ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.15 by root, Sun Mar 25 02:32:40 2007 UTC vs.
Revision 1.126 by root, Sun Feb 21 16:18:19 2016 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	// compatibility with perl <5.18
		23	#ifndef HvNAMELEN_get
		24	# define HvNAMELEN_get(hv) strlen (HvNAME (hv))
		25	#endif
		26	#ifndef HvNAMELEN
		27	# define HvNAMELEN(hv) HvNAMELEN_get (hv)
		28	#endif
		29	#ifndef HvNAMEUTF8
		30	# define HvNAMEUTF8(hv) 0
		31	#endif
		32
		33	// three extra for rounding, sign, and end of string
		34	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
		35
9	#define F_ASCII 0x00000001	36	#define F_ASCII 0x00000001UL
		37	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	38	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	39	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	40	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	41	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	42	#define F_SPACE_AFTER 0x00000040UL
15	#define F_ALLOW_NONREF 0x00000080	43	#define F_ALLOW_NONREF 0x00000100UL
16	#define F_SHRINK 0x00000100	44	#define F_SHRINK 0x00000200UL
		45	#define F_ALLOW_BLESSED 0x00000400UL
		46	#define F_CONV_BLESSED 0x00000800UL
		47	#define F_RELAXED 0x00001000UL
		48	#define F_ALLOW_UNKNOWN 0x00002000UL
		49	#define F_ALLOW_TAGS 0x00004000UL
		50	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
17		51
18	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	52	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
19	#define F_DEFAULT 0
20		53
21	#define INIT_SIZE 32 // initial scalar size to be allocated	54	#define INIT_SIZE 32 // initial scalar size to be allocated
22	#define INDENT_STEP 3 // spaces per indentation level	55	#define INDENT_STEP 3 // spaces per indentation level
23		56
24	#define UTF8_MAX_LEN 11 // for perls UTF-X: max. number of octets per character
25	#define SHORT_STRING_LEN 512 // special-case strings of up to this size	57	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
		58
		59	#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
26		60
27	#define SB do {	61	#define SB do {
28	#define SE } while (0)	62	#define SE } while (0)
29		63
30	static HV *json_stash; // JSON::XS::	64	#if __GNUC__ >= 3
		65	# define expect(expr,value) __builtin_expect ((expr), (value))
		66	# define INLINE static inline
		67	#else
		68	# define expect(expr,value) (expr)
		69	# define INLINE static
		70	#endif
		71
		72	#define expect_false(expr) expect ((expr) != 0, 0)
		73	#define expect_true(expr) expect ((expr) != 0, 1)
		74
		75	#define IN_RANGE_INC(type,val,beg,end) \
		76	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		77	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		78
		79	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
		80
		81	#ifdef USE_ITHREADS
		82	# define JSON_SLOW 1
		83	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		84	# define BOOL_STASH (bool_stash ? bool_stash : gv_stashpv ("Types::Serialiser::Boolean", 1))
		85	#else
		86	# define JSON_SLOW 0
		87	# define JSON_STASH json_stash
		88	# define BOOL_STASH bool_stash
		89	#endif
		90
		91	// the amount of HEs to allocate on the stack, when sorting keys
		92	#define STACK_HES 64
		93
		94	static HV *json_stash, *bool_stash; // JSON::XS::, Types::Serialiser::Boolean::
		95	static SV *bool_true, *bool_false, *sv_json;
		96
		97	enum {
		98	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		99	INCR_M_STR, // inside string
		100	INCR_M_BS, // inside backslash
		101	INCR_M_C0, // inside comment in initial whitespace sequence
		102	INCR_M_C1, // inside comment in other places
		103	INCR_M_JSON // outside anything, count nesting
		104	};
		105
		106	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		107
		108	typedef struct {
		109	U32 flags;
		110	U32 max_depth;
		111	STRLEN max_size;
		112
		113	SV *cb_object;
		114	HV *cb_sk_object;
		115
		116	// for the incremental parser
		117	SV *incr_text; // the source text so far
		118	STRLEN incr_pos; // the current offset into the text
		119	int incr_nest; // {[]}-nesting level
		120	unsigned char incr_mode;
		121	} JSON;
		122
		123	INLINE void
		124	json_init (JSON *json)
		125	{
		126	Zero (json, 1, JSON);
		127	json->max_depth = 512;
		128	}
31		129
32	/////////////////////////////////////////////////////////////////////////////	130	/////////////////////////////////////////////////////////////////////////////
33	// utility functions	131	// utility functions
34		132
35	static UV *	133	INLINE SV *
36	SvJSON (SV *sv)	134	get_bool (const char *name)
37	{	135	{
38	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))	136	SV *sv = get_sv (name, 1);
39	croak ("object is not of type JSON::XS");
40		137
41	return &SvUVX (SvRV (sv));	138	SvREADONLY_on (sv);
42	}	139	SvREADONLY_on (SvRV (sv));
43		140
44	static void	141	return sv;
		142	}
		143
		144	INLINE void
45	shrink (SV *sv)	145	shrink (SV *sv)
46	{	146	{
47	sv_utf8_downgrade (sv, 1);	147	sv_utf8_downgrade (sv, 1);
		148
48	if (SvLEN (sv) > SvCUR (sv) + 1)	149	if (SvLEN (sv) > SvCUR (sv) + 1)
49	{	150	{
50	#ifdef SvPV_shrink_to_cur	151	#ifdef SvPV_shrink_to_cur
51	SvPV_shrink_to_cur (sv);	152	SvPV_shrink_to_cur (sv);
52	#elif defined (SvPV_renew)	153	#elif defined (SvPV_renew)
…		…
58	// decode an utf-8 character and return it, or (UV)-1 in	159	// decode an utf-8 character and return it, or (UV)-1 in
59	// case of an error.	160	// case of an error.
60	// we special-case "safe" characters from U+80 .. U+7FF,	161	// we special-case "safe" characters from U+80 .. U+7FF,
61	// but use the very good perl function to parse anything else.	162	// but use the very good perl function to parse anything else.
62	// note that we never call this function for a ascii codepoints	163	// note that we never call this function for a ascii codepoints
63	static UV	164	INLINE UV
64	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)	165	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
65	{	166	{
66	if (s[0] > 0xdf || s[0] < 0xc2)	167	if (expect_true (len >= 2
67	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	168	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
68	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	169	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
69	{	170	{
70	*clen = 2;	171	*clen = 2;
71	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);	172	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
72	}	173	}
73	else	174	else
		175	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		176	}
		177
		178	// likewise for encoding, also never called for ascii codepoints
		179	// this function takes advantage of this fact, although current gccs
		180	// seem to optimise the check for >= 0x80 away anyways
		181	INLINE unsigned char *
		182	encode_utf8 (unsigned char *s, UV ch)
		183	{
		184	if (expect_false (ch < 0x000080))
		185	*s++ = ch;
		186	else if (expect_true (ch < 0x000800))
		187	*s++ = 0xc0 | ( ch >> 6),
		188	*s++ = 0x80 | ( ch & 0x3f);
		189	else if ( ch < 0x010000)
		190	*s++ = 0xe0 | ( ch >> 12),
		191	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		192	*s++ = 0x80 | ( ch & 0x3f);
		193	else if ( ch < 0x110000)
		194	*s++ = 0xf0 | ( ch >> 18),
		195	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		196	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		197	*s++ = 0x80 | ( ch & 0x3f);
		198
		199	return s;
		200	}
		201
		202	// convert offset pointer to character index, sv must be string
		203	static STRLEN
		204	ptr_to_index (SV *sv, char *offset)
		205	{
		206	return SvUTF8 (sv)
		207	? utf8_distance (offset, SvPVX (sv))
		208	: offset - SvPVX (sv);
		209	}
		210
		211	/////////////////////////////////////////////////////////////////////////////
		212	// fp hell
		213
		214	// scan a group of digits, and a trailing exponent
		215	static void
		216	json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
		217	{
		218	UV uaccum = 0;
		219	int eaccum = 0;
		220
		221	// if we recurse too deep, skip all remaining digits
		222	// to avoid a stack overflow attack
		223	if (expect_false (--maxdepth <= 0))
		224	while (((U8)*s - '0') < 10)
		225	++s;
		226
		227	for (;;)
		228	{
		229	U8 dig = (U8)*s - '0';
		230
		231	if (expect_false (dig >= 10))
		232	{
		233	if (dig == (U8)((U8)'.' - (U8)'0'))
		234	{
		235	++s;
		236	json_atof_scan1 (s, accum, expo, 1, maxdepth);
		237	}
		238	else if ((dig | ' ') == 'e' - '0')
		239	{
		240	int exp2 = 0;
		241	int neg = 0;
		242
		243	++s;
		244
		245	if (*s == '-')
		246	{
		247	++s;
		248	neg = 1;
		249	}
		250	else if (*s == '+')
		251	++s;
		252
		253	while ((dig = (U8)*s - '0') < 10)
		254	exp2 = exp2 * 10 + *s++ - '0';
		255
		256	*expo += neg ? -exp2 : exp2;
		257	}
		258
		259	break;
		260	}
		261
		262	++s;
		263
		264	uaccum = uaccum * 10 + dig;
		265	++eaccum;
		266
		267	// if we have too many digits, then recurse for more
		268	// we actually do this for rather few digits
		269	if (uaccum >= (UV_MAX - 9) / 10)
		270	{
		271	if (postdp) *expo -= eaccum;
		272	json_atof_scan1 (s, accum, expo, postdp, maxdepth);
		273	if (postdp) *expo += eaccum;
		274
		275	break;
		276	}
		277	}
		278
		279	// this relies greatly on the quality of the pow ()
		280	// implementation of the platform, but a good
		281	// implementation is hard to beat.
		282	// (IEEE 754 conformant ones are required to be exact)
		283	if (postdp) *expo -= eaccum;
		284	*accum += uaccum * Perl_pow (10., *expo);
		285	*expo += eaccum;
		286	}
		287
		288	static NV
		289	json_atof (const char *s)
		290	{
		291	NV accum = 0.;
		292	int expo = 0;
		293	int neg = 0;
		294
		295	if (*s == '-')
		296	{
		297	++s;
		298	neg = 1;
		299	}
		300
		301	// a recursion depth of ten gives us >>500 bits
		302	json_atof_scan1 (s, &accum, &expo, 0, 10);
		303
		304	return neg ? -accum : accum;
		305	}
		306
		307	// target of scalar reference is bool? -1 == nope, 0 == false, 1 == true
		308	static int
		309	ref_bool_type (SV *sv)
		310	{
		311	svtype svt = SvTYPE (sv);
		312
		313	if (svt < SVt_PVAV)
		314	{
		315	STRLEN len = 0;
		316	char *pv = svt ? SvPV (sv, len) : 0;
		317
		318	if (len == 1)
		319	if (*pv == '1')
		320	return 1;
		321	else if (*pv == '0')
		322	return 0;
		323
		324	}
		325
		326	return -1;
		327	}
		328
		329	// returns whether scalar is not a reference in the sense of allow_nonref
		330	static int
		331	json_nonref (SV *scalar)
		332	{
		333	if (!SvROK (scalar))
74	return (UV)-1;	334	return 1;
		335
		336	scalar = SvRV (scalar);
		337
		338	if (SvSTASH (scalar) == bool_stash)
		339	return 1;
		340
		341	if (!SvOBJECT (scalar) && ref_bool_type (scalar) >= 0)
		342	return 1;
		343
		344	return 0;
75	}	345	}
76		346
77	/////////////////////////////////////////////////////////////////////////////	347	/////////////////////////////////////////////////////////////////////////////
78	// encoder	348	// encoder
79		349
…		…
81	typedef struct	351	typedef struct
82	{	352	{
83	char *cur; // SvPVX (sv) + current output position	353	char *cur; // SvPVX (sv) + current output position
84	char *end; // SvEND (sv)	354	char *end; // SvEND (sv)
85	SV *sv; // result scalar	355	SV *sv; // result scalar
86	UV flags; // F_*	356	JSON json;
87	int indent; // indentation level	357	U32 indent; // indentation level
88	int max_depth; // max. recursion level	358	UV limit; // escape character values >= this value when encoding
89	} enc_t;	359	} enc_t;
90		360
91	static void	361	INLINE void
92	need (enc_t *enc, STRLEN len)	362	need (enc_t *enc, STRLEN len)
93	{	363	{
94	if (enc->cur + len >= enc->end)	364	if (expect_false (enc->cur + len >= enc->end))
95	{	365	{
96	STRLEN cur = enc->cur - SvPVX (enc->sv);	366	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
97	SvGROW (enc->sv, cur + len + 1);	367	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
98	enc->cur = SvPVX (enc->sv) + cur;	368	enc->cur = SvPVX (enc->sv) + cur;
99	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	369	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
100	}	370	}
101	}	371	}
102		372
103	static void	373	INLINE void
104	encode_ch (enc_t *enc, char ch)	374	encode_ch (enc_t *enc, char ch)
105	{	375	{
106	need (enc, 1);	376	need (enc, 1);
107	*enc->cur++ = ch;	377	*enc->cur++ = ch;
108	}	378	}
…		…
116		386
117	while (str < end)	387	while (str < end)
118	{	388	{
119	unsigned char ch = *(unsigned char *)str;	389	unsigned char ch = *(unsigned char *)str;
120		390
121	if (ch >= 0x20 && ch < 0x80) // most common case	391	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
122	{	392	{
123	if (ch == '"') // but with slow exceptions	393	if (expect_false (ch == '"')) // but with slow exceptions
124	{	394	{
125	need (enc, len += 1);	395	need (enc, len += 1);
126	*enc->cur++ = '\\';	396	*enc->cur++ = '\\';
127	*enc->cur++ = '"';	397	*enc->cur++ = '"';
128	}	398	}
129	else if (ch == '\\')	399	else if (expect_false (ch == '\\'))
130	{	400	{
131	need (enc, len += 1);	401	need (enc, len += 1);
132	*enc->cur++ = '\\';	402	*enc->cur++ = '\\';
133	*enc->cur++ = '\\';	403	*enc->cur++ = '\\';
134	}	404	}
…		…
152	STRLEN clen;	422	STRLEN clen;
153	UV uch;	423	UV uch;
154		424
155	if (is_utf8)	425	if (is_utf8)
156	{	426	{
157	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
158	uch = decode_utf8 (str, end - str, &clen);	427	uch = decode_utf8 (str, end - str, &clen);
159	if (clen == (STRLEN)-1)	428	if (clen == (STRLEN)-1)
160	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	429	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
161	}	430	}
162	else	431	else
163	{	432	{
164	uch = ch;	433	uch = ch;
165	clen = 1;	434	clen = 1;
166	}	435	}
167		436
168	if (uch > 0x10FFFFUL)	437	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
169	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
170
171	if (uch < 0x80 || enc->flags & F_ASCII)
172	{	438	{
173	if (uch > 0xFFFFUL)	439	if (uch >= 0x10000UL)
174	{	440	{
		441	if (uch >= 0x110000UL)
		442	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		443
175	need (enc, len += 11);	444	need (enc, len += 11);
176	sprintf (enc->cur, "\\u%04x\\u%04x",	445	sprintf (enc->cur, "\\u%04x\\u%04x",
177	(int)((uch - 0x10000) / 0x400 + 0xD800),	446	(int)((uch - 0x10000) / 0x400 + 0xD800),
178	(int)((uch - 0x10000) % 0x400 + 0xDC00));	447	(int)((uch - 0x10000) % 0x400 + 0xDC00));
179	enc->cur += 12;	448	enc->cur += 12;
180	}	449	}
181	else	450	else
182	{	451	{
183	static char hexdigit [16] = "0123456789abcdef";
184	need (enc, len += 5);	452	need (enc, len += 5);
185	*enc->cur++ = '\\';	453	*enc->cur++ = '\\';
186	*enc->cur++ = 'u';	454	*enc->cur++ = 'u';
187	*enc->cur++ = hexdigit [ uch >> 12 ];	455	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
188	*enc->cur++ = hexdigit [(uch >> 8) & 15];	456	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
189	*enc->cur++ = hexdigit [(uch >> 4) & 15];	457	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
190	*enc->cur++ = hexdigit [(uch >> 0) & 15];	458	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
191	}	459	}
192		460
		461	str += clen;
		462	}
		463	else if (enc->json.flags & F_LATIN1)
		464	{
		465	*enc->cur++ = uch;
193	str += clen;	466	str += clen;
194	}	467	}
195	else if (is_utf8)	468	else if (is_utf8)
196	{	469	{
197	need (enc, len += clen);	470	need (enc, len += clen);
…		…
201	}	474	}
202	while (--clen);	475	while (--clen);
203	}	476	}
204	else	477	else
205	{	478	{
206	need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed	479	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
207	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	480	enc->cur = encode_utf8 (enc->cur, uch);
208	++str;	481	++str;
209	}	482	}
210	}	483	}
211	}	484	}
212	}	485	}
213		486
214	--len;	487	--len;
215	}	488	}
216	}	489	}
217		490
218	static void	491	INLINE void
219	encode_indent (enc_t *enc)	492	encode_indent (enc_t *enc)
220	{	493	{
221	if (enc->flags & F_INDENT)	494	if (enc->json.flags & F_INDENT)
222	{	495	{
223	int spaces = enc->indent * INDENT_STEP;	496	int spaces = enc->indent * INDENT_STEP;
224		497
225	need (enc, spaces);	498	need (enc, spaces);
226	memset (enc->cur, ' ', spaces);	499	memset (enc->cur, ' ', spaces);
227	enc->cur += spaces;	500	enc->cur += spaces;
228	}	501	}
229	}	502	}
230		503
231	static void	504	INLINE void
232	encode_space (enc_t *enc)	505	encode_space (enc_t *enc)
233	{	506	{
234	need (enc, 1);	507	need (enc, 1);
235	encode_ch (enc, ' ');	508	encode_ch (enc, ' ');
236	}	509	}
237		510
238	static void	511	INLINE void
239	encode_nl (enc_t *enc)	512	encode_nl (enc_t *enc)
240	{	513	{
241	if (enc->flags & F_INDENT)	514	if (enc->json.flags & F_INDENT)
242	{	515	{
243	need (enc, 1);	516	need (enc, 1);
244	encode_ch (enc, '\n');	517	encode_ch (enc, '\n');
245	}	518	}
246	}	519	}
247		520
248	static void	521	INLINE void
249	encode_comma (enc_t *enc)	522	encode_comma (enc_t *enc)
250	{	523	{
251	encode_ch (enc, ',');	524	encode_ch (enc, ',');
252		525
253	if (enc->flags & F_INDENT)	526	if (enc->json.flags & F_INDENT)
254	encode_nl (enc);	527	encode_nl (enc);
255	else if (enc->flags & F_SPACE_AFTER)	528	else if (enc->json.flags & F_SPACE_AFTER)
256	encode_space (enc);	529	encode_space (enc);
257	}	530	}
258		531
259	static void encode_sv (enc_t *enc, SV *sv);	532	static void encode_sv (enc_t *enc, SV *sv);
260		533
261	static void	534	static void
262	encode_av (enc_t *enc, AV *av)	535	encode_av (enc_t *enc, AV *av)
263	{	536	{
264	int i, len = av_len (av);	537	int i, len = av_len (av);
265		538
266	encode_ch (enc, '['); encode_nl (enc);	539	if (enc->indent >= enc->json.max_depth)
267	++enc->indent;	540	croak (ERR_NESTING_EXCEEDED);
268		541
		542	encode_ch (enc, '[');
		543
		544	if (len >= 0)
		545	{
		546	encode_nl (enc); ++enc->indent;
		547
269	for (i = 0; i <= len; ++i)	548	for (i = 0; i <= len; ++i)
270	{	549	{
		550	SV **svp = av_fetch (av, i, 0);
		551
271	encode_indent (enc);	552	encode_indent (enc);
272	encode_sv (enc, *av_fetch (av, i, 0));
273		553
		554	if (svp)
		555	encode_sv (enc, *svp);
		556	else
		557	encode_str (enc, "null", 4, 0);
		558
274	if (i < len)	559	if (i < len)
275	encode_comma (enc);	560	encode_comma (enc);
276	}	561	}
277		562
		563	encode_nl (enc); --enc->indent; encode_indent (enc);
		564	}
		565
278	encode_nl (enc);	566	encode_ch (enc, ']');
279
280	--enc->indent;
281	encode_indent (enc); encode_ch (enc, ']');
282	}	567	}
283		568
284	static void	569	static void
285	encode_he (enc_t *enc, HE *he)	570	encode_hk (enc_t *enc, HE *he)
286	{	571	{
287	encode_ch (enc, '"');	572	encode_ch (enc, '"');
288		573
289	if (HeKLEN (he) == HEf_SVKEY)	574	if (HeKLEN (he) == HEf_SVKEY)
290	{	575	{
291	SV *sv = HeSVKEY (he);	576	SV *sv = HeSVKEY (he);
292	STRLEN len;	577	STRLEN len;
293	char *str;	578	char *str;
294		579
295	SvGETMAGIC (sv);	580	SvGETMAGIC (sv);
296	str = SvPV (sv, len);	581	str = SvPV (sv, len);
297		582
298	encode_str (enc, str, len, SvUTF8 (sv));	583	encode_str (enc, str, len, SvUTF8 (sv));
299	}	584	}
300	else	585	else
301	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	586	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
302		587
303	encode_ch (enc, '"');	588	encode_ch (enc, '"');
304		589
305	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	590	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
306	encode_ch (enc, ':');	591	encode_ch (enc, ':');
307	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	592	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
308	encode_sv (enc, HeVAL (he));
309	}	593	}
310		594
311	// compare hash entries, used when all keys are bytestrings	595	// compare hash entries, used when all keys are bytestrings
312	static int	596	static int
313	he_cmp_fast (const void *a_, const void *b_)	597	he_cmp_fast (const void *a_, const void *b_)
…		…
318	HE *b = *(HE **)b_;	602	HE *b = *(HE **)b_;
319		603
320	STRLEN la = HeKLEN (a);	604	STRLEN la = HeKLEN (a);
321	STRLEN lb = HeKLEN (b);	605	STRLEN lb = HeKLEN (b);
322		606
323	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	607	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
324	cmp = la - lb;	608	cmp = lb - la;
325		609
326	return cmp;	610	return cmp;
327	}	611	}
328		612
329	// compare hash entries, used when some keys are sv's or utf-x	613	// compare hash entries, used when some keys are sv's or utf-x
330	static int	614	static int
331	he_cmp_slow (const void *a, const void *b)	615	he_cmp_slow (const void *a, const void *b)
332	{	616	{
333	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	617	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
334	}	618	}
335		619
336	static void	620	static void
337	encode_hv (enc_t *enc, HV *hv)	621	encode_hv (enc_t *enc, HV *hv)
338	{	622	{
339	int count, i;	623	HE *he;
340		624
341	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	625	if (enc->indent >= enc->json.max_depth)
		626	croak (ERR_NESTING_EXCEEDED);
342		627
343	if ((count = hv_iterinit (hv)))	628	encode_ch (enc, '{');
344	{	629
345	// for canonical output we have to sort by keys first	630	// for canonical output we have to sort by keys first
346	// actually, this is mostly due to the stupid so-called	631	// actually, this is mostly due to the stupid so-called
347	// security workaround added somewhere in 5.8.x.	632	// security workaround added somewhere in 5.8.x
348	// that randomises hash orderings	633	// that randomises hash orderings
349	if (enc->flags & F_CANONICAL)	634	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
		635	{
		636	int count = hv_iterinit (hv);
		637
		638	if (SvMAGICAL (hv))
350	{	639	{
351	HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode	640	// need to count by iterating. could improve by dynamically building the vector below
		641	// but I don't care for the speed of this special case.
		642	// note also that we will run into undefined behaviour when the two iterations
		643	// do not result in the same count, something I might care for in some later release.
		644
		645	count = 0;
		646	while (hv_iternext (hv))
		647	++count;
		648
		649	hv_iterinit (hv);
		650	}
		651
		652	if (count)
		653	{
352	int fast = 1;	654	int i, fast = 1;
		655	HE *hes_stack [STACK_HES];
		656	HE **hes = hes_stack;
		657
		658	// allocate larger arrays on the heap
		659	if (count > STACK_HES)
		660	{
		661	SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
		662	hes = (HE **)SvPVX (sv);
		663	}
353		664
354	i = 0;	665	i = 0;
355	while ((he = hv_iternext (hv)))	666	while ((he = hv_iternext (hv)))
356	{	667	{
357	hes [i++] = he;	668	hes [i++] = he;
…		…
379		690
380	FREETMPS;	691	FREETMPS;
381	LEAVE;	692	LEAVE;
382	}	693	}
383		694
384	for (i = 0; i < count; ++i)	695	encode_nl (enc); ++enc->indent;
		696
		697	while (count--)
385	{	698	{
386	encode_indent (enc);	699	encode_indent (enc);
		700	he = hes [count];
387	encode_he (enc, hes [i]);	701	encode_hk (enc, he);
		702	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
388		703
389	if (i < count - 1)	704	if (count)
390	encode_comma (enc);	705	encode_comma (enc);
391	}	706	}
392		707
393	encode_nl (enc);	708	encode_nl (enc); --enc->indent; encode_indent (enc);
394	}	709	}
		710	}
		711	else
		712	{
		713	if (hv_iterinit (hv) || SvMAGICAL (hv))
		714	if ((he = hv_iternext (hv)))
		715	{
		716	encode_nl (enc); ++enc->indent;
		717
		718	for (;;)
		719	{
		720	encode_indent (enc);
		721	encode_hk (enc, he);
		722	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		723
		724	if (!(he = hv_iternext (hv)))
		725	break;
		726
		727	encode_comma (enc);
		728	}
		729
		730	encode_nl (enc); --enc->indent; encode_indent (enc);
		731	}
		732	}
		733
		734	encode_ch (enc, '}');
		735	}
		736
		737	// encode objects, arrays and special \0=false and \1=true values.
		738	static void
		739	encode_rv (enc_t *enc, SV *sv)
		740	{
		741	svtype svt;
		742	GV *method;
		743
		744	SvGETMAGIC (sv);
		745	svt = SvTYPE (sv);
		746
		747	if (expect_false (SvOBJECT (sv)))
		748	{
		749	HV *stash = SvSTASH (sv);
		750
		751	if (stash == bool_stash)
		752	{
		753	if (SvIV (sv))
		754	encode_str (enc, "true", 4, 0);
		755	else
		756	encode_str (enc, "false", 5, 0);
		757	}
		758	else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
		759	{
		760	int count;
		761	dSP;
		762
		763	ENTER; SAVETMPS;
		764	SAVESTACK_POS ();
		765	PUSHMARK (SP);
		766	EXTEND (SP, 2);
		767	// we re-bless the reference to get overload and other niceties right
		768	PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		769	PUSHs (sv_json);
		770
		771	PUTBACK;
		772	count = call_sv ((SV *)GvCV (method), G_ARRAY);
		773	SPAGAIN;
		774
		775	// catch this surprisingly common error
		776	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		777	croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		778
		779	encode_ch (enc, '(');
		780	encode_ch (enc, '"');
		781	encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
		782	encode_ch (enc, '"');
		783	encode_ch (enc, ')');
		784	encode_ch (enc, '[');
		785
		786	while (count)
		787	{
		788	encode_sv (enc, SP[1 - count--]);
		789
		790	if (count)
		791	encode_ch (enc, ',');
		792	}
		793
		794	encode_ch (enc, ']');
		795
		796	FREETMPS; LEAVE;
		797	}
		798	else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
		799	{
		800	dSP;
		801
		802	ENTER; SAVETMPS;
		803	PUSHMARK (SP);
		804	// we re-bless the reference to get overload and other niceties right
		805	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		806
		807	// calling with G_SCALAR ensures that we always get a 1 return value
		808	PUTBACK;
		809	call_sv ((SV *)GvCV (method), G_SCALAR);
		810	SPAGAIN;
		811
		812	// catch this surprisingly common error
		813	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		814	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		815
		816	sv = POPs;
		817	PUTBACK;
		818
		819	encode_sv (enc, sv);
		820
		821	FREETMPS; LEAVE;
		822	}
		823	else if (enc->json.flags & F_ALLOW_BLESSED)
		824	encode_str (enc, "null", 4, 0);
395	else	825	else
396	{	826	croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
397	SV *sv;	827	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
398	HE *he = hv_iternext (hv);	828	}
399		829	else if (svt == SVt_PVHV)
400	for (;;)	830	encode_hv (enc, (HV *)sv);
401	{	831	else if (svt == SVt_PVAV)
402	encode_indent (enc);	832	encode_av (enc, (AV *)sv);
403	encode_he (enc, he);	833	else if (svt < SVt_PVAV)
404
405	if (!(he = hv_iternext (hv)))
406	break;
407
408	encode_comma (enc);
409	}
410
411	encode_nl (enc);
412	}
413	}	834	{
		835	int bool_type = ref_bool_type (sv);
414		836
415	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	837	if (bool_type == 1)
		838	encode_str (enc, "true", 4, 0);
		839	else if (bool_type == 0)
		840	encode_str (enc, "false", 5, 0);
		841	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		842	encode_str (enc, "null", 4, 0);
		843	else
		844	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		845	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		846	}
		847	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		848	encode_str (enc, "null", 4, 0);
		849	else
		850	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		851	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
416	}	852	}
417		853
418	static void	854	static void
419	encode_sv (enc_t *enc, SV *sv)	855	encode_sv (enc_t *enc, SV *sv)
420	{	856	{
…		…
428	encode_str (enc, str, len, SvUTF8 (sv));	864	encode_str (enc, str, len, SvUTF8 (sv));
429	encode_ch (enc, '"');	865	encode_ch (enc, '"');
430	}	866	}
431	else if (SvNOKp (sv))	867	else if (SvNOKp (sv))
432	{	868	{
		869	// trust that perl will do the right thing w.r.t. JSON syntax.
433	need (enc, NV_DIG + 32);	870	need (enc, NV_DIG + 32);
434	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	871	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
435	enc->cur += strlen (enc->cur);	872	enc->cur += strlen (enc->cur);
436	}	873	}
437	else if (SvIOKp (sv))	874	else if (SvIOKp (sv))
438	{	875	{
		876	// we assume we can always read an IV as a UV and vice versa
		877	// we assume two's complement
		878	// we assume no aliasing issues in the union
		879	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		880	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		881	{
		882	// optimise the "small number case"
		883	// code will likely be branchless and use only a single multiplication
		884	// works for numbers up to 59074
		885	I32 i = SvIVX (sv);
		886	U32 u;
		887	char digit, nz = 0;
		888
439	need (enc, 64);	889	need (enc, 6);
		890
		891	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		892	u = i < 0 ? -i : i;
		893
		894	// convert to 4.28 fixed-point representation
		895	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		896
		897	// now output digit by digit, each time masking out the integer part
		898	// and multiplying by 5 while moving the decimal point one to the right,
		899	// resulting in a net multiplication by 10.
		900	// we always write the digit to memory but conditionally increment
		901	// the pointer, to enable the use of conditional move instructions.
		902	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		903	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		904	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		905	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		906	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		907	}
		908	else
		909	{
		910	// large integer, use the (rather slow) snprintf way.
		911	need (enc, IVUV_MAXCHARS);
440	enc->cur +=	912	enc->cur +=
441	SvIsUV(sv)	913	SvIsUV(sv)
442	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	914	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
443	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	915	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		916	}
444	}	917	}
445	else if (SvROK (sv))	918	else if (SvROK (sv))
446	{	919	encode_rv (enc, SvRV (sv));
447	SV *rv = SvRV (sv);	920	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
448
449	if (enc->indent >= enc->max_depth)
450	croak ("data structure too deep (hit recursion limit)");
451
452	switch (SvTYPE (rv))
453	{
454	case SVt_PVAV: encode_av (enc, (AV *)rv); break;
455	case SVt_PVHV: encode_hv (enc, (HV *)rv); break;
456
457	default:
458	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
459	SvPV_nolen (sv));
460	}
461	}
462	else if (!SvOK (sv))
463	encode_str (enc, "null", 4, 0);	921	encode_str (enc, "null", 4, 0);
464	else	922	else
465	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	923	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
466	SvPV_nolen (sv), SvFLAGS (sv));	924	SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
467	}	925	}
468		926
469	static SV *	927	static SV *
470	encode_json (SV *scalar, UV flags)	928	encode_json (SV *scalar, JSON *json)
471	{	929	{
		930	enc_t enc;
		931
472	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	932	if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar))
473	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	933	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
474		934
475	enc_t enc;	935	enc.json = *json;
476	enc.flags = flags;
477	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	936	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
478	enc.cur = SvPVX (enc.sv);	937	enc.cur = SvPVX (enc.sv);
479	enc.end = SvEND (enc.sv);	938	enc.end = SvEND (enc.sv);
480	enc.indent = 0;	939	enc.indent = 0;
481	enc.max_depth = 0x7fffffffUL;	940	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		941	: enc.json.flags & F_LATIN1 ? 0x000100UL
		942	: 0x110000UL;
482		943
483	SvPOK_only (enc.sv);	944	SvPOK_only (enc.sv);
484	encode_sv (&enc, scalar);	945	encode_sv (&enc, scalar);
		946	encode_nl (&enc);
485		947
		948	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		949	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		950
486	if (!(flags & (F_ASCII | F_UTF8)))	951	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
487	SvUTF8_on (enc.sv);	952	SvUTF8_on (enc.sv);
488		953
489	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
490
491	if (enc.flags & F_SHRINK)	954	if (enc.json.flags & F_SHRINK)
492	shrink (enc.sv);	955	shrink (enc.sv);
493		956
494	return enc.sv;	957	return enc.sv;
495	}	958	}
496		959
…		…
501	typedef struct	964	typedef struct
502	{	965	{
503	char *cur; // current parser pointer	966	char *cur; // current parser pointer
504	char *end; // end of input string	967	char *end; // end of input string
505	const char *err; // parse error, if != 0	968	const char *err; // parse error, if != 0
506	UV flags; // F_*	969	JSON json;
		970	U32 depth; // recursion depth
		971	U32 maxdepth; // recursion depth limit
507	} dec_t;	972	} dec_t;
508		973
509	static void	974	INLINE void
		975	decode_comment (dec_t *dec)
		976	{
		977	// only '#'-style comments allowed a.t.m.
		978
		979	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		980	++dec->cur;
		981	}
		982
		983	INLINE void
510	decode_ws (dec_t *dec)	984	decode_ws (dec_t *dec)
511	{	985	{
512	for (;;)	986	for (;;)
513	{	987	{
514	char ch = *dec->cur;	988	char ch = *dec->cur;
515		989
516	if (ch > 0x20	990	if (ch > 0x20)
		991	{
		992	if (expect_false (ch == '#'))
		993	{
		994	if (dec->json.flags & F_RELAXED)
		995	decode_comment (dec);
		996	else
		997	break;
		998	}
		999	else
		1000	break;
		1001	}
517	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	1002	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
518	break;	1003	break; // parse error, but let higher level handle it, gives better error messages
519		1004
520	++dec->cur;	1005	++dec->cur;
521	}	1006	}
522	}	1007	}
523		1008
524	#define ERR(reason) SB dec->err = reason; goto fail; SE	1009	#define ERR(reason) SB dec->err = reason; goto fail; SE
		1010
525	#define EXPECT_CH(ch) SB \	1011	#define EXPECT_CH(ch) SB \
526	if (*dec->cur != ch) \	1012	if (*dec->cur != ch) \
527	ERR (# ch " expected"); \	1013	ERR (# ch " expected"); \
528	++dec->cur; \	1014	++dec->cur; \
529	SE	1015	SE
530		1016
		1017	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
		1018	#define DEC_DEC_DEPTH --dec->depth
		1019
531	static SV *decode_sv (dec_t *dec);	1020	static SV *decode_sv (dec_t *dec);
532		1021
533	static signed char decode_hexdigit[256];	1022	static signed char decode_hexdigit[256];
534		1023
535	static UV	1024	static UV
536	decode_4hex (dec_t *dec)	1025	decode_4hex (dec_t *dec)
537	{	1026	{
538	signed char d1, d2, d3, d4;	1027	signed char d1, d2, d3, d4;
539	unsigned char *cur = (unsigned char *)dec->cur;	1028	unsigned char *cur = (unsigned char *)dec->cur;
540		1029
541	d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected");	1030	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
542	d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected");	1031	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
543	d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected");	1032	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
544	d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected");	1033	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
545		1034
546	dec->cur += 4;	1035	dec->cur += 4;
547		1036
548	return ((UV)d1) << 12	1037	return ((UV)d1) << 12
549	| ((UV)d2) << 8	1038	| ((UV)d2) << 8
…		…
557	static SV *	1046	static SV *
558	decode_str (dec_t *dec)	1047	decode_str (dec_t *dec)
559	{	1048	{
560	SV *sv = 0;	1049	SV *sv = 0;
561	int utf8 = 0;	1050	int utf8 = 0;
		1051	char *dec_cur = dec->cur;
562		1052
563	do	1053	do
564	{	1054	{
565	char buf [SHORT_STRING_LEN + UTF8_MAX_LEN];	1055	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
566	char *cur = buf;	1056	char *cur = buf;
567		1057
568	do	1058	do
569	{	1059	{
570	unsigned char ch = *(unsigned char *)dec->cur++;	1060	unsigned char ch = *(unsigned char *)dec_cur++;
571		1061
572	if (ch == '"')	1062	if (expect_false (ch == '"'))
573	{	1063	{
574	--dec->cur;	1064	--dec_cur;
575	break;	1065	break;
576	}	1066	}
577	else if (ch == '\\')	1067	else if (expect_false (ch == '\\'))
578	{	1068	{
579	switch (*dec->cur)	1069	switch (*dec_cur)
580	{	1070	{
581	case '\\':	1071	case '\\':
582	case '/':	1072	case '/':
583	case '"': *cur++ = *dec->cur++; break;	1073	case '"': *cur++ = *dec_cur++; break;
584		1074
585	case 'b': ++dec->cur; *cur++ = '\010'; break;	1075	case 'b': ++dec_cur; *cur++ = '\010'; break;
586	case 't': ++dec->cur; *cur++ = '\011'; break;	1076	case 't': ++dec_cur; *cur++ = '\011'; break;
587	case 'n': ++dec->cur; *cur++ = '\012'; break;	1077	case 'n': ++dec_cur; *cur++ = '\012'; break;
588	case 'f': ++dec->cur; *cur++ = '\014'; break;	1078	case 'f': ++dec_cur; *cur++ = '\014'; break;
589	case 'r': ++dec->cur; *cur++ = '\015'; break;	1079	case 'r': ++dec_cur; *cur++ = '\015'; break;
590		1080
591	case 'u':	1081	case 'u':
592	{	1082	{
593	UV lo, hi;	1083	UV lo, hi;
594	++dec->cur;	1084	++dec_cur;
595		1085
		1086	dec->cur = dec_cur;
596	hi = decode_4hex (dec);	1087	hi = decode_4hex (dec);
		1088	dec_cur = dec->cur;
597	if (hi == (UV)-1)	1089	if (hi == (UV)-1)
598	goto fail;	1090	goto fail;
599		1091
600	// possibly a surrogate pair	1092	// possibly a surrogate pair
601	if (hi >= 0xd800)	1093	if (hi >= 0xd800)
602	if (hi < 0xdc00)	1094	if (hi < 0xdc00)
603	{	1095	{
604	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	1096	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
605	ERR ("missing low surrogate character in surrogate pair");	1097	ERR ("missing low surrogate character in surrogate pair");
606		1098
607	dec->cur += 2;	1099	dec_cur += 2;
608		1100
		1101	dec->cur = dec_cur;
609	lo = decode_4hex (dec);	1102	lo = decode_4hex (dec);
		1103	dec_cur = dec->cur;
610	if (lo == (UV)-1)	1104	if (lo == (UV)-1)
611	goto fail;	1105	goto fail;
612		1106
613	if (lo < 0xdc00 || lo >= 0xe000)	1107	if (lo < 0xdc00 || lo >= 0xe000)
614	ERR ("surrogate pair expected");	1108	ERR ("surrogate pair expected");
…		…
620		1114
621	if (hi >= 0x80)	1115	if (hi >= 0x80)
622	{	1116	{
623	utf8 = 1;	1117	utf8 = 1;
624		1118
625	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	1119	cur = encode_utf8 (cur, hi);
626	}	1120	}
627	else	1121	else
628	*cur++ = hi;	1122	*cur++ = hi;
629	}	1123	}
630	break;	1124	break;
631		1125
632	default:	1126	default:
633	--dec->cur;	1127	--dec_cur;
634	ERR ("illegal backslash escape sequence in string");	1128	ERR ("illegal backslash escape sequence in string");
635	}	1129	}
636	}	1130	}
637	else if (ch >= 0x20 && ch <= 0x7f)	1131	else if (expect_true (ch >= 0x20 && ch < 0x80))
638	*cur++ = ch;	1132	*cur++ = ch;
639	else if (ch >= 0x80)	1133	else if (ch >= 0x80)
640	{	1134	{
641	--dec->cur;
642
643	STRLEN clen;	1135	STRLEN clen;
		1136
		1137	--dec_cur;
		1138
644	UV uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	1139	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
645	if (clen == (STRLEN)-1)	1140	if (clen == (STRLEN)-1)
646	ERR ("malformed UTF-8 character in JSON string");	1141	ERR ("malformed UTF-8 character in JSON string");
647		1142
648	do	1143	do
649	{
650	*cur++ = *dec->cur++;	1144	*cur++ = *dec_cur++;
651	}
652	while (--clen);	1145	while (--clen);
653		1146
654	utf8 = 1;	1147	utf8 = 1;
655	}	1148	}
656	else if (!ch)	1149	else if (ch == '\t' && dec->json.flags & F_RELAXED)
657	ERR ("unexpected end of string while parsing json string");	1150	*cur++ = ch;
658	else	1151	else
		1152	{
		1153	--dec_cur;
		1154
		1155	if (!ch)
		1156	ERR ("unexpected end of string while parsing JSON string");
		1157	else
659	ERR ("invalid character encountered");	1158	ERR ("invalid character encountered while parsing JSON string");
660		1159	}
661	}	1160	}
662	while (cur < buf + SHORT_STRING_LEN);	1161	while (cur < buf + SHORT_STRING_LEN);
663		1162
		1163	{
664	STRLEN len = cur - buf;	1164	STRLEN len = cur - buf;
665		1165
666	if (sv)	1166	if (sv)
667	{	1167	{
668	SvGROW (sv, SvCUR (sv) + len + 1);	1168	STRLEN cur = SvCUR (sv);
		1169
		1170	if (SvLEN (sv) <= cur + len)
		1171	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1172
669	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	1173	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
670	SvCUR_set (sv, SvCUR (sv) + len);	1174	SvCUR_set (sv, SvCUR (sv) + len);
671	}	1175	}
672	else	1176	else
673	sv = newSVpvn (buf, len);	1177	sv = newSVpvn (buf, len);
674	}	1178	}
		1179	}
675	while (*dec->cur != '"');	1180	while (*dec_cur != '"');
676		1181
677	++dec->cur;	1182	++dec_cur;
678		1183
679	if (sv)	1184	if (sv)
680	{	1185	{
681	SvPOK_only (sv);	1186	SvPOK_only (sv);
682	*SvEND (sv) = 0;	1187	*SvEND (sv) = 0;
…		…
685	SvUTF8_on (sv);	1190	SvUTF8_on (sv);
686	}	1191	}
687	else	1192	else
688	sv = newSVpvn ("", 0);	1193	sv = newSVpvn ("", 0);
689		1194
		1195	dec->cur = dec_cur;
690	return sv;	1196	return sv;
691		1197
692	fail:	1198	fail:
		1199	dec->cur = dec_cur;
693	return 0;	1200	return 0;
694	}	1201	}
695		1202
696	static SV *	1203	static SV *
697	decode_num (dec_t *dec)	1204	decode_num (dec_t *dec)
…		…
755	is_nv = 1;	1262	is_nv = 1;
756	}	1263	}
757		1264
758	if (!is_nv)	1265	if (!is_nv)
759	{	1266	{
760	UV uv;	1267	int len = dec->cur - start;
761	int numtype = grok_number (start, dec->cur - start, &uv);	1268
762	if (numtype & IS_NUMBER_IN_UV)	1269	// special case the rather common 1..5-digit-int case
763	if (numtype & IS_NUMBER_NEG)	1270	if (*start == '-')
		1271	switch (len)
764	{	1272	{
765	if (uv < (UV)IV_MIN)	1273	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
766	return newSViv (-(IV)uv);	1274	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
		1275	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1276	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1277	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
767	}	1278	}
		1279	else
		1280	switch (len)
		1281	{
		1282	case 1: return newSViv ( start [0] - '0' * 1);
		1283	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1284	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1285	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1286	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1287	}
		1288
		1289	{
		1290	UV uv;
		1291	int numtype = grok_number (start, len, &uv);
		1292	if (numtype & IS_NUMBER_IN_UV)
		1293	if (numtype & IS_NUMBER_NEG)
		1294	{
		1295	if (uv < (UV)IV_MIN)
		1296	return newSViv (-(IV)uv);
		1297	}
768	else	1298	else
769	return newSVuv (uv);	1299	return newSVuv (uv);
770	}	1300	}
771		1301
		1302	len -= *start == '-' ? 1 : 0;
		1303
		1304	// does not fit into IV or UV, try NV
		1305	if (len <= NV_DIG)
		1306	// fits into NV without loss of precision
		1307	return newSVnv (json_atof (start));
		1308
		1309	// everything else fails, convert it to a string
		1310	return newSVpvn (start, dec->cur - start);
		1311	}
		1312
		1313	// loss of precision here
772	return newSVnv (Atof (start));	1314	return newSVnv (json_atof (start));
773		1315
774	fail:	1316	fail:
775	return 0;	1317	return 0;
776	}	1318	}
777		1319
778	static SV *	1320	static SV *
779	decode_av (dec_t *dec)	1321	decode_av (dec_t *dec)
780	{	1322	{
781	AV *av = newAV ();	1323	AV *av = newAV ();
782		1324
		1325	DEC_INC_DEPTH;
783	decode_ws (dec);	1326	decode_ws (dec);
		1327
784	if (*dec->cur == ']')	1328	if (*dec->cur == ']')
785	++dec->cur;	1329	++dec->cur;
786	else	1330	else
787	for (;;)	1331	for (;;)
788	{	1332	{
…		…
799	if (*dec->cur == ']')	1343	if (*dec->cur == ']')
800	{	1344	{
801	++dec->cur;	1345	++dec->cur;
802	break;	1346	break;
803	}	1347	}
804		1348
805	if (*dec->cur != ',')	1349	if (*dec->cur != ',')
806	ERR (", or ] expected while parsing array");	1350	ERR (", or ] expected while parsing array");
807		1351
808	++dec->cur;	1352	++dec->cur;
		1353
		1354	decode_ws (dec);
		1355
		1356	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1357	{
		1358	++dec->cur;
		1359	break;
		1360	}
809	}	1361	}
810		1362
		1363	DEC_DEC_DEPTH;
811	return newRV_noinc ((SV *)av);	1364	return newRV_noinc ((SV *)av);
812		1365
813	fail:	1366	fail:
814	SvREFCNT_dec (av);	1367	SvREFCNT_dec (av);
		1368	DEC_DEC_DEPTH;
815	return 0;	1369	return 0;
816	}	1370	}
817		1371
818	static SV *	1372	static SV *
819	decode_hv (dec_t *dec)	1373	decode_hv (dec_t *dec)
820	{	1374	{
		1375	SV *sv;
821	HV *hv = newHV ();	1376	HV *hv = newHV ();
822		1377
		1378	DEC_INC_DEPTH;
823	decode_ws (dec);	1379	decode_ws (dec);
		1380
824	if (*dec->cur == '}')	1381	if (*dec->cur == '}')
825	++dec->cur;	1382	++dec->cur;
826	else	1383	else
827	for (;;)	1384	for (;;)
828	{	1385	{
		1386	EXPECT_CH ('"');
		1387
		1388	// heuristic: assume that
		1389	// a) decode_str + hv_store_ent are abysmally slow.
		1390	// b) most hash keys are short, simple ascii text.
		1391	// => try to "fast-match" such strings to avoid
		1392	// the overhead of decode_str + hv_store_ent.
		1393	{
829	SV *key, *value;	1394	SV *value;
		1395	char *p = dec->cur;
		1396	char *e = p + 24; // only try up to 24 bytes
830		1397
831	decode_ws (dec); EXPECT_CH ('"');	1398	for (;;)
832
833	key = decode_str (dec);
834	if (!key)
835	goto fail;
836
837	decode_ws (dec); EXPECT_CH (':');
838
839	value = decode_sv (dec);
840	if (!value)
841	{	1399	{
		1400	// the >= 0x80 is false on most architectures
		1401	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1402	{
		1403	// slow path, back up and use decode_str
		1404	SV *key = decode_str (dec);
		1405	if (!key)
		1406	goto fail;
		1407
		1408	decode_ws (dec); EXPECT_CH (':');
		1409
		1410	decode_ws (dec);
		1411	value = decode_sv (dec);
		1412	if (!value)
		1413	{
		1414	SvREFCNT_dec (key);
		1415	goto fail;
		1416	}
		1417
		1418	hv_store_ent (hv, key, value, 0);
842	SvREFCNT_dec (key);	1419	SvREFCNT_dec (key);
		1420
		1421	break;
		1422	}
		1423	else if (*p == '"')
		1424	{
		1425	// fast path, got a simple key
		1426	char *key = dec->cur;
		1427	int len = p - key;
		1428	dec->cur = p + 1;
		1429
		1430	decode_ws (dec); EXPECT_CH (':');
		1431
		1432	decode_ws (dec);
		1433	value = decode_sv (dec);
		1434	if (!value)
843	goto fail;	1435	goto fail;
		1436
		1437	hv_store (hv, key, len, value, 0);
		1438
		1439	break;
		1440	}
		1441
		1442	++p;
844	}	1443	}
845		1444	}
846	//TODO: optimise
847	hv_store_ent (hv, key, value, 0);
848		1445
849	decode_ws (dec);	1446	decode_ws (dec);
850		1447
851	if (*dec->cur == '}')	1448	if (*dec->cur == '}')
852	{	1449	{
…		…
856		1453
857	if (*dec->cur != ',')	1454	if (*dec->cur != ',')
858	ERR (", or } expected while parsing object/hash");	1455	ERR (", or } expected while parsing object/hash");
859		1456
860	++dec->cur;	1457	++dec->cur;
		1458
		1459	decode_ws (dec);
		1460
		1461	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1462	{
		1463	++dec->cur;
		1464	break;
		1465	}
861	}	1466	}
862		1467
		1468	DEC_DEC_DEPTH;
863	return newRV_noinc ((SV *)hv);	1469	sv = newRV_noinc ((SV *)hv);
		1470
		1471	// check filter callbacks
		1472	if (dec->json.flags & F_HOOK)
		1473	{
		1474	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1475	{
		1476	HE *cb, *he;
		1477
		1478	hv_iterinit (hv);
		1479	he = hv_iternext (hv);
		1480	hv_iterinit (hv);
		1481
		1482	// the next line creates a mortal sv each time it's called.
		1483	// might want to optimise this for common cases.
		1484	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1485
		1486	if (cb)
		1487	{
		1488	dSP;
		1489	int count;
		1490
		1491	ENTER; SAVETMPS;
		1492	SAVESTACK_POS ();
		1493	PUSHMARK (SP);
		1494	XPUSHs (HeVAL (he));
		1495	sv_2mortal (sv);
		1496
		1497	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1498
		1499	if (count == 1)
		1500	{
		1501	sv = newSVsv (POPs);
		1502	FREETMPS; LEAVE;
		1503	return sv;
		1504	}
		1505
		1506	SvREFCNT_inc (sv);
		1507	FREETMPS; LEAVE;
		1508	}
		1509	}
		1510
		1511	if (dec->json.cb_object)
		1512	{
		1513	dSP;
		1514	int count;
		1515
		1516	ENTER; SAVETMPS;
		1517	SAVESTACK_POS ();
		1518	PUSHMARK (SP);
		1519	XPUSHs (sv_2mortal (sv));
		1520
		1521	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1522
		1523	if (count == 1)
		1524	{
		1525	sv = newSVsv (POPs);
		1526	FREETMPS; LEAVE;
		1527	return sv;
		1528	}
		1529
		1530	SvREFCNT_inc (sv);
		1531	FREETMPS; LEAVE;
		1532	}
		1533	}
		1534
		1535	return sv;
864		1536
865	fail:	1537	fail:
866	SvREFCNT_dec (hv);	1538	SvREFCNT_dec (hv);
		1539	DEC_DEC_DEPTH;
		1540	return 0;
		1541	}
		1542
		1543	static SV *
		1544	decode_tag (dec_t *dec)
		1545	{
		1546	SV *tag = 0;
		1547	SV *val = 0;
		1548
		1549	if (!(dec->json.flags & F_ALLOW_TAGS))
		1550	ERR ("malformed JSON string, neither array, object, number, string or atom");
		1551
		1552	++dec->cur;
		1553
		1554	decode_ws (dec);
		1555
		1556	tag = decode_sv (dec);
		1557	if (!tag)
		1558	goto fail;
		1559
		1560	if (!SvPOK (tag))
		1561	ERR ("malformed JSON string, (tag) must be a string");
		1562
		1563	decode_ws (dec);
		1564
		1565	if (*dec->cur != ')')
		1566	ERR (") expected after tag");
		1567
		1568	++dec->cur;
		1569
		1570	decode_ws (dec);
		1571
		1572	val = decode_sv (dec);
		1573	if (!val)
		1574	goto fail;
		1575
		1576	if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
		1577	ERR ("malformed JSON string, tag value must be an array");
		1578
		1579	{
		1580	AV *av = (AV *)SvRV (val);
		1581	int i, len = av_len (av) + 1;
		1582	HV *stash = gv_stashsv (tag, 0);
		1583	SV *sv;
		1584
		1585	if (!stash)
		1586	ERR ("cannot decode perl-object (package does not exist)");
		1587
		1588	GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
		1589
		1590	if (!method)
		1591	ERR ("cannot decode perl-object (package does not have a THAW method)");
		1592
		1593	dSP;
		1594
		1595	ENTER; SAVETMPS;
		1596	PUSHMARK (SP);
		1597	EXTEND (SP, len + 2);
		1598	// we re-bless the reference to get overload and other niceties right
		1599	PUSHs (tag);
		1600	PUSHs (sv_json);
		1601
		1602	for (i = 0; i < len; ++i)
		1603	PUSHs (*av_fetch (av, i, 1));
		1604
		1605	PUTBACK;
		1606	call_sv ((SV *)GvCV (method), G_SCALAR);
		1607	SPAGAIN;
		1608
		1609	SvREFCNT_dec (tag);
		1610	SvREFCNT_dec (val);
		1611	sv = SvREFCNT_inc (POPs);
		1612
		1613	PUTBACK;
		1614
		1615	FREETMPS; LEAVE;
		1616
		1617	return sv;
		1618	}
		1619
		1620	fail:
		1621	SvREFCNT_dec (tag);
		1622	SvREFCNT_dec (val);
867	return 0;	1623	return 0;
868	}	1624	}
869		1625
870	static SV *	1626	static SV *
871	decode_sv (dec_t *dec)	1627	decode_sv (dec_t *dec)
872	{	1628	{
873	decode_ws (dec);	1629	// the beauty of JSON: you need exactly one character lookahead
		1630	// to parse everything.
874	switch (*dec->cur)	1631	switch (*dec->cur)
875	{	1632	{
876	case '"': ++dec->cur; return decode_str (dec);	1633	case '"': ++dec->cur; return decode_str (dec);
877	case '[': ++dec->cur; return decode_av (dec);	1634	case '[': ++dec->cur; return decode_av (dec);
878	case '{': ++dec->cur; return decode_hv (dec);	1635	case '{': ++dec->cur; return decode_hv (dec);
		1636	case '(': return decode_tag (dec);
879		1637
880	case '-':	1638	case '-':
881	case '0': case '1': case '2': case '3': case '4':	1639	case '0': case '1': case '2': case '3': case '4':
882	case '5': case '6': case '7': case '8': case '9':	1640	case '5': case '6': case '7': case '8': case '9':
883	return decode_num (dec);	1641	return decode_num (dec);
884		1642
885	case 't':	1643	case 't':
886	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1644	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
887	{	1645	{
888	dec->cur += 4;	1646	dec->cur += 4;
		1647	#if JSON_SLOW
		1648	bool_true = get_bool ("Types::Serialiser::true");
		1649	#endif
889	return newSViv (1);	1650	return newSVsv (bool_true);
890	}	1651	}
891	else	1652	else
892	ERR ("'true' expected");	1653	ERR ("'true' expected");
893		1654
894	break;	1655	break;
895		1656
896	case 'f':	1657	case 'f':
897	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1658	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
898	{	1659	{
899	dec->cur += 5;	1660	dec->cur += 5;
		1661	#if JSON_SLOW
		1662	bool_false = get_bool ("Types::Serialiser::false");
		1663	#endif
900	return newSViv (0);	1664	return newSVsv (bool_false);
901	}	1665	}
902	else	1666	else
903	ERR ("'false' expected");	1667	ERR ("'false' expected");
904		1668
905	break;	1669	break;
…		…
914	ERR ("'null' expected");	1678	ERR ("'null' expected");
915		1679
916	break;	1680	break;
917		1681
918	default:	1682	default:
919	ERR ("malformed json string, neither array, object, number, string or atom");	1683	ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
920	break;	1684	break;
921	}	1685	}
922		1686
923	fail:	1687	fail:
924	return 0;	1688	return 0;
925	}	1689	}
926		1690
927	static SV *	1691	static SV *
928	decode_json (SV *string, UV flags)	1692	decode_json (SV *string, JSON *json, char **offset_return)
929	{	1693	{
		1694	dec_t dec;
930	SV *sv;	1695	SV *sv;
931		1696
932	if (flags & F_UTF8)	1697	/* work around bugs in 5.10 where manipulating magic values
		1698	* makes perl ignore the magic in subsequent accesses.
		1699	* also make a copy of non-PV values, to get them into a clean
		1700	* state (SvPV should do that, but it's buggy, see below).
		1701	*/
		1702	/*SvGETMAGIC (string);*/
		1703	if (SvMAGICAL (string) || !SvPOK (string))
		1704	string = sv_2mortal (newSVsv (string));
		1705
		1706	SvUPGRADE (string, SVt_PV);
		1707
		1708	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1709	* assertion with -DDEBUGGING, although SvCUR is documented to
		1710	* return the xpv_cur field which certainly exists after upgrading.
		1711	* according to nicholas clark, calling SvPOK fixes this.
		1712	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1713	* and hope for the best.
		1714	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1715	* assertion business is seriously broken, try yet another workaround
		1716	* for the broken -DDEBUGGING.
		1717	*/
		1718	{
		1719	#ifdef DEBUGGING
		1720	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1721	#else
		1722	STRLEN offset = SvCUR (string);
		1723	#endif
		1724
		1725	if (offset > json->max_size && json->max_size)
		1726	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1727	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1728	}
		1729
		1730	if (DECODE_WANTS_OCTETS (json))
933	sv_utf8_downgrade (string, 0);	1731	sv_utf8_downgrade (string, 0);
934	else	1732	else
935	sv_utf8_upgrade (string);	1733	sv_utf8_upgrade (string);
936		1734
937	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1735	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
938		1736
939	dec_t dec;	1737	dec.json = *json;
940	dec.flags = flags;
941	dec.cur = SvPVX (string);	1738	dec.cur = SvPVX (string);
942	dec.end = SvEND (string);	1739	dec.end = SvEND (string);
943	dec.err = 0;	1740	dec.err = 0;
		1741	dec.depth = 0;
944		1742
		1743	if (dec.json.cb_object || dec.json.cb_sk_object)
		1744	dec.json.flags |= F_HOOK;
		1745
		1746	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1747
		1748	decode_ws (&dec);
945	sv = decode_sv (&dec);	1749	sv = decode_sv (&dec);
946		1750
		1751	if (offset_return)
		1752	*offset_return = dec.cur;
		1753
		1754	if (!(offset_return || !sv))
		1755	{
		1756	// check for trailing garbage
		1757	decode_ws (&dec);
		1758
		1759	if (*dec.cur)
		1760	{
		1761	dec.err = "garbage after JSON object";
		1762	SvREFCNT_dec (sv);
		1763	sv = 0;
		1764	}
		1765	}
		1766
947	if (!sv)	1767	if (!sv)
948	{	1768	{
949	IV offset = dec.flags & F_UTF8
950	? dec.cur - SvPVX (string)
951	: utf8_distance (dec.cur, SvPVX (string));
952	SV *uni = sv_newmortal ();	1769	SV *uni = sv_newmortal ();
953		1770
954	// horrible hack to silence warning inside pv_uni_display	1771	// horrible hack to silence warning inside pv_uni_display
955	COP cop = *PL_curcop;	1772	COP cop = *PL_curcop;
956	cop.cop_warnings = pWARN_NONE;	1773	cop.cop_warnings = pWARN_NONE;
…		…
958	SAVEVPTR (PL_curcop);	1775	SAVEVPTR (PL_curcop);
959	PL_curcop = &cop;	1776	PL_curcop = &cop;
960	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1777	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
961	LEAVE;	1778	LEAVE;
962		1779
963	croak ("%s, at character offset %d (%s)",	1780	croak ("%s, at character offset %d (before \"%s\")",
964	dec.err,	1781	dec.err,
965	(int)offset,	1782	(int)ptr_to_index (string, dec.cur),
966	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1783	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
967	}	1784	}
968		1785
969	sv = sv_2mortal (sv);	1786	sv = sv_2mortal (sv);
970		1787
971	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1788	if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv))
972	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1789	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
973		1790
974	return sv;	1791	return sv;
		1792	}
		1793
		1794	/////////////////////////////////////////////////////////////////////////////
		1795	// incremental parser
		1796
		1797	static void
		1798	incr_parse (JSON *self)
		1799	{
		1800	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1801
		1802	// the state machine here is a bit convoluted and could be simplified a lot
		1803	// but this would make it slower, so...
		1804
		1805	for (;;)
		1806	{
		1807	//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
		1808	switch (self->incr_mode)
		1809	{
		1810	// only used for initial whitespace skipping
		1811	case INCR_M_WS:
		1812	for (;;)
		1813	{
		1814	if (*p > 0x20)
		1815	{
		1816	if (*p == '#')
		1817	{
		1818	self->incr_mode = INCR_M_C0;
		1819	goto incr_m_c;
		1820	}
		1821	else
		1822	{
		1823	self->incr_mode = INCR_M_JSON;
		1824	goto incr_m_json;
		1825	}
		1826	}
		1827	else if (!*p)
		1828	goto interrupt;
		1829
		1830	++p;
		1831	}
		1832
		1833	// skip a single char inside a string (for \\-processing)
		1834	case INCR_M_BS:
		1835	if (!*p)
		1836	goto interrupt;
		1837
		1838	++p;
		1839	self->incr_mode = INCR_M_STR;
		1840	goto incr_m_str;
		1841
		1842	// inside #-style comments
		1843	case INCR_M_C0:
		1844	case INCR_M_C1:
		1845	incr_m_c:
		1846	for (;;)
		1847	{
		1848	if (*p == '\n')
		1849	{
		1850	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1851	break;
		1852	}
		1853	else if (!*p)
		1854	goto interrupt;
		1855
		1856	++p;
		1857	}
		1858
		1859	break;
		1860
		1861	// inside a string
		1862	case INCR_M_STR:
		1863	incr_m_str:
		1864	for (;;)
		1865	{
		1866	if (*p == '"')
		1867	{
		1868	++p;
		1869	self->incr_mode = INCR_M_JSON;
		1870
		1871	if (!self->incr_nest)
		1872	goto interrupt;
		1873
		1874	goto incr_m_json;
		1875	}
		1876	else if (*p == '\\')
		1877	{
		1878	++p; // "virtually" consumes character after \
		1879
		1880	if (!*p) // if at end of string we have to switch modes
		1881	{
		1882	self->incr_mode = INCR_M_BS;
		1883	goto interrupt;
		1884	}
		1885	}
		1886	else if (!*p)
		1887	goto interrupt;
		1888
		1889	++p;
		1890	}
		1891
		1892	// after initial ws, outside string
		1893	case INCR_M_JSON:
		1894	incr_m_json:
		1895	for (;;)
		1896	{
		1897	switch (*p++)
		1898	{
		1899	case 0:
		1900	--p;
		1901	goto interrupt;
		1902
		1903	case 0x09:
		1904	case 0x0a:
		1905	case 0x0d:
		1906	case 0x20:
		1907	if (!self->incr_nest)
		1908	{
		1909	--p; // do not eat the whitespace, let the next round do it
		1910	goto interrupt;
		1911	}
		1912	break;
		1913
		1914	case '"':
		1915	self->incr_mode = INCR_M_STR;
		1916	goto incr_m_str;
		1917
		1918	case '[':
		1919	case '{':
		1920	case '(':
		1921	if (++self->incr_nest > self->max_depth)
		1922	croak (ERR_NESTING_EXCEEDED);
		1923	break;
		1924
		1925	case ']':
		1926	case '}':
		1927	if (--self->incr_nest <= 0)
		1928	goto interrupt;
		1929	break;
		1930
		1931	case ')':
		1932	--self->incr_nest;
		1933	break;
		1934
		1935	case '#':
		1936	self->incr_mode = INCR_M_C1;
		1937	goto incr_m_c;
		1938	}
		1939	}
		1940	}
		1941
		1942	modechange:
		1943	;
		1944	}
		1945
		1946	interrupt:
		1947	self->incr_pos = p - SvPVX (self->incr_text);
		1948	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1949	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
975	}	1950	}
976		1951
977	/////////////////////////////////////////////////////////////////////////////	1952	/////////////////////////////////////////////////////////////////////////////
978	// XS interface functions	1953	// XS interface functions
979		1954
…		…
981		1956
982	BOOT:	1957	BOOT:
983	{	1958	{
984	int i;	1959	int i;
985		1960
986	memset (decode_hexdigit, 0xff, 256);
987	for (i = 10; i--; )	1961	for (i = 0; i < 256; ++i)
988	decode_hexdigit ['0' + i] = i;	1962	decode_hexdigit [i] =
		1963	i >= '0' && i <= '9' ? i - '0'
		1964	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1965	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1966	: -1;
989		1967
990	for (i = 7; i--; )
991	{
992	decode_hexdigit ['a' + i] = 10 + i;
993	decode_hexdigit ['A' + i] = 10 + i;
994	}
995
996	json_stash = gv_stashpv ("JSON::XS", 1);	1968	json_stash = gv_stashpv ("JSON::XS" , 1);
		1969	bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
		1970	bool_true = get_bool ("Types::Serialiser::true");
		1971	bool_false = get_bool ("Types::Serialiser::false");
		1972
		1973	sv_json = newSVpv ("JSON", 0);
		1974	SvREADONLY_on (sv_json);
		1975
		1976	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
997	}	1977	}
998		1978
999	PROTOTYPES: DISABLE	1979	PROTOTYPES: DISABLE
1000		1980
1001	SV *new (char *dummy)	1981	void CLONE (...)
1002	CODE:	1982	CODE:
1003	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1983	json_stash = 0;
		1984	bool_stash = 0;
		1985
		1986	void new (char *klass)
		1987	PPCODE:
		1988	{
		1989	SV *pv = NEWSV (0, sizeof (JSON));
		1990	SvPOK_only (pv);
		1991	json_init ((JSON *)SvPVX (pv));
		1992	XPUSHs (sv_2mortal (sv_bless (
		1993	newRV_noinc (pv),
		1994	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1995	)));
		1996	}
		1997
		1998	void ascii (JSON *self, int enable = 1)
		1999	ALIAS:
		2000	ascii = F_ASCII
		2001	latin1 = F_LATIN1
		2002	utf8 = F_UTF8
		2003	indent = F_INDENT
		2004	canonical = F_CANONICAL
		2005	space_before = F_SPACE_BEFORE
		2006	space_after = F_SPACE_AFTER
		2007	pretty = F_PRETTY
		2008	allow_nonref = F_ALLOW_NONREF
		2009	shrink = F_SHRINK
		2010	allow_blessed = F_ALLOW_BLESSED
		2011	convert_blessed = F_CONV_BLESSED
		2012	relaxed = F_RELAXED
		2013	allow_unknown = F_ALLOW_UNKNOWN
		2014	allow_tags = F_ALLOW_TAGS
		2015	PPCODE:
		2016	{
		2017	if (enable)
		2018	self->flags |= ix;
		2019	else
		2020	self->flags &= ~ix;
		2021
		2022	XPUSHs (ST (0));
		2023	}
		2024
		2025	void get_ascii (JSON *self)
		2026	ALIAS:
		2027	get_ascii = F_ASCII
		2028	get_latin1 = F_LATIN1
		2029	get_utf8 = F_UTF8
		2030	get_indent = F_INDENT
		2031	get_canonical = F_CANONICAL
		2032	get_space_before = F_SPACE_BEFORE
		2033	get_space_after = F_SPACE_AFTER
		2034	get_allow_nonref = F_ALLOW_NONREF
		2035	get_shrink = F_SHRINK
		2036	get_allow_blessed = F_ALLOW_BLESSED
		2037	get_convert_blessed = F_CONV_BLESSED
		2038	get_relaxed = F_RELAXED
		2039	get_allow_unknown = F_ALLOW_UNKNOWN
		2040	get_allow_tags = F_ALLOW_TAGS
		2041	PPCODE:
		2042	XPUSHs (boolSV (self->flags & ix));
		2043
		2044	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		2045	PPCODE:
		2046	self->max_depth = max_depth;
		2047	XPUSHs (ST (0));
		2048
		2049	U32 get_max_depth (JSON *self)
		2050	CODE:
		2051	RETVAL = self->max_depth;
1004	OUTPUT:	2052	OUTPUT:
1005	RETVAL	2053	RETVAL
1006		2054
1007	SV *ascii (SV *self, int enable = 1)	2055	void max_size (JSON *self, U32 max_size = 0)
1008	ALIAS:	2056	PPCODE:
1009	ascii = F_ASCII	2057	self->max_size = max_size;
1010	utf8 = F_UTF8	2058	XPUSHs (ST (0));
1011	indent = F_INDENT	2059
1012	canonical = F_CANONICAL	2060	int get_max_size (JSON *self)
1013	space_before = F_SPACE_BEFORE
1014	space_after = F_SPACE_AFTER
1015	pretty = F_PRETTY
1016	allow_nonref = F_ALLOW_NONREF
1017	shrink = F_SHRINK
1018	CODE:	2061	CODE:
1019	{	2062	RETVAL = self->max_size;
1020	UV *uv = SvJSON (self);
1021	if (enable)
1022	*uv |= ix;
1023	else
1024	*uv &= ~ix;
1025
1026	RETVAL = newSVsv (self);
1027	}
1028	OUTPUT:	2063	OUTPUT:
1029	RETVAL	2064	RETVAL
1030		2065
1031	void encode (SV *self, SV *scalar)	2066	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1032	PPCODE:	2067	PPCODE:
1033	XPUSHs (encode_json (scalar, *SvJSON (self)));	2068	{
		2069	SvREFCNT_dec (self->cb_object);
		2070	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
1034		2071
1035	void decode (SV *self, SV *jsonstr)	2072	XPUSHs (ST (0));
		2073	}
		2074
		2075	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1036	PPCODE:	2076	PPCODE:
1037	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	2077	{
		2078	if (!self->cb_sk_object)
		2079	self->cb_sk_object = newHV ();
		2080
		2081	if (SvOK (cb))
		2082	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		2083	else
		2084	{
		2085	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		2086
		2087	if (!HvKEYS (self->cb_sk_object))
		2088	{
		2089	SvREFCNT_dec (self->cb_sk_object);
		2090	self->cb_sk_object = 0;
		2091	}
		2092	}
		2093
		2094	XPUSHs (ST (0));
		2095	}
		2096
		2097	void encode (JSON *self, SV *scalar)
		2098	PPCODE:
		2099	PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
		2100	XPUSHs (scalar);
		2101
		2102	void decode (JSON *self, SV *jsonstr)
		2103	PPCODE:
		2104	PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
		2105	XPUSHs (jsonstr);
		2106
		2107	void decode_prefix (JSON *self, SV *jsonstr)
		2108	PPCODE:
		2109	{
		2110	SV *sv;
		2111	char *offset;
		2112	PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
		2113	EXTEND (SP, 2);
		2114	PUSHs (sv);
		2115	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		2116	}
		2117
		2118	void incr_parse (JSON *self, SV *jsonstr = 0)
		2119	PPCODE:
		2120	{
		2121	if (!self->incr_text)
		2122	self->incr_text = newSVpvn ("", 0);
		2123
		2124	/* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
		2125	if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
		2126	if (DECODE_WANTS_OCTETS (self))
		2127	{
		2128	if (self->incr_pos)
		2129	self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
		2130	(U8 *)SvPVX (self->incr_text) + self->incr_pos);
		2131
		2132	sv_utf8_downgrade (self->incr_text, 0);
		2133	}
		2134	else
		2135	{
		2136	sv_utf8_upgrade (self->incr_text);
		2137
		2138	if (self->incr_pos)
		2139	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		2140	- (U8 *)SvPVX (self->incr_text);
		2141	}
		2142
		2143	// append data, if any
		2144	if (jsonstr)
		2145	{
		2146	/* make sure both strings have same encoding */
		2147	if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
		2148	if (SvUTF8 (jsonstr))
		2149	sv_utf8_downgrade (jsonstr, 0);
		2150	else
		2151	sv_utf8_upgrade (jsonstr);
		2152
		2153	/* and then just blindly append */
		2154	{
		2155	STRLEN len;
		2156	const char *str = SvPV (jsonstr, len);
		2157	STRLEN cur = SvCUR (self->incr_text);
		2158
		2159	if (SvLEN (self->incr_text) <= cur + len)
		2160	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		2161
		2162	Move (str, SvEND (self->incr_text), len, char);
		2163	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		2164	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		2165	}
		2166	}
		2167
		2168	if (GIMME_V != G_VOID)
		2169	do
		2170	{
		2171	SV *sv;
		2172	char *offset;
		2173
		2174	if (!INCR_DONE (self))
		2175	{
		2176	incr_parse (self);
		2177
		2178	if (self->incr_pos > self->max_size && self->max_size)
		2179	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		2180	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		2181
		2182	if (!INCR_DONE (self))
		2183	{
		2184	// as an optimisation, do not accumulate white space in the incr buffer
		2185	if (self->incr_mode == INCR_M_WS && self->incr_pos)
		2186	{
		2187	self->incr_pos = 0;
		2188	SvCUR_set (self->incr_text, 0);
		2189	}
		2190
		2191	break;
		2192	}
		2193	}
		2194
		2195	PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
		2196	XPUSHs (sv);
		2197
		2198	self->incr_pos -= offset - SvPVX (self->incr_text);
		2199	self->incr_nest = 0;
		2200	self->incr_mode = 0;
		2201
		2202	sv_chop (self->incr_text, offset);
		2203	}
		2204	while (GIMME_V == G_ARRAY);
		2205	}
		2206
		2207	SV *incr_text (JSON *self)
		2208	ATTRS: lvalue
		2209	CODE:
		2210	{
		2211	if (self->incr_pos)
		2212	croak ("incr_text can not be called when the incremental parser already started parsing");
		2213
		2214	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		2215	}
		2216	OUTPUT:
		2217	RETVAL
		2218
		2219	void incr_skip (JSON *self)
		2220	CODE:
		2221	{
		2222	if (self->incr_pos)
		2223	{
		2224	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2225	self->incr_pos = 0;
		2226	self->incr_nest = 0;
		2227	self->incr_mode = 0;
		2228	}
		2229	}
		2230
		2231	void incr_reset (JSON *self)
		2232	CODE:
		2233	{
		2234	SvREFCNT_dec (self->incr_text);
		2235	self->incr_text = 0;
		2236	self->incr_pos = 0;
		2237	self->incr_nest = 0;
		2238	self->incr_mode = 0;
		2239	}
		2240
		2241	void DESTROY (JSON *self)
		2242	CODE:
		2243	SvREFCNT_dec (self->cb_sk_object);
		2244	SvREFCNT_dec (self->cb_object);
		2245	SvREFCNT_dec (self->incr_text);
1038		2246
1039	PROTOTYPES: ENABLE	2247	PROTOTYPES: ENABLE
1040		2248
1041	void to_json (SV *scalar)	2249	void encode_json (SV *scalar)
1042	ALIAS:
1043	objToJson = 0
1044	PPCODE:	2250	PPCODE:
1045	XPUSHs (encode_json (scalar, F_UTF8));	2251	{
		2252	JSON json;
		2253	json_init (&json);
		2254	json.flags |= F_UTF8;
		2255	PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
		2256	XPUSHs (scalar);
		2257	}
1046		2258
1047	void from_json (SV *jsonstr)	2259	void decode_json (SV *jsonstr)
1048	ALIAS:
1049	jsonToObj = 0
1050	PPCODE:	2260	PPCODE:
1051	XPUSHs (decode_json (jsonstr, F_UTF8));	2261	{
		2262	JSON json;
		2263	json_init (&json);
		2264	json.flags |= F_UTF8;
		2265	PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
		2266	XPUSHs (jsonstr);
		2267	}
1052		2268

Diff Legend

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