_scan1 (s, &accum, &expo, 0, 10);
		286
		287	return neg ? -accum : accum;
		288	}
81	/////////////////////////////////////////////////////////////////////////////	289	/////////////////////////////////////////////////////////////////////////////
82	// encoder	290	// encoder
83		291
84	// structure used for encoding JSON	292	// structure used for encoding JSON
85	typedef struct	293	typedef struct
86	{	294	{
87	char *cur; // SvPVX (sv) + current output position	295	char *cur; // SvPVX (sv) + current output position
88	char *end; // SvEND (sv)	296	char *end; // SvEND (sv)
89	SV *sv; // result scalar	297	SV *sv; // result scalar
90	UV flags; // F_*	298	JSON json;
91	int indent; // indentation level	299	U32 indent; // indentation level
92	int max_depth; // max. recursion level	300	UV limit; // escape character values >= this value when encoding
93	} enc_t;	301	} enc_t;
94		302
95	static void	303	INLINE void
96	need (enc_t *enc, STRLEN len)	304	need (enc_t *enc, STRLEN len)
97	{	305	{
98	if (enc->cur + len >= enc->end)	306	if (expect_false (enc->cur + len >= enc->end))
99	{	307	{
100	STRLEN cur = enc->cur - SvPVX (enc->sv);	308	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
101	SvGROW (enc->sv, cur + len + 1);	309	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
102	enc->cur = SvPVX (enc->sv) + cur;	310	enc->cur = SvPVX (enc->sv) + cur;
103	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	311	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
104	}	312	}
105	}	313	}
106		314
107	static void	315	INLINE void
108	encode_ch (enc_t *enc, char ch)	316	encode_ch (enc_t *enc, char ch)
109	{	317	{
110	need (enc, 1);	318	need (enc, 1);
111	*enc->cur++ = ch;	319	*enc->cur++ = ch;
112	}	320	}
…		…
120		328
121	while (str < end)	329	while (str < end)
122	{	330	{
123	unsigned char ch = *(unsigned char *)str;	331	unsigned char ch = *(unsigned char *)str;
124		332
125	if (ch >= 0x20 && ch < 0x80) // most common case	333	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
126	{	334	{
127	if (ch == '"') // but with slow exceptions	335	if (expect_false (ch == '"')) // but with slow exceptions
128	{	336	{
129	need (enc, len += 1);	337	need (enc, len += 1);
130	*enc->cur++ = '\\';	338	*enc->cur++ = '\\';
131	*enc->cur++ = '"';	339	*enc->cur++ = '"';
132	}	340	}
133	else if (ch == '\\')	341	else if (expect_false (ch == '\\'))
134	{	342	{
135	need (enc, len += 1);	343	need (enc, len += 1);
136	*enc->cur++ = '\\';	344	*enc->cur++ = '\\';
137	*enc->cur++ = '\\';	345	*enc->cur++ = '\\';
138	}	346	}
…		…
156	STRLEN clen;	364	STRLEN clen;
157	UV uch;	365	UV uch;
158		366
159	if (is_utf8)	367	if (is_utf8)
160	{	368	{
161	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
162	uch = decode_utf8 (str, end - str, &clen);	369	uch = decode_utf8 (str, end - str, &clen);
163	if (clen == (STRLEN)-1)	370	if (clen == (STRLEN)-1)
164	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	371	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
165	}	372	}
166	else	373	else
167	{	374	{
168	uch = ch;	375	uch = ch;
169	clen = 1;	376	clen = 1;
170	}	377	}
171		378
172	if (uch > 0x10FFFFUL)	379	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
173	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
174
175	if (uch < 0x80 || enc->flags & F_ASCII)
176	{	380	{
177	if (uch > 0xFFFFUL)	381	if (uch >= 0x10000UL)
178	{	382	{
		383	if (uch >= 0x110000UL)
		384	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		385
179	need (enc, len += 11);	386	need (enc, len += 11);
180	sprintf (enc->cur, "\\u%04x\\u%04x",	387	sprintf (enc->cur, "\\u%04x\\u%04x",
181	(int)((uch - 0x10000) / 0x400 + 0xD800),	388	(int)((uch - 0x10000) / 0x400 + 0xD800),
182	(int)((uch - 0x10000) % 0x400 + 0xDC00));	389	(int)((uch - 0x10000) % 0x400 + 0xDC00));
183	enc->cur += 12;	390	enc->cur += 12;
184	}	391	}
185	else	392	else
186	{	393	{
187	static char hexdigit [16] = "0123456789abcdef";
188	need (enc, len += 5);	394	need (enc, len += 5);
189	*enc->cur++ = '\\';	395	*enc->cur++ = '\\';
190	*enc->cur++ = 'u';	396	*enc->cur++ = 'u';
191	*enc->cur++ = hexdigit [ uch >> 12 ];	397	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
192	*enc->cur++ = hexdigit [(uch >> 8) & 15];	398	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
193	*enc->cur++ = hexdigit [(uch >> 4) & 15];	399	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
194	*enc->cur++ = hexdigit [(uch >> 0) & 15];	400	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
195	}	401	}
196		402
		403	str += clen;
		404	}
		405	else if (enc->json.flags & F_LATIN1)
		406	{
		407	*enc->cur++ = uch;
197	str += clen;	408	str += clen;
198	}	409	}
199	else if (is_utf8)	410	else if (is_utf8)
200	{	411	{
201	need (enc, len += clen);	412	need (enc, len += clen);
…		…
205	}	416	}
206	while (--clen);	417	while (--clen);
207	}	418	}
208	else	419	else
209	{	420	{
210	need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed	421	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
211	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	422	enc->cur = encode_utf8 (enc->cur, uch);
212	++str;	423	++str;
213	}	424	}
214	}	425	}
215	}	426	}
216	}	427	}
217		428
218	--len;	429	--len;
219	}	430	}
220	}	431	}
221		432
222	static void	433	INLINE void
223	encode_indent (enc_t *enc)	434	encode_indent (enc_t *enc)
224	{	435	{
225	if (enc->flags & F_INDENT)	436	if (enc->json.flags & F_INDENT)
226	{	437	{
227	int spaces = enc->indent * INDENT_STEP;	438	int spaces = enc->indent * INDENT_STEP;
228		439
229	need (enc, spaces);	440	need (enc, spaces);
230	memset (enc->cur, ' ', spaces);	441	memset (enc->cur, ' ', spaces);
231	enc->cur += spaces;	442	enc->cur += spaces;
232	}	443	}
233	}	444	}
234		445
235	static void	446	INLINE void
236	encode_space (enc_t *enc)	447	encode_space (enc_t *enc)
237	{	448	{
238	need (enc, 1);	449	need (enc, 1);
239	encode_ch (enc, ' ');	450	encode_ch (enc, ' ');
240	}	451	}
241		452
242	static void	453	INLINE void
243	encode_nl (enc_t *enc)	454	encode_nl (enc_t *enc)
244	{	455	{
245	if (enc->flags & F_INDENT)	456	if (enc->json.flags & F_INDENT)
246	{	457	{
247	need (enc, 1);	458	need (enc, 1);
248	encode_ch (enc, '\n');	459	encode_ch (enc, '\n');
249	}	460	}
250	}	461	}
251		462
252	static void	463	INLINE void
253	encode_comma (enc_t *enc)	464	encode_comma (enc_t *enc)
254	{	465	{
255	encode_ch (enc, ',');	466	encode_ch (enc, ',');
256		467
257	if (enc->flags & F_INDENT)	468	if (enc->json.flags & F_INDENT)
258	encode_nl (enc);	469	encode_nl (enc);
259	else if (enc->flags & F_SPACE_AFTER)	470	else if (enc->json.flags & F_SPACE_AFTER)
260	encode_space (enc);	471	encode_space (enc);
261	}	472	}
262		473
263	static void encode_sv (enc_t *enc, SV *sv);	474	static void encode_sv (enc_t *enc, SV *sv);
264		475
265	static void	476	static void
266	encode_av (enc_t *enc, AV *av)	477	encode_av (enc_t *enc, AV *av)
267	{	478	{
268	int i, len = av_len (av);	479	int i, len = av_len (av);
269		480
270	encode_ch (enc, '['); encode_nl (enc);	481	if (enc->indent >= enc->json.max_depth)
271	++enc->indent;	482	croak (ERR_NESTING_EXCEEDED);
272		483
		484	encode_ch (enc, '[');
		485
		486	if (len >= 0)
		487	{
		488	encode_nl (enc); ++enc->indent;
		489
273	for (i = 0; i <= len; ++i)	490	for (i = 0; i <= len; ++i)
274	{	491	{
		492	SV **svp = av_fetch (av, i, 0);
		493
275	encode_indent (enc);	494	encode_indent (enc);
276	encode_sv (enc, *av_fetch (av, i, 0));
277		495
		496	if (svp)
		497	encode_sv (enc, *svp);
		498	else
		499	encode_str (enc, "null", 4, 0);
		500
278	if (i < len)	501	if (i < len)
279	encode_comma (enc);	502	encode_comma (enc);
280	}	503	}
281		504
		505	encode_nl (enc); --enc->indent; encode_indent (enc);
		506	}
		507
282	encode_nl (enc);	508	encode_ch (enc, ']');
283
284	--enc->indent;
285	encode_indent (enc); encode_ch (enc, ']');
286	}	509	}
287		510
288	static void	511	static void
289	encode_he (enc_t *enc, HE *he)	512	encode_hk (enc_t *enc, HE *he)
290	{	513	{
291	encode_ch (enc, '"');	514	encode_ch (enc, '"');
292		515
293	if (HeKLEN (he) == HEf_SVKEY)	516	if (HeKLEN (he) == HEf_SVKEY)
294	{	517	{
…		…
304	else	527	else
305	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	528	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
306		529
307	encode_ch (enc, '"');	530	encode_ch (enc, '"');
308		531
309	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	532	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
310	encode_ch (enc, ':');	533	encode_ch (enc, ':');
311	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	534	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
312	encode_sv (enc, HeVAL (he));
313	}	535	}
314		536
315	// compare hash entries, used when all keys are bytestrings	537	// compare hash entries, used when all keys are bytestrings
316	static int	538	static int
317	he_cmp_fast (const void *a_, const void *b_)	539	he_cmp_fast (const void *a_, const void *b_)
…		…
322	HE *b = *(HE **)b_;	544	HE *b = *(HE **)b_;
323		545
324	STRLEN la = HeKLEN (a);	546	STRLEN la = HeKLEN (a);
325	STRLEN lb = HeKLEN (b);	547	STRLEN lb = HeKLEN (b);
326		548
327	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	549	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
328	cmp = la - lb;	550	cmp = lb - la;
329		551
330	return cmp;	552	return cmp;
331	}	553	}
332		554
333	// compare hash entries, used when some keys are sv's or utf-x	555	// compare hash entries, used when some keys are sv's or utf-x
334	static int	556	static int
335	he_cmp_slow (const void *a, const void *b)	557	he_cmp_slow (const void *a, const void *b)
336	{	558	{
337	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	559	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
338	}	560	}
339		561
340	static void	562	static void
341	encode_hv (enc_t *enc, HV *hv)	563	encode_hv (enc_t *enc, HV *hv)
342	{	564	{
343	int count, i;	565	HE *he;
344		566
345	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	567	if (enc->indent >= enc->json.max_depth)
		568	croak (ERR_NESTING_EXCEEDED);
346		569
347	if ((count = hv_iterinit (hv)))	570	encode_ch (enc, '{');
348	{	571
349	// for canonical output we have to sort by keys first	572	// for canonical output we have to sort by keys first
350	// actually, this is mostly due to the stupid so-called	573	// actually, this is mostly due to the stupid so-called
351	// security workaround added somewhere in 5.8.x.	574	// security workaround added somewhere in 5.8.x
352	// that randomises hash orderings	575	// that randomises hash orderings
353	if (enc->flags & F_CANONICAL)	576	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
		577	{
		578	int count = hv_iterinit (hv);
		579
		580	if (SvMAGICAL (hv))
354	{	581	{
		582	// need to count by iterating. could improve by dynamically building the vector below
		583	// but I don't care for the speed of this special case.
		584	// note also that we will run into undefined behaviour when the two iterations
		585	// do not result in the same count, something I might care for in some later release.
		586
		587	count = 0;
		588	while (hv_iternext (hv))
		589	++count;
		590
		591	hv_iterinit (hv);
		592	}
		593
		594	if (count)
		595	{
		596	int i, fast = 1;
		597	#if defined(__BORLANDC__) || defined(_MSC_VER)
		598	HE **hes = _alloca (count * sizeof (HE));
		599	#else
355	HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode	600	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
356	int fast = 1;	601	#endif
357		602
358	i = 0;	603	i = 0;
359	while ((he = hv_iternext (hv)))	604	while ((he = hv_iternext (hv)))
360	{	605	{
361	hes [i++] = he;	606	hes [i++] = he;
…		…
383		628
384	FREETMPS;	629	FREETMPS;
385	LEAVE;	630	LEAVE;
386	}	631	}
387		632
388	for (i = 0; i < count; ++i)	633	encode_nl (enc); ++enc->indent;
		634
		635	while (count--)
389	{	636	{
390	encode_indent (enc);	637	encode_indent (enc);
		638	he = hes [count];
391	encode_he (enc, hes [i]);	639	encode_hk (enc, he);
		640	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
392		641
393	if (i < count - 1)	642	if (count)
394	encode_comma (enc);	643	encode_comma (enc);
395	}	644	}
396		645
		646	encode_nl (enc); --enc->indent; encode_indent (enc);
		647	}
		648	}
		649	else
		650	{
		651	if (hv_iterinit (hv) || SvMAGICAL (hv))
		652	if ((he = hv_iternext (hv)))
		653	{
		654	encode_nl (enc); ++enc->indent;
		655
		656	for (;;)
		657	{
397	encode_nl (enc);	658	encode_indent (enc);
		659	encode_hk (enc, he);
		660	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		661
		662	if (!(he = hv_iternext (hv)))
		663	break;
		664
		665	encode_comma (enc);
		666	}
		667
		668	encode_nl (enc); --enc->indent; encode_indent (enc);
		669	}
		670	}
		671
		672	encode_ch (enc, '}');
		673	}
		674
		675	// encode objects, arrays and special \0=false and \1=true values.
		676	static void
		677	encode_rv (enc_t *enc, SV *sv)
		678	{
		679	svtype svt;
		680
		681	SvGETMAGIC (sv);
		682	svt = SvTYPE (sv);
		683
		684	if (expect_false (SvOBJECT (sv)))
		685	{
		686	HV *stash = !JSON_SLOW || json_boolean_stash
		687	? json_boolean_stash
		688	: gv_stashpv ("JSON::XS::Boolean", 1);
		689
		690	if (SvSTASH (sv) == stash)
		691	{
		692	if (SvIV (sv))
		693	encode_str (enc, "true", 4, 0);
		694	else
		695	encode_str (enc, "false", 5, 0);
398	}	696	}
399	else	697	else
400	{	698	{
401	SV *sv;	699	#if 0
402	HE *he = hv_iternext (hv);	700	if (0 && sv_derived_from (rv, "JSON::Literal"))
403
404	for (;;)
405	{
406	encode_indent (enc);
407	encode_he (enc, he);
408
409	if (!(he = hv_iternext (hv)))
410	break;
411
412	encode_comma (enc);
413	}	701	{
		702	// not yet
		703	}
		704	#endif
		705	if (enc->json.flags & F_CONV_BLESSED)
		706	{
		707	// we re-bless the reference to get overload and other niceties right
		708	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
414		709
		710	if (to_json)
		711	{
		712	dSP;
		713
		714	ENTER; SAVETMPS; PUSHMARK (SP);
		715	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		716
		717	// calling with G_SCALAR ensures that we always get a 1 return value
		718	PUTBACK;
		719	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		720	SPAGAIN;
		721
		722	// catch this surprisingly common error
		723	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		724	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		725
		726	sv = POPs;
		727	PUTBACK;
		728
415	encode_nl (enc);	729	encode_sv (enc, sv);
		730
		731	FREETMPS; LEAVE;
		732	}
		733	else if (enc->json.flags & F_ALLOW_BLESSED)
		734	encode_str (enc, "null", 4, 0);
		735	else
		736	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		737	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		738	}
		739	else if (enc->json.flags & F_ALLOW_BLESSED)
		740	encode_str (enc, "null", 4, 0);
		741	else
		742	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		743	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
416	}	744	}
417	}	745	}
		746	else if (svt == SVt_PVHV)
		747	encode_hv (enc, (HV *)sv);
		748	else if (svt == SVt_PVAV)
		749	encode_av (enc, (AV *)sv);
		750	else if (svt < SVt_PVAV)
		751	{
		752	STRLEN len = 0;
		753	char *pv = svt ? SvPV (sv, len) : 0;
418		754
419	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	755	if (len == 1 && *pv == '1')
		756	encode_str (enc, "true", 4, 0);
		757	else if (len == 1 && *pv == '0')
		758	encode_str (enc, "false", 5, 0);
		759	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		760	encode_str (enc, "null", 4, 0);
		761	else
		762	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		763	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		764	}
		765	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		766	encode_str (enc, "null", 4, 0);
		767	else
		768	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		769	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
420	}	770	}
421		771
422	static void	772	static void
423	encode_sv (enc_t *enc, SV *sv)	773	encode_sv (enc_t *enc, SV *sv)
424	{	774	{
…		…
432	encode_str (enc, str, len, SvUTF8 (sv));	782	encode_str (enc, str, len, SvUTF8 (sv));
433	encode_ch (enc, '"');	783	encode_ch (enc, '"');
434	}	784	}
435	else if (SvNOKp (sv))	785	else if (SvNOKp (sv))
436	{	786	{
		787	// trust that perl will do the right thing w.r.t. JSON syntax.
437	need (enc, NV_DIG + 32);	788	need (enc, NV_DIG + 32);
438	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	789	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
439	enc->cur += strlen (enc->cur);	790	enc->cur += strlen (enc->cur);
440	}	791	}
441	else if (SvIOKp (sv))	792	else if (SvIOKp (sv))
442	{	793	{
		794	// we assume we can always read an IV as a UV and vice versa
		795	// we assume two's complement
		796	// we assume no aliasing issues in the union
		797	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		798	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		799	{
		800	// optimise the "small number case"
		801	// code will likely be branchless and use only a single multiplication
		802	// works for numbers up to 59074
		803	I32 i = SvIVX (sv);
		804	U32 u;
		805	char digit, nz = 0;
		806
443	need (enc, 64);	807	need (enc, 6);
		808
		809	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		810	u = i < 0 ? -i : i;
		811
		812	// convert to 4.28 fixed-point representation
		813	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		814
		815	// now output digit by digit, each time masking out the integer part
		816	// and multiplying by 5 while moving the decimal point one to the right,
		817	// resulting in a net multiplication by 10.
		818	// we always write the digit to memory but conditionally increment
		819	// the pointer, to enable the use of conditional move instructions.
		820	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		821	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		822	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		823	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		824	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		825	}
		826	else
		827	{
		828	// large integer, use the (rather slow) snprintf way.
		829	need (enc, IVUV_MAXCHARS);
444	enc->cur +=	830	enc->cur +=
445	SvIsUV(sv)	831	SvIsUV(sv)
446	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	832	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
447	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	833	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		834	}
448	}	835	}
449	else if (SvROK (sv))	836	else if (SvROK (sv))
450	{	837	encode_rv (enc, SvRV (sv));
451	SV *rv = SvRV (sv);	838	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
452
453	if (enc->indent >= enc->max_depth)
454	croak ("data structure too deep (hit recursion limit)");
455
456	switch (SvTYPE (rv))
457	{
458	case SVt_PVAV: encode_av (enc, (AV *)rv); break;
459	case SVt_PVHV: encode_hv (enc, (HV *)rv); break;
460
461	default:
462	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
463	SvPV_nolen (sv));
464	}
465	}
466	else if (!SvOK (sv))
467	encode_str (enc, "null", 4, 0);	839	encode_str (enc, "null", 4, 0);
468	else	840	else
469	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	841	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
470	SvPV_nolen (sv), SvFLAGS (sv));	842	SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
471	}	843	}
472		844
473	static SV *	845	static SV *
474	encode_json (SV *scalar, UV flags)	846	encode_json (SV *scalar, JSON *json)
475	{	847	{
		848	enc_t enc;
		849
476	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	850	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
477	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	851	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
478		852
479	enc_t enc;	853	enc.json = *json;
480	enc.flags = flags;
481	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	854	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
482	enc.cur = SvPVX (enc.sv);	855	enc.cur = SvPVX (enc.sv);
483	enc.end = SvEND (enc.sv);	856	enc.end = SvEND (enc.sv);
484	enc.indent = 0;	857	enc.indent = 0;
485	enc.max_depth = 0x7fffffffUL;	858	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		859	: enc.json.flags & F_LATIN1 ? 0x000100UL
		860	: 0x110000UL;
486		861
487	SvPOK_only (enc.sv);	862	SvPOK_only (enc.sv);
488	encode_sv (&enc, scalar);	863	encode_sv (&enc, scalar);
		864	encode_nl (&enc);
489		865
		866	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		867	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		868
490	if (!(flags & (F_ASCII | F_UTF8)))	869	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
491	SvUTF8_on (enc.sv);	870	SvUTF8_on (enc.sv);
492		871
493	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
494
495	if (enc.flags & F_SHRINK)	872	if (enc.json.flags & F_SHRINK)
496	shrink (enc.sv);	873	shrink (enc.sv);
497		874
498	return enc.sv;	875	return enc.sv;
499	}	876	}
500		877
…		…
505	typedef struct	882	typedef struct
506	{	883	{
507	char *cur; // current parser pointer	884	char *cur; // current parser pointer
508	char *end; // end of input string	885	char *end; // end of input string
509	const char *err; // parse error, if != 0	886	const char *err; // parse error, if != 0
510	UV flags; // F_*	887	JSON json;
		888	U32 depth; // recursion depth
		889	U32 maxdepth; // recursion depth limit
511	} dec_t;	890	} dec_t;
512		891
513	static void	892	INLINE void
		893	decode_comment (dec_t *dec)
		894	{
		895	// only '#'-style comments allowed a.t.m.
		896
		897	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		898	++dec->cur;
		899	}
		900
		901	INLINE void
514	decode_ws (dec_t *dec)	902	decode_ws (dec_t *dec)
515	{	903	{
516	for (;;)	904	for (;;)
517	{	905	{
518	char ch = *dec->cur;	906	char ch = *dec->cur;
519		907
520	if (ch > 0x20	908	if (ch > 0x20)
		909	{
		910	if (expect_false (ch == '#'))
		911	{
		912	if (dec->json.flags & F_RELAXED)
		913	decode_comment (dec);
		914	else
		915	break;
		916	}
		917	else
		918	break;
		919	}
521	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	920	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
522	break;	921	break; // parse error, but let higher level handle it, gives better error messages
523		922
524	++dec->cur;	923	++dec->cur;
525	}	924	}
526	}	925	}
527		926
528	#define ERR(reason) SB dec->err = reason; goto fail; SE	927	#define ERR(reason) SB dec->err = reason; goto fail; SE
		928
529	#define EXPECT_CH(ch) SB \	929	#define EXPECT_CH(ch) SB \
530	if (*dec->cur != ch) \	930	if (*dec->cur != ch) \
531	ERR (# ch " expected"); \	931	ERR (# ch " expected"); \
532	++dec->cur; \	932	++dec->cur; \
533	SE	933	SE
534		934
		935	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
		936	#define DEC_DEC_DEPTH --dec->depth
		937
535	static SV *decode_sv (dec_t *dec);	938	static SV *decode_sv (dec_t *dec);
536		939
537	static signed char decode_hexdigit[256];	940	static signed char decode_hexdigit[256];
538		941
539	static UV	942	static UV
540	decode_4hex (dec_t *dec)	943	decode_4hex (dec_t *dec)
541	{	944	{
542	signed char d1, d2, d3, d4;	945	signed char d1, d2, d3, d4;
543	unsigned char *cur = (unsigned char *)dec->cur;	946	unsigned char *cur = (unsigned char *)dec->cur;
544		947
545	d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected");	948	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
546	d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected");	949	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
547	d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected");	950	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
548	d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected");	951	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
549		952
550	dec->cur += 4;	953	dec->cur += 4;
551		954
552	return ((UV)d1) << 12	955	return ((UV)d1) << 12
553	| ((UV)d2) << 8	956	| ((UV)d2) << 8
…		…
561	static SV *	964	static SV *
562	decode_str (dec_t *dec)	965	decode_str (dec_t *dec)
563	{	966	{
564	SV *sv = 0;	967	SV *sv = 0;
565	int utf8 = 0;	968	int utf8 = 0;
		969	char *dec_cur = dec->cur;
566		970
567	do	971	do
568	{	972	{
569	char buf [SHORT_STRING_LEN + UTF8_MAX_LEN];	973	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
570	char *cur = buf;	974	char *cur = buf;
571		975
572	do	976	do
573	{	977	{
574	unsigned char ch = *(unsigned char *)dec->cur++;	978	unsigned char ch = *(unsigned char *)dec_cur++;
575		979
576	if (ch == '"')	980	if (expect_false (ch == '"'))
577	{	981	{
578	--dec->cur;	982	--dec_cur;
579	break;	983	break;
580	}	984	}
581	else if (ch == '\\')	985	else if (expect_false (ch == '\\'))
582	{	986	{
583	switch (*dec->cur)	987	switch (*dec_cur)
584	{	988	{
585	case '\\':	989	case '\\':
586	case '/':	990	case '/':
587	case '"': *cur++ = *dec->cur++; break;	991	case '"': *cur++ = *dec_cur++; break;
588		992
589	case 'b': ++dec->cur; *cur++ = '\010'; break;	993	case 'b': ++dec_cur; *cur++ = '\010'; break;
590	case 't': ++dec->cur; *cur++ = '\011'; break;	994	case 't': ++dec_cur; *cur++ = '\011'; break;
591	case 'n': ++dec->cur; *cur++ = '\012'; break;	995	case 'n': ++dec_cur; *cur++ = '\012'; break;
592	case 'f': ++dec->cur; *cur++ = '\014'; break;	996	case 'f': ++dec_cur; *cur++ = '\014'; break;
593	case 'r': ++dec->cur; *cur++ = '\015'; break;	997	case 'r': ++dec_cur; *cur++ = '\015'; break;
594		998
595	case 'u':	999	case 'u':
596	{	1000	{
597	UV lo, hi;	1001	UV lo, hi;
598	++dec->cur;	1002	++dec_cur;
599		1003
		1004	dec->cur = dec_cur;
600	hi = decode_4hex (dec);	1005	hi = decode_4hex (dec);
		1006	dec_cur = dec->cur;
601	if (hi == (UV)-1)	1007	if (hi == (UV)-1)
602	goto fail;	1008	goto fail;
603		1009
604	// possibly a surrogate pair	1010	// possibly a surrogate pair
605	if (hi >= 0xd800)	1011	if (hi >= 0xd800)
606	if (hi < 0xdc00)	1012	if (hi < 0xdc00)
607	{	1013	{
608	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	1014	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
609	ERR ("missing low surrogate character in surrogate pair");	1015	ERR ("missing low surrogate character in surrogate pair");
610		1016
611	dec->cur += 2;	1017	dec_cur += 2;
612		1018
		1019	dec->cur = dec_cur;
613	lo = decode_4hex (dec);	1020	lo = decode_4hex (dec);
		1021	dec_cur = dec->cur;
614	if (lo == (UV)-1)	1022	if (lo == (UV)-1)
615	goto fail;	1023	goto fail;
616		1024
617	if (lo < 0xdc00 || lo >= 0xe000)	1025	if (lo < 0xdc00 || lo >= 0xe000)
618	ERR ("surrogate pair expected");	1026	ERR ("surrogate pair expected");
…		…
624		1032
625	if (hi >= 0x80)	1033	if (hi >= 0x80)
626	{	1034	{
627	utf8 = 1;	1035	utf8 = 1;
628		1036
629	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	1037	cur = encode_utf8 (cur, hi);
630	}	1038	}
631	else	1039	else
632	*cur++ = hi;	1040	*cur++ = hi;
633	}	1041	}
634	break;	1042	break;
635		1043
636	default:	1044	default:
637	--dec->cur;	1045	--dec_cur;
638	ERR ("illegal backslash escape sequence in string");	1046	ERR ("illegal backslash escape sequence in string");
639	}	1047	}
640	}	1048	}
641	else if (ch >= 0x20 && ch <= 0x7f)	1049	else if (expect_true (ch >= 0x20 && ch < 0x80))
642	*cur++ = ch;	1050	*cur++ = ch;
643	else if (ch >= 0x80)	1051	else if (ch >= 0x80)
644	{	1052	{
645	--dec->cur;
646
647	STRLEN clen;	1053	STRLEN clen;
		1054
		1055	--dec_cur;
		1056
648	UV uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	1057	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
649	if (clen == (STRLEN)-1)	1058	if (clen == (STRLEN)-1)
650	ERR ("malformed UTF-8 character in JSON string");	1059	ERR ("malformed UTF-8 character in JSON string");
651		1060
652	do	1061	do
653	{
654	*cur++ = *dec->cur++;	1062	*cur++ = *dec_cur++;
655	}
656	while (--clen);	1063	while (--clen);
657		1064
658	utf8 = 1;	1065	utf8 = 1;
659	}	1066	}
660	else if (!ch)
661	ERR ("unexpected end of string while parsing json string");
662	else	1067	else
		1068	{
		1069	--dec_cur;
		1070
		1071	if (!ch)
		1072	ERR ("unexpected end of string while parsing JSON string");
		1073	else
663	ERR ("invalid character encountered");	1074	ERR ("invalid character encountered while parsing JSON string");
664		1075	}
665	}	1076	}
666	while (cur < buf + SHORT_STRING_LEN);	1077	while (cur < buf + SHORT_STRING_LEN);
667		1078
		1079	{
668	STRLEN len = cur - buf;	1080	STRLEN len = cur - buf;
669		1081
670	if (sv)	1082	if (sv)
671	{	1083	{
672	SvGROW (sv, SvCUR (sv) + len + 1);	1084	STRLEN cur = SvCUR (sv);
		1085
		1086	if (SvLEN (sv) <= cur + len)
		1087	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1088
673	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	1089	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
674	SvCUR_set (sv, SvCUR (sv) + len);	1090	SvCUR_set (sv, SvCUR (sv) + len);
675	}	1091	}
676	else	1092	else
677	sv = newSVpvn (buf, len);	1093	sv = newSVpvn (buf, len);
678	}	1094	}
		1095	}
679	while (*dec->cur != '"');	1096	while (*dec_cur != '"');
680		1097
681	++dec->cur;	1098	++dec_cur;
682		1099
683	if (sv)	1100	if (sv)
684	{	1101	{
685	SvPOK_only (sv);	1102	SvPOK_only (sv);
686	*SvEND (sv) = 0;	1103	*SvEND (sv) = 0;
…		…
689	SvUTF8_on (sv);	1106	SvUTF8_on (sv);
690	}	1107	}
691	else	1108	else
692	sv = newSVpvn ("", 0);	1109	sv = newSVpvn ("", 0);
693		1110
		1111	dec->cur = dec_cur;
694	return sv;	1112	return sv;
695		1113
696	fail:	1114	fail:
		1115	dec->cur = dec_cur;
697	return 0;	1116	return 0;
698	}	1117	}
699		1118
700	static SV *	1119	static SV *
701	decode_num (dec_t *dec)	1120	decode_num (dec_t *dec)
…		…
759	is_nv = 1;	1178	is_nv = 1;
760	}	1179	}
761		1180
762	if (!is_nv)	1181	if (!is_nv)
763	{	1182	{
764	UV uv;	1183	int len = dec->cur - start;
765	int numtype = grok_number (start, dec->cur - start, &uv);	1184
766	if (numtype & IS_NUMBER_IN_UV)	1185	// special case the rather common 1..5-digit-int case
767	if (numtype & IS_NUMBER_NEG)	1186	if (*start == '-')
		1187	switch (len)
768	{	1188	{
769	if (uv < (UV)IV_MIN)	1189	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
770	return newSViv (-(IV)uv);	1190	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
		1191	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1192	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1193	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
771	}	1194	}
		1195	else
		1196	switch (len)
		1197	{
		1198	case 1: return newSViv ( start [0] - '0' * 1);
		1199	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1200	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1201	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1202	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1203	}
		1204
		1205	{
		1206	UV uv;
		1207	int numtype = grok_number (start, len, &uv);
		1208	if (numtype & IS_NUMBER_IN_UV)
		1209	if (numtype & IS_NUMBER_NEG)
		1210	{
		1211	if (uv < (UV)IV_MIN)
		1212	return newSViv (-(IV)uv);
		1213	}
772	else	1214	else
773	return newSVuv (uv);	1215	return newSVuv (uv);
774	}	1216	}
775		1217
		1218	len -= *start == '-' ? 1 : 0;
		1219
		1220	// does not fit into IV or UV, try NV
		1221	if (len <= NV_DIG)
		1222	// fits into NV without loss of precision
		1223	return newSVnv (json_atof (start));
		1224
		1225	// everything else fails, convert it to a string
		1226	return newSVpvn (start, dec->cur - start);
		1227	}
		1228
		1229	// loss of precision here
776	return newSVnv (Atof (start));	1230	return newSVnv (json_atof (start));
777		1231
778	fail:	1232	fail:
779	return 0;	1233	return 0;
780	}	1234	}
781		1235
782	static SV *	1236	static SV *
783	decode_av (dec_t *dec)	1237	decode_av (dec_t *dec)
784	{	1238	{
785	AV *av = newAV ();	1239	AV *av = newAV ();
786		1240
		1241	DEC_INC_DEPTH;
787	decode_ws (dec);	1242	decode_ws (dec);
		1243
788	if (*dec->cur == ']')	1244	if (*dec->cur == ']')
789	++dec->cur;	1245	++dec->cur;
790	else	1246	else
791	for (;;)	1247	for (;;)
792	{	1248	{
…		…
808		1264
809	if (*dec->cur != ',')	1265	if (*dec->cur != ',')
810	ERR (", or ] expected while parsing array");	1266	ERR (", or ] expected while parsing array");
811		1267
812	++dec->cur;	1268	++dec->cur;
		1269
		1270	decode_ws (dec);
		1271
		1272	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1273	{
		1274	++dec->cur;
		1275	break;
		1276	}
813	}	1277	}
814		1278
		1279	DEC_DEC_DEPTH;
815	return newRV_noinc ((SV *)av);	1280	return newRV_noinc ((SV *)av);
816		1281
817	fail:	1282	fail:
818	SvREFCNT_dec (av);	1283	SvREFCNT_dec (av);
		1284	DEC_DEC_DEPTH;
819	return 0;	1285	return 0;
820	}	1286	}
821		1287
822	static SV *	1288	static SV *
823	decode_hv (dec_t *dec)	1289	decode_hv (dec_t *dec)
824	{	1290	{
		1291	SV *sv;
825	HV *hv = newHV ();	1292	HV *hv = newHV ();
826		1293
		1294	DEC_INC_DEPTH;
827	decode_ws (dec);	1295	decode_ws (dec);
		1296
828	if (*dec->cur == '}')	1297	if (*dec->cur == '}')
829	++dec->cur;	1298	++dec->cur;
830	else	1299	else
831	for (;;)	1300	for (;;)
832	{	1301	{
		1302	EXPECT_CH ('"');
		1303
		1304	// heuristic: assume that
		1305	// a) decode_str + hv_store_ent are abysmally slow.
		1306	// b) most hash keys are short, simple ascii text.
		1307	// => try to "fast-match" such strings to avoid
		1308	// the overhead of decode_str + hv_store_ent.
		1309	{
833	SV *key, *value;	1310	SV *value;
		1311	char *p = dec->cur;
		1312	char *e = p + 24; // only try up to 24 bytes
834		1313
835	decode_ws (dec); EXPECT_CH ('"');	1314	for (;;)
836
837	key = decode_str (dec);
838	if (!key)
839	goto fail;
840
841	decode_ws (dec); EXPECT_CH (':');
842
843	value = decode_sv (dec);
844	if (!value)
845	{	1315	{
		1316	// the >= 0x80 is false on most architectures
		1317	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1318	{
		1319	// slow path, back up and use decode_str
		1320	SV *key = decode_str (dec);
		1321	if (!key)
		1322	goto fail;
		1323
		1324	decode_ws (dec); EXPECT_CH (':');
		1325
		1326	decode_ws (dec);
		1327	value = decode_sv (dec);
		1328	if (!value)
		1329	{
		1330	SvREFCNT_dec (key);
		1331	goto fail;
		1332	}
		1333
		1334	hv_store_ent (hv, key, value, 0);
846	SvREFCNT_dec (key);	1335	SvREFCNT_dec (key);
		1336
		1337	break;
		1338	}
		1339	else if (*p == '"')
		1340	{
		1341	// fast path, got a simple key
		1342	char *key = dec->cur;
		1343	int len = p - key;
		1344	dec->cur = p + 1;
		1345
		1346	decode_ws (dec); EXPECT_CH (':');
		1347
		1348	decode_ws (dec);
		1349	value = decode_sv (dec);
		1350	if (!value)
847	goto fail;	1351	goto fail;
		1352
		1353	hv_store (hv, key, len, value, 0);
		1354
		1355	break;
		1356	}
		1357
		1358	++p;
848	}	1359	}
849		1360	}
850	//TODO: optimise
851	hv_store_ent (hv, key, value, 0);
852		1361
853	decode_ws (dec);	1362	decode_ws (dec);
854		1363
855	if (*dec->cur == '}')	1364	if (*dec->cur == '}')
856	{	1365	{
…		…
860		1369
861	if (*dec->cur != ',')	1370	if (*dec->cur != ',')
862	ERR (", or } expected while parsing object/hash");	1371	ERR (", or } expected while parsing object/hash");
863		1372
864	++dec->cur;	1373	++dec->cur;
		1374
		1375	decode_ws (dec);
		1376
		1377	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1378	{
		1379	++dec->cur;
		1380	break;
		1381	}
865	}	1382	}
866		1383
		1384	DEC_DEC_DEPTH;
867	return newRV_noinc ((SV *)hv);	1385	sv = newRV_noinc ((SV *)hv);
		1386
		1387	// check filter callbacks
		1388	if (dec->json.flags & F_HOOK)
		1389	{
		1390	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1391	{
		1392	HE *cb, *he;
		1393
		1394	hv_iterinit (hv);
		1395	he = hv_iternext (hv);
		1396	hv_iterinit (hv);
		1397
		1398	// the next line creates a mortal sv each time its called.
		1399	// might want to optimise this for common cases.
		1400	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1401
		1402	if (cb)
		1403	{
		1404	dSP;
		1405	int count;
		1406
		1407	ENTER; SAVETMPS; PUSHMARK (SP);
		1408	XPUSHs (HeVAL (he));
		1409	sv_2mortal (sv);
		1410
		1411	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1412
		1413	if (count == 1)
		1414	{
		1415	sv = newSVsv (POPs);
		1416	FREETMPS; LEAVE;
		1417	return sv;
		1418	}
		1419
		1420	SvREFCNT_inc (sv);
		1421	FREETMPS; LEAVE;
		1422	}
		1423	}
		1424
		1425	if (dec->json.cb_object)
		1426	{
		1427	dSP;
		1428	int count;
		1429
		1430	ENTER; SAVETMPS; PUSHMARK (SP);
		1431	XPUSHs (sv_2mortal (sv));
		1432
		1433	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1434
		1435	if (count == 1)
		1436	{
		1437	sv = newSVsv (POPs);
		1438	FREETMPS; LEAVE;
		1439	return sv;
		1440	}
		1441
		1442	SvREFCNT_inc (sv);
		1443	FREETMPS; LEAVE;
		1444	}
		1445	}
		1446
		1447	return sv;
868		1448
869	fail:	1449	fail:
870	SvREFCNT_dec (hv);	1450	SvREFCNT_dec (hv);
		1451	DEC_DEC_DEPTH;
871	return 0;	1452	return 0;
872	}	1453	}
873		1454
874	static SV *	1455	static SV *
875	decode_sv (dec_t *dec)	1456	decode_sv (dec_t *dec)
876	{	1457	{
877	decode_ws (dec);	1458	// the beauty of JSON: you need exactly one character lookahead
		1459	// to parse everything.
878	switch (*dec->cur)	1460	switch (*dec->cur)
879	{	1461	{
880	case '"': ++dec->cur; return decode_str (dec);	1462	case '"': ++dec->cur; return decode_str (dec);
881	case '[': ++dec->cur; return decode_av (dec);	1463	case '[': ++dec->cur; return decode_av (dec);
882	case '{': ++dec->cur; return decode_hv (dec);	1464	case '{': ++dec->cur; return decode_hv (dec);
883		1465
884	case '-':	1466	case '-':
885	case '0': case '1': case '2': case '3': case '4':	1467	case '0': case '1': case '2': case '3': case '4':
886	case '5': case '6': case '7': case '8': case '9':	1468	case '5': case '6': case '7': case '8': case '9':
887	return decode_num (dec);	1469	return decode_num (dec);
888		1470
889	case 't':	1471	case 't':
890	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1472	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
891	{	1473	{
892	dec->cur += 4;	1474	dec->cur += 4;
		1475	#if JSON_SLOW
		1476	json_true = get_bool ("JSON::XS::true");
		1477	#endif
893	return newSViv (1);	1478	return newSVsv (json_true);
894	}	1479	}
895	else	1480	else
896	ERR ("'true' expected");	1481	ERR ("'true' expected");
897		1482
898	break;	1483	break;
899		1484
900	case 'f':	1485	case 'f':
901	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1486	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
902	{	1487	{
903	dec->cur += 5;	1488	dec->cur += 5;
		1489	#if JSON_SLOW
		1490	json_false = get_bool ("JSON::XS::false");
		1491	#endif
904	return newSViv (0);	1492	return newSVsv (json_false);
905	}	1493	}
906	else	1494	else
907	ERR ("'false' expected");	1495	ERR ("'false' expected");
908		1496
909	break;	1497	break;
…		…
918	ERR ("'null' expected");	1506	ERR ("'null' expected");
919		1507
920	break;	1508	break;
921		1509
922	default:	1510	default:
923	ERR ("malformed json string, neither array, object, number, string or atom");	1511	ERR ("malformed JSON string, neither array, object, number, string or atom");
924	break;	1512	break;
925	}	1513	}
926		1514
927	fail:	1515	fail:
928	return 0;	1516	return 0;
929	}	1517	}
930		1518
931	static SV *	1519	static SV *
932	decode_json (SV *string, UV flags)	1520	decode_json (SV *string, JSON *json, char **offset_return)
933	{	1521	{
		1522	dec_t dec;
934	SV *sv;	1523	SV *sv;
935		1524
936	if (flags & F_UTF8)	1525	/* work around bugs in 5.10 where manipulating magic values
		1526	* will perl ignore the magic in subsequent accesses.
		1527	* also make a copy of non-PV values, to get them into a clean
		1528	* state (SvPV should do that, but it's buggy, see below).
		1529	*/
		1530	/*SvGETMAGIC (string);*/
		1531	if (SvMAGICAL (string) || !SvPOK (string))
		1532	string = sv_2mortal (newSVsv (string));
		1533
		1534	SvUPGRADE (string, SVt_PV);
		1535
		1536	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1537	* assertion with -DDEBUGGING, although SvCUR is documented to
		1538	* return the xpv_cur field which certainly exists after upgrading.
		1539	* according to nicholas clark, calling SvPOK fixes this.
		1540	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1541	* and hope for the best.
		1542	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1543	* assertion business is seriously broken, try yet another workaround
		1544	* for the broken -DDEBUGGING.
		1545	*/
		1546	{
		1547	#ifdef DEBUGGING
		1548	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1549	#else
		1550	STRLEN offset = SvCUR (string);
		1551	#endif
		1552
		1553	if (offset > json->max_size && json->max_size)
		1554	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1555	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1556	}
		1557
		1558	if (DECODE_WANTS_OCTETS (json))
937	sv_utf8_downgrade (string, 0);	1559	sv_utf8_downgrade (string, 0);
938	else	1560	else
939	sv_utf8_upgrade (string);	1561	sv_utf8_upgrade (string);
940		1562
941	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1563	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
942		1564
943	dec_t dec;	1565	dec.json = *json;
944	dec.flags = flags;
945	dec.cur = SvPVX (string);	1566	dec.cur = SvPVX (string);
946	dec.end = SvEND (string);	1567	dec.end = SvEND (string);
947	dec.err = 0;	1568	dec.err = 0;
		1569	dec.depth = 0;
948		1570
		1571	if (dec.json.cb_object || dec.json.cb_sk_object)
		1572	dec.json.flags |= F_HOOK;
		1573
		1574	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1575
		1576	decode_ws (&dec);
949	sv = decode_sv (&dec);	1577	sv = decode_sv (&dec);
950		1578
		1579	if (offset_return)
		1580	*offset_return = dec.cur;
		1581
		1582	if (!(offset_return || !sv))
		1583	{
		1584	// check for trailing garbage
		1585	decode_ws (&dec);
		1586
		1587	if (*dec.cur)
		1588	{
		1589	dec.err = "garbage after JSON object";
		1590	SvREFCNT_dec (sv);
		1591	sv = 0;
		1592	}
		1593	}
		1594
951	if (!sv)	1595	if (!sv)
952	{	1596	{
953	IV offset = dec.flags & F_UTF8
954	? dec.cur - SvPVX (string)
955	: utf8_distance (dec.cur, SvPVX (string));
956	SV *uni = sv_newmortal ();	1597	SV *uni = sv_newmortal ();
957		1598
958	// horrible hack to silence warning inside pv_uni_display	1599	// horrible hack to silence warning inside pv_uni_display
959	COP cop = *PL_curcop;	1600	COP cop = *PL_curcop;
960	cop.cop_warnings = pWARN_NONE;	1601	cop.cop_warnings = pWARN_NONE;
…		…
962	SAVEVPTR (PL_curcop);	1603	SAVEVPTR (PL_curcop);
963	PL_curcop = &cop;	1604	PL_curcop = &cop;
964	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1605	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
965	LEAVE;	1606	LEAVE;
966		1607
967	croak ("%s, at character offset %d (%s)",	1608	croak ("%s, at character offset %d (before \"%s\")",
968	dec.err,	1609	dec.err,
969	(int)offset,	1610	(int)ptr_to_index (string, dec.cur),
970	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1611	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
971	}	1612	}
972		1613
973	sv = sv_2mortal (sv);	1614	sv = sv_2mortal (sv);
974		1615
975	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1616	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
976	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1617	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
977		1618
978	return sv;	1619	return sv;
		1620	}
		1621
		1622	/////////////////////////////////////////////////////////////////////////////
		1623	// incremental parser
		1624
		1625	static void
		1626	incr_parse (JSON *self)
		1627	{
		1628	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1629
		1630	// the state machine here is a bit convoluted and could be simplified a lot
		1631	// but this would make it slower, so...
		1632
		1633	for (;;)
		1634	{
		1635	//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
		1636	switch (self->incr_mode)
		1637	{
		1638	// only used for initial whitespace skipping
		1639	case INCR_M_WS:
		1640	for (;;)
		1641	{
		1642	if (*p > 0x20)
		1643	{
		1644	if (*p == '#')
		1645	{
		1646	self->incr_mode = INCR_M_C0;
		1647	goto incr_m_c;
		1648	}
		1649	else
		1650	{
		1651	self->incr_mode = INCR_M_JSON;
		1652	goto incr_m_json;
		1653	}
		1654	}
		1655	else if (!*p)
		1656	goto interrupt;
		1657
		1658	++p;
		1659	}
		1660
		1661	// skip a single char inside a string (for \\-processing)
		1662	case INCR_M_BS:
		1663	if (!*p)
		1664	goto interrupt;
		1665
		1666	++p;
		1667	self->incr_mode = INCR_M_STR;
		1668	goto incr_m_str;
		1669
		1670	// inside #-style comments
		1671	case INCR_M_C0:
		1672	case INCR_M_C1:
		1673	incr_m_c:
		1674	for (;;)
		1675	{
		1676	if (*p == '\n')
		1677	{
		1678	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1679	break;
		1680	}
		1681	else if (!*p)
		1682	goto interrupt;
		1683
		1684	++p;
		1685	}
		1686
		1687	break;
		1688
		1689	// inside a string
		1690	case INCR_M_STR:
		1691	incr_m_str:
		1692	for (;;)
		1693	{
		1694	if (*p == '"')
		1695	{
		1696	++p;
		1697	self->incr_mode = INCR_M_JSON;
		1698
		1699	if (!self->incr_nest)
		1700	goto interrupt;
		1701
		1702	goto incr_m_json;
		1703	}
		1704	else if (*p == '\\')
		1705	{
		1706	++p; // "virtually" consumes character after \
		1707
		1708	if (!*p) // if at end of string we have to switch modes
		1709	{
		1710	self->incr_mode = INCR_M_BS;
		1711	goto interrupt;
		1712	}
		1713	}
		1714	else if (!*p)
		1715	goto interrupt;
		1716
		1717	++p;
		1718	}
		1719
		1720	// after initial ws, outside string
		1721	case INCR_M_JSON:
		1722	incr_m_json:
		1723	for (;;)
		1724	{
		1725	switch (*p++)
		1726	{
		1727	case 0:
		1728	--p;
		1729	goto interrupt;
		1730
		1731	case 0x09:
		1732	case 0x0a:
		1733	case 0x0d:
		1734	case 0x20:
		1735	if (!self->incr_nest)
		1736	{
		1737	--p; // do not eat the whitespace, let the next round do it
		1738	goto interrupt;
		1739	}
		1740	break;
		1741
		1742	case '"':
		1743	self->incr_mode = INCR_M_STR;
		1744	goto incr_m_str;
		1745
		1746	case '[':
		1747	case '{':
		1748	if (++self->incr_nest > self->max_depth)
		1749	croak (ERR_NESTING_EXCEEDED);
		1750	break;
		1751
		1752	case ']':
		1753	case '}':
		1754	if (--self->incr_nest <= 0)
		1755	goto interrupt;
		1756	break;
		1757
		1758	case '#':
		1759	self->incr_mode = INCR_M_C1;
		1760	goto incr_m_c;
		1761	}
		1762	}
		1763	}
		1764
		1765	modechange:
		1766	;
		1767	}
		1768
		1769	interrupt:
		1770	self->incr_pos = p - SvPVX (self->incr_text);
		1771	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1772	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
979	}	1773	}
980		1774
981	/////////////////////////////////////////////////////////////////////////////	1775	/////////////////////////////////////////////////////////////////////////////
982	// XS interface functions	1776	// XS interface functions
983		1777
…		…
985		1779
986	BOOT:	1780	BOOT:
987	{	1781	{
988	int i;	1782	int i;
989		1783
990	memset (decode_hexdigit, 0xff, 256);
991	for (i = 10; i--; )	1784	for (i = 0; i < 256; ++i)
992	decode_hexdigit ['0' + i] = i;	1785	decode_hexdigit [i] =
		1786	i >= '0' && i <= '9' ? i - '0'
		1787	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1788	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1789	: -1;
993		1790
994	for (i = 7; i--; )
995	{
996	decode_hexdigit ['a' + i] = 10 + i;
997	decode_hexdigit ['A' + i] = 10 + i;
998	}
999
1000	json_stash = gv_stashpv ("JSON::XS", 1);	1791	json_stash = gv_stashpv ("JSON::XS" , 1);
		1792	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1793
		1794	json_true = get_bool ("JSON::XS::true");
		1795	json_false = get_bool ("JSON::XS::false");
		1796
		1797	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1001	}	1798	}
1002		1799
1003	PROTOTYPES: DISABLE	1800	PROTOTYPES: DISABLE
1004		1801
1005	SV *new (char *dummy)	1802	void CLONE (...)
1006	CODE:	1803	CODE:
1007	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1804	json_stash = 0;
		1805	json_boolean_stash = 0;
		1806
		1807	void new (char *klass)
		1808	PPCODE:
		1809	{
		1810	SV *pv = NEWSV (0, sizeof (JSON));
		1811	SvPOK_only (pv);
		1812	json_init ((JSON *)SvPVX (pv));
		1813	XPUSHs (sv_2mortal (sv_bless (
		1814	newRV_noinc (pv),
		1815	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1816	)));
		1817	}
		1818
		1819	void ascii (JSON *self, int enable = 1)
		1820	ALIAS:
		1821	ascii = F_ASCII
		1822	latin1 = F_LATIN1
		1823	utf8 = F_UTF8
		1824	indent = F_INDENT
		1825	canonical = F_CANONICAL
		1826	space_before = F_SPACE_BEFORE
		1827	space_after = F_SPACE_AFTER
		1828	pretty = F_PRETTY
		1829	allow_nonref = F_ALLOW_NONREF
		1830	shrink = F_SHRINK
		1831	allow_blessed = F_ALLOW_BLESSED
		1832	convert_blessed = F_CONV_BLESSED
		1833	relaxed = F_RELAXED
		1834	allow_unknown = F_ALLOW_UNKNOWN
		1835	PPCODE:
		1836	{
		1837	if (enable)
		1838	self->flags |= ix;
		1839	else
		1840	self->flags &= ~ix;
		1841
		1842	XPUSHs (ST (0));
		1843	}
		1844
		1845	void get_ascii (JSON *self)
		1846	ALIAS:
		1847	get_ascii = F_ASCII
		1848	get_latin1 = F_LATIN1
		1849	get_utf8 = F_UTF8
		1850	get_indent = F_INDENT
		1851	get_canonical = F_CANONICAL
		1852	get_space_before = F_SPACE_BEFORE
		1853	get_space_after = F_SPACE_AFTER
		1854	get_allow_nonref = F_ALLOW_NONREF
		1855	get_shrink = F_SHRINK
		1856	get_allow_blessed = F_ALLOW_BLESSED
		1857	get_convert_blessed = F_CONV_BLESSED
		1858	get_relaxed = F_RELAXED
		1859	get_allow_unknown = F_ALLOW_UNKNOWN
		1860	PPCODE:
		1861	XPUSHs (boolSV (self->flags & ix));
		1862
		1863	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		1864	PPCODE:
		1865	self->max_depth = max_depth;
		1866	XPUSHs (ST (0));
		1867
		1868	U32 get_max_depth (JSON *self)
		1869	CODE:
		1870	RETVAL = self->max_depth;
1008	OUTPUT:	1871	OUTPUT:
1009	RETVAL	1872	RETVAL
1010		1873
1011	SV *ascii (SV *self, int enable = 1)	1874	void max_size (JSON *self, U32 max_size = 0)
1012	ALIAS:	1875	PPCODE:
1013	ascii = F_ASCII	1876	self->max_size = max_size;
1014	utf8 = F_UTF8	1877	XPUSHs (ST (0));
1015	indent = F_INDENT	1878
1016	canonical = F_CANONICAL	1879	int get_max_size (JSON *self)
1017	space_before = F_SPACE_BEFORE
1018	space_after = F_SPACE_AFTER
1019	pretty = F_PRETTY
1020	allow_nonref = F_ALLOW_NONREF
1021	shrink = F_SHRINK
1022	CODE:	1880	CODE:
1023	{	1881	RETVAL = self->max_size;
1024	UV *uv = SvJSON (self);
1025	if (enable)
1026	*uv |= ix;
1027	else
1028	*uv &= ~ix;
1029
1030	RETVAL = newSVsv (self);
1031	}
1032	OUTPUT:	1882	OUTPUT:
1033	RETVAL	1883	RETVAL
1034		1884
1035	void encode (SV *self, SV *scalar)	1885	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1036	PPCODE:	1886	PPCODE:
1037	XPUSHs (encode_json (scalar, *SvJSON (self)));	1887	{
		1888	SvREFCNT_dec (self->cb_object);
		1889	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
1038		1890
1039	void decode (SV *self, SV *jsonstr)	1891	XPUSHs (ST (0));
		1892	}
		1893
		1894	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1040	PPCODE:	1895	PPCODE:
1041	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1896	{
		1897	if (!self->cb_sk_object)
		1898	self->cb_sk_object = newHV ();
		1899
		1900	if (SvOK (cb))
		1901	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1902	else
		1903	{
		1904	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1905
		1906	if (!HvKEYS (self->cb_sk_object))
		1907	{
		1908	SvREFCNT_dec (self->cb_sk_object);
		1909	self->cb_sk_object = 0;
		1910	}
		1911	}
		1912
		1913	XPUSHs (ST (0));
		1914	}
		1915
		1916	void encode (JSON *self, SV *scalar)
		1917	PPCODE:
		1918	PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
		1919	XPUSHs (scalar);
		1920
		1921	void decode (JSON *self, SV *jsonstr)
		1922	PPCODE:
		1923	PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
		1924	XPUSHs (jsonstr);
		1925
		1926	void decode_prefix (JSON *self, SV *jsonstr)
		1927	PPCODE:
		1928	{
		1929	SV *sv;
		1930	char *offset;
		1931	PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
		1932	EXTEND (SP, 2);
		1933	PUSHs (sv);
		1934	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		1935	}
		1936
		1937	void incr_parse (JSON *self, SV *jsonstr = 0)
		1938	PPCODE:
		1939	{
		1940	if (!self->incr_text)
		1941	self->incr_text = newSVpvn ("", 0);
		1942
		1943	/* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
		1944	if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
		1945	if (DECODE_WANTS_OCTETS (self))
		1946	{
		1947	if (self->incr_pos)
		1948	self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
		1949	(U8 *)SvPVX (self->incr_text) + self->incr_pos);
		1950
		1951	sv_utf8_downgrade (self->incr_text, 0);
		1952	}
		1953	else
		1954	{
		1955	sv_utf8_upgrade (self->incr_text);
		1956
		1957	if (self->incr_pos)
		1958	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1959	- (U8 *)SvPVX (self->incr_text);
		1960	}
		1961
		1962	// append data, if any
		1963	if (jsonstr)
		1964	{
		1965	/* make sure both strings have same encoding */
		1966	if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
		1967	if (SvUTF8 (jsonstr))
		1968	sv_utf8_downgrade (jsonstr, 0);
		1969	else
		1970	sv_utf8_upgrade (jsonstr);
		1971
		1972	/* and then just blindly append */
		1973	{
		1974	STRLEN len;
		1975	const char *str = SvPV (jsonstr, len);
		1976	STRLEN cur = SvCUR (self->incr_text);
		1977
		1978	if (SvLEN (self->incr_text) <= cur + len)
		1979	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1980
		1981	Move (str, SvEND (self->incr_text), len, char);
		1982	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1983	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1984	}
		1985	}
		1986
		1987	if (GIMME_V != G_VOID)
		1988	do
		1989	{
		1990	SV *sv;
		1991	char *offset;
		1992
		1993	if (!INCR_DONE (self))
		1994	{
		1995	incr_parse (self);
		1996
		1997	if (self->incr_pos > self->max_size && self->max_size)
		1998	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1999	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		2000
		2001	if (!INCR_DONE (self))
		2002	{
		2003	// as an optimisation, do not accumulate white space in the incr buffer
		2004	if (self->incr_mode == INCR_M_WS && self->incr_pos)
		2005	{
		2006	self->incr_pos = 0;
		2007	SvCUR_set (self->incr_text, 0);
		2008	}
		2009
		2010	break;
		2011	}
		2012	}
		2013
		2014	PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
		2015	XPUSHs (sv);
		2016
		2017	self->incr_pos -= offset - SvPVX (self->incr_text);
		2018	self->incr_nest = 0;
		2019	self->incr_mode = 0;
		2020
		2021	sv_chop (self->incr_text, offset);
		2022	}
		2023	while (GIMME_V == G_ARRAY);
		2024	}
		2025
		2026	SV *incr_text (JSON *self)
		2027	ATTRS: lvalue
		2028	CODE:
		2029	{
		2030	if (self->incr_pos)
		2031	croak ("incr_text can not be called when the incremental parser already started parsing");
		2032
		2033	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		2034	}
		2035	OUTPUT:
		2036	RETVAL
		2037
		2038	void incr_skip (JSON *self)
		2039	CODE:
		2040	{
		2041	if (self->incr_pos)
		2042	{
		2043	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2044	self->incr_pos = 0;
		2045	self->incr_nest = 0;
		2046	self->incr_mode = 0;
		2047	}
		2048	}
		2049
		2050	void incr_reset (JSON *self)
		2051	CODE:
		2052	{
		2053	SvREFCNT_dec (self->incr_text);
		2054	self->incr_text = 0;
		2055	self->incr_pos = 0;
		2056	self->incr_nest = 0;
		2057	self->incr_mode = 0;
		2058	}
		2059
		2060	void DESTROY (JSON *self)
		2061	CODE:
		2062	SvREFCNT_dec (self->cb_sk_object);
		2063	SvREFCNT_dec (self->cb_object);
		2064	SvREFCNT_dec (self->incr_text);
1042		2065
1043	PROTOTYPES: ENABLE	2066	PROTOTYPES: ENABLE
1044		2067
1045	void to_json (SV *scalar)	2068	void encode_json (SV *scalar)
1046	ALIAS:	2069	ALIAS:
1047	objToJson = 0	2070	to_json_ = 0
		2071	encode_json = F_UTF8
1048	PPCODE:	2072	PPCODE:
1049	XPUSHs (encode_json (scalar, F_UTF8));	2073	{
		2074	JSON json;
		2075	json_init (&json);
		2076	json.flags |= ix;
		2077	PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
		2078	XPUSHs (scalar);
		2079	}
1050		2080
1051	void from_json (SV *jsonstr)	2081	void decode_json (SV *jsonstr)
1052	ALIAS:	2082	ALIAS:
1053	jsonToObj = 0	2083	from_json_ = 0
		2084	decode_json = F_UTF8
1054	PPCODE:	2085	PPCODE:
1055	XPUSHs (decode_json (jsonstr, F_UTF8));	2086	{
		2087	JSON json;
		2088	json_init (&json);
		2089	json.flags |= ix;
		2090	PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
		2091	XPUSHs (jsonstr);
		2092	}
1056		2093

Diff Legend

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