ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.14 by root, Sat Mar 24 22:55:43 2007 UTC vs.
Revision 1.130 by root, Sat Nov 26 06:09:29 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>
		11	#include <inttypes.h>
8		12
		13	#if defined(__BORLANDC__) || defined(_MSC_VER)
		14	# define snprintf _snprintf // C compilers have this in stdio.h
		15	#endif
		16
		17	// some old perls do not have this, try to make it work, no
		18	// guarantees, though. if it breaks, you get to keep the pieces.
		19	#ifndef UTF8_MAXBYTES
		20	# define UTF8_MAXBYTES 13
		21	#endif
		22
		23	// compatibility with perl <5.18
		24	#ifndef HvNAMELEN_get
		25	# define HvNAMELEN_get(hv) strlen (HvNAME (hv))
		26	#endif
		27	#ifndef HvNAMELEN
		28	# define HvNAMELEN(hv) HvNAMELEN_get (hv)
		29	#endif
		30	#ifndef HvNAMEUTF8
		31	# define HvNAMEUTF8(hv) 0
		32	#endif
		33
		34	// three extra for rounding, sign, and end of string
		35	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
		36
9	#define F_ASCII 0x00000001	37	#define F_ASCII 0x00000001UL
		38	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	39	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	40	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	41	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	42	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	43	#define F_SPACE_AFTER 0x00000040UL
15	#define F_ALLOW_NONREF 0x00000080	44	#define F_ALLOW_NONREF 0x00000100UL
16	#define F_SHRINK 0x00000100	45	#define F_SHRINK 0x00000200UL
		46	#define F_ALLOW_BLESSED 0x00000400UL
		47	#define F_CONV_BLESSED 0x00000800UL
		48	#define F_RELAXED 0x00001000UL
		49	#define F_ALLOW_UNKNOWN 0x00002000UL
		50	#define F_ALLOW_TAGS 0x00004000UL
		51	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
17		52
18	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	53	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
19	#define F_DEFAULT 0
20		54
21	#define INIT_SIZE 32 // initial scalar size to be allocated	55	#define INIT_SIZE 32 // initial scalar size to be allocated
22	#define INDENT_STEP 3 // spaces per indentation level	56	#define INDENT_STEP 3 // spaces per indentation level
23		57
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	58	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
		59
		60	#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
26		61
27	#define SB do {	62	#define SB do {
28	#define SE } while (0)	63	#define SE } while (0)
29		64
30	static HV *json_stash; // JSON::XS::	65	#if __GNUC__ >= 3
		66	# define expect(expr,value) __builtin_expect ((expr), (value))
		67	# define INLINE static inline
		68	#else
		69	# define expect(expr,value) (expr)
		70	# define INLINE static
		71	#endif
		72
		73	#define expect_false(expr) expect ((expr) != 0, 0)
		74	#define expect_true(expr) expect ((expr) != 0, 1)
		75
		76	#define IN_RANGE_INC(type,val,beg,end) \
		77	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		78	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		79
		80	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
		81
		82	#ifdef USE_ITHREADS
		83	# define JSON_SLOW 1
		84	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		85	# define BOOL_STASH (bool_stash ? bool_stash : gv_stashpv ("Types::Serialiser::Boolean", 1))
		86	#else
		87	# define JSON_SLOW 0
		88	# define JSON_STASH json_stash
		89	# define BOOL_STASH bool_stash
		90	#endif
		91
		92	// the amount of HEs to allocate on the stack, when sorting keys
		93	#define STACK_HES 64
		94
		95	static HV *json_stash, *bool_stash; // JSON::XS::, Types::Serialiser::Boolean::
		96	static SV *bool_true, *bool_false, *sv_json;
		97
		98	enum {
		99	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		100	INCR_M_STR, // inside string
		101	INCR_M_BS, // inside backslash
		102	INCR_M_C0, // inside comment in initial whitespace sequence
		103	INCR_M_C1, // inside comment in other places
		104	INCR_M_JSON // outside anything, count nesting
		105	};
		106
		107	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		108
		109	typedef struct {
		110	U32 flags;
		111	U32 max_depth;
		112	STRLEN max_size;
		113
		114	SV *cb_object;
		115	HV *cb_sk_object;
		116
		117	// for the incremental parser
		118	SV *incr_text; // the source text so far
		119	STRLEN incr_pos; // the current offset into the text
		120	int incr_nest; // {[]}-nesting level
		121	unsigned char incr_mode;
		122	} JSON;
		123
		124	INLINE void
		125	json_init (JSON *json)
		126	{
		127	Zero (json, 1, JSON);
		128	json->max_depth = 512;
		129	}
31		130
32	/////////////////////////////////////////////////////////////////////////////	131	/////////////////////////////////////////////////////////////////////////////
33	// utility functions	132	// utility functions
34		133
35	static UV *	134	INLINE SV *
36	SvJSON (SV *sv)	135	get_bool (const char *name)
37	{	136	{
38	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))	137	SV *sv = get_sv (name, 1);
39	croak ("object is not of type JSON::XS");
40		138
41	return &SvUVX (SvRV (sv));	139	SvREADONLY_on (sv);
42	}	140	SvREADONLY_on (SvRV (sv));
43		141
44	static void	142	return sv;
		143	}
		144
		145	INLINE void
45	shrink (SV *sv)	146	shrink (SV *sv)
46	{	147	{
47	sv_utf8_downgrade (sv, 1);	148	sv_utf8_downgrade (sv, 1);
		149
48	if (SvLEN (sv) > SvCUR (sv) + 1)	150	if (SvLEN (sv) > SvCUR (sv) + 1)
49	{	151	{
50	#ifdef SvPV_shrink_to_cur	152	#ifdef SvPV_shrink_to_cur
51	SvPV_shrink_to_cur (sv);	153	SvPV_shrink_to_cur (sv);
52	#elif defined (SvPV_renew)	154	#elif defined (SvPV_renew)
…		…
58	// decode an utf-8 character and return it, or (UV)-1 in	160	// decode an utf-8 character and return it, or (UV)-1 in
59	// case of an error.	161	// case of an error.
60	// we special-case "safe" characters from U+80 .. U+7FF,	162	// we special-case "safe" characters from U+80 .. U+7FF,
61	// but use the very good perl function to parse anything else.	163	// but use the very good perl function to parse anything else.
62	// note that we never call this function for a ascii codepoints	164	// note that we never call this function for a ascii codepoints
63	static UV	165	INLINE UV
64	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)	166	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
65	{	167	{
66	if (s[0] > 0xdf || s[0] < 0xc2)	168	if (expect_true (len >= 2
67	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	169	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
68	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	170	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
69	{	171	{
70	*clen = 2;	172	*clen = 2;
71	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);	173	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
72	}	174	}
73	else	175	else
		176	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		177	}
		178
		179	// likewise for encoding, also never called for ascii codepoints
		180	// this function takes advantage of this fact, although current gccs
		181	// seem to optimise the check for >= 0x80 away anyways
		182	INLINE unsigned char *
		183	encode_utf8 (unsigned char *s, UV ch)
		184	{
		185	if (expect_false (ch < 0x000080))
		186	*s++ = ch;
		187	else if (expect_true (ch < 0x000800))
		188	*s++ = 0xc0 | ( ch >> 6),
		189	*s++ = 0x80 | ( ch & 0x3f);
		190	else if ( ch < 0x010000)
		191	*s++ = 0xe0 | ( ch >> 12),
		192	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		193	*s++ = 0x80 | ( ch & 0x3f);
		194	else if ( ch < 0x110000)
		195	*s++ = 0xf0 | ( ch >> 18),
		196	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		197	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		198	*s++ = 0x80 | ( ch & 0x3f);
		199
		200	return s;
		201	}
		202
		203	// convert offset pointer to character index, sv must be string
		204	static STRLEN
		205	ptr_to_index (SV *sv, char *offset)
		206	{
		207	return SvUTF8 (sv)
		208	? utf8_distance (offset, SvPVX (sv))
		209	: offset - SvPVX (sv);
		210	}
		211
		212	/////////////////////////////////////////////////////////////////////////////
		213	// fp hell
		214
		215	// scan a group of digits, and a trailing exponent
		216	static void
		217	json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
		218	{
		219	UV uaccum = 0;
		220	int eaccum = 0;
		221
		222	// if we recurse too deep, skip all remaining digits
		223	// to avoid a stack overflow attack
		224	if (expect_false (--maxdepth <= 0))
		225	while (((U8)*s - '0') < 10)
		226	++s;
		227
		228	for (;;)
		229	{
		230	U8 dig = (U8)*s - '0';
		231
		232	if (expect_false (dig >= 10))
		233	{
		234	if (dig == (U8)((U8)'.' - (U8)'0'))
		235	{
		236	++s;
		237	json_atof_scan1 (s, accum, expo, 1, maxdepth);
		238	}
		239	else if ((dig | ' ') == 'e' - '0')
		240	{
		241	int exp2 = 0;
		242	int neg = 0;
		243
		244	++s;
		245
		246	if (*s == '-')
		247	{
		248	++s;
		249	neg = 1;
		250	}
		251	else if (*s == '+')
		252	++s;
		253
		254	while ((dig = (U8)*s - '0') < 10)
		255	exp2 = exp2 * 10 + *s++ - '0';
		256
		257	*expo += neg ? -exp2 : exp2;
		258	}
		259
		260	break;
		261	}
		262
		263	++s;
		264
		265	uaccum = uaccum * 10 + dig;
		266	++eaccum;
		267
		268	// if we have too many digits, then recurse for more
		269	// we actually do this for rather few digits
		270	if (uaccum >= (UV_MAX - 9) / 10)
		271	{
		272	if (postdp) *expo -= eaccum;
		273	json_atof_scan1 (s, accum, expo, postdp, maxdepth);
		274	if (postdp) *expo += eaccum;
		275
		276	break;
		277	}
		278	}
		279
		280	// this relies greatly on the quality of the pow ()
		281	// implementation of the platform, but a good
		282	// implementation is hard to beat.
		283	// (IEEE 754 conformant ones are required to be exact)
		284	if (postdp) *expo -= eaccum;
		285	*accum += uaccum * Perl_pow (10., *expo);
		286	*expo += eaccum;
		287	}
		288
		289	static NV
		290	json_atof (const char *s)
		291	{
		292	NV accum = 0.;
		293	int expo = 0;
		294	int neg = 0;
		295
		296	if (*s == '-')
		297	{
		298	++s;
		299	neg = 1;
		300	}
		301
		302	// a recursion depth of ten gives us >>500 bits
		303	json_atof_scan1 (s, &accum, &expo, 0, 10);
		304
		305	return neg ? -accum : accum;
		306	}
		307
		308	// target of scalar reference is bool? -1 == nope, 0 == false, 1 == true
		309	static int
		310	ref_bool_type (SV *sv)
		311	{
		312	svtype svt = SvTYPE (sv);
		313
		314	if (svt < SVt_PVAV)
		315	{
		316	STRLEN len = 0;
		317	char *pv = svt ? SvPV (sv, len) : 0;
		318
		319	if (len == 1)
		320	if (*pv == '1')
		321	return 1;
		322	else if (*pv == '0')
		323	return 0;
		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 (SvTYPE (scalar) >= SVt_PVMG)
		339	{
		340	if (SvSTASH (scalar) == bool_stash)
		341	return 1;
		342
		343	if (!SvOBJECT (scalar) && ref_bool_type (scalar) >= 0)
		344	return 1;
		345	}
		346
		347	return 0;
75	}	348	}
76		349
77	/////////////////////////////////////////////////////////////////////////////	350	/////////////////////////////////////////////////////////////////////////////
78	// encoder	351	// encoder
79		352
…		…
81	typedef struct	354	typedef struct
82	{	355	{
83	char *cur; // SvPVX (sv) + current output position	356	char *cur; // SvPVX (sv) + current output position
84	char *end; // SvEND (sv)	357	char *end; // SvEND (sv)
85	SV *sv; // result scalar	358	SV *sv; // result scalar
86	UV flags; // F_*	359	JSON json;
87	int indent; // indentation level	360	U32 indent; // indentation level
88	int max_depth; // max. recursion level	361	UV limit; // escape character values >= this value when encoding
89	} enc_t;	362	} enc_t;
90		363
91	static void	364	INLINE void
92	need (enc_t *enc, STRLEN len)	365	need (enc_t *enc, STRLEN len)
93	{	366	{
94	if (enc->cur + len >= enc->end)	367	if (expect_false ((uintptr_t)(enc->end - enc->cur) < len))
95	{	368	{
96	STRLEN cur = enc->cur - SvPVX (enc->sv);	369	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
97	SvGROW (enc->sv, cur + len + 1);	370	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
98	enc->cur = SvPVX (enc->sv) + cur;	371	enc->cur = SvPVX (enc->sv) + cur;
99	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	372	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
100	}	373	}
101	}	374	}
102		375
103	static void	376	INLINE void
104	encode_ch (enc_t *enc, char ch)	377	encode_ch (enc_t *enc, char ch)
105	{	378	{
106	need (enc, 1);	379	need (enc, 1);
107	*enc->cur++ = ch;	380	*enc->cur++ = ch;
108	}	381	}
…		…
116		389
117	while (str < end)	390	while (str < end)
118	{	391	{
119	unsigned char ch = *(unsigned char *)str;	392	unsigned char ch = *(unsigned char *)str;
120		393
121	if (ch >= 0x20 && ch < 0x80) // most common case	394	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
122	{	395	{
123	if (ch == '"') // but with slow exceptions	396	if (expect_false (ch == '"')) // but with slow exceptions
124	{	397	{
125	need (enc, len += 1);	398	need (enc, len += 1);
126	*enc->cur++ = '\\';	399	*enc->cur++ = '\\';
127	*enc->cur++ = '"';	400	*enc->cur++ = '"';
128	}	401	}
129	else if (ch == '\\')	402	else if (expect_false (ch == '\\'))
130	{	403	{
131	need (enc, len += 1);	404	need (enc, len += 1);
132	*enc->cur++ = '\\';	405	*enc->cur++ = '\\';
133	*enc->cur++ = '\\';	406	*enc->cur++ = '\\';
134	}	407	}
…		…
152	STRLEN clen;	425	STRLEN clen;
153	UV uch;	426	UV uch;
154		427
155	if (is_utf8)	428	if (is_utf8)
156	{	429	{
157	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
158	uch = decode_utf8 (str, end - str, &clen);	430	uch = decode_utf8 (str, end - str, &clen);
159	if (clen == (STRLEN)-1)	431	if (clen == (STRLEN)-1)
160	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	432	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
161	}	433	}
162	else	434	else
163	{	435	{
164	uch = ch;	436	uch = ch;
165	clen = 1;	437	clen = 1;
166	}	438	}
167		439
168	if (uch > 0x10FFFFUL)	440	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	{	441	{
173	if (uch > 0xFFFFUL)	442	if (uch >= 0x10000UL)
174	{	443	{
		444	if (uch >= 0x110000UL)
		445	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		446
175	need (enc, len += 11);	447	need (enc, len += 11);
176	sprintf (enc->cur, "\\u%04x\\u%04x",	448	sprintf (enc->cur, "\\u%04x\\u%04x",
177	(int)((uch - 0x10000) / 0x400 + 0xD800),	449	(int)((uch - 0x10000) / 0x400 + 0xD800),
178	(int)((uch - 0x10000) % 0x400 + 0xDC00));	450	(int)((uch - 0x10000) % 0x400 + 0xDC00));
179	enc->cur += 12;	451	enc->cur += 12;
180	}	452	}
181	else	453	else
182	{	454	{
183	static char hexdigit [16] = "0123456789abcdef";
184	need (enc, len += 5);	455	need (enc, len += 5);
185	*enc->cur++ = '\\';	456	*enc->cur++ = '\\';
186	*enc->cur++ = 'u';	457	*enc->cur++ = 'u';
187	*enc->cur++ = hexdigit [ uch >> 12 ];	458	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
188	*enc->cur++ = hexdigit [(uch >> 8) & 15];	459	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
189	*enc->cur++ = hexdigit [(uch >> 4) & 15];	460	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
190	*enc->cur++ = hexdigit [(uch >> 0) & 15];	461	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
191	}	462	}
192		463
		464	str += clen;
		465	}
		466	else if (enc->json.flags & F_LATIN1)
		467	{
		468	*enc->cur++ = uch;
193	str += clen;	469	str += clen;
194	}	470	}
195	else if (is_utf8)	471	else if (is_utf8)
196	{	472	{
197	need (enc, len += clen);	473	need (enc, len += clen);
…		…
201	}	477	}
202	while (--clen);	478	while (--clen);
203	}	479	}
204	else	480	else
205	{	481	{
206	need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed	482	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
207	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	483	enc->cur = encode_utf8 (enc->cur, uch);
208	++str;	484	++str;
209	}	485	}
210	}	486	}
211	}	487	}
212	}	488	}
213		489
214	--len;	490	--len;
215	}	491	}
216	}	492	}
217		493
218	static void	494	INLINE void
219	encode_indent (enc_t *enc)	495	encode_indent (enc_t *enc)
220	{	496	{
221	if (enc->flags & F_INDENT)	497	if (enc->json.flags & F_INDENT)
222	{	498	{
223	int spaces = enc->indent * INDENT_STEP;	499	int spaces = enc->indent * INDENT_STEP;
224		500
225	need (enc, spaces);	501	need (enc, spaces);
226	memset (enc->cur, ' ', spaces);	502	memset (enc->cur, ' ', spaces);
227	enc->cur += spaces;	503	enc->cur += spaces;
228	}	504	}
229	}	505	}
230		506
231	static void	507	INLINE void
232	encode_space (enc_t *enc)	508	encode_space (enc_t *enc)
233	{	509	{
234	need (enc, 1);	510	need (enc, 1);
235	encode_ch (enc, ' ');	511	encode_ch (enc, ' ');
236	}	512	}
237		513
238	static void	514	INLINE void
239	encode_nl (enc_t *enc)	515	encode_nl (enc_t *enc)
240	{	516	{
241	if (enc->flags & F_INDENT)	517	if (enc->json.flags & F_INDENT)
242	{	518	{
243	need (enc, 1);	519	need (enc, 1);
244	encode_ch (enc, '\n');	520	encode_ch (enc, '\n');
245	}	521	}
246	}	522	}
247		523
248	static void	524	INLINE void
249	encode_comma (enc_t *enc)	525	encode_comma (enc_t *enc)
250	{	526	{
251	encode_ch (enc, ',');	527	encode_ch (enc, ',');
252		528
253	if (enc->flags & F_INDENT)	529	if (enc->json.flags & F_INDENT)
254	encode_nl (enc);	530	encode_nl (enc);
255	else if (enc->flags & F_SPACE_AFTER)	531	else if (enc->json.flags & F_SPACE_AFTER)
256	encode_space (enc);	532	encode_space (enc);
257	}	533	}
258		534
259	static void encode_sv (enc_t *enc, SV *sv);	535	static void encode_sv (enc_t *enc, SV *sv);
260		536
261	static void	537	static void
262	encode_av (enc_t *enc, AV *av)	538	encode_av (enc_t *enc, AV *av)
263	{	539	{
264	int i, len = av_len (av);	540	int i, len = av_len (av);
265		541
266	encode_ch (enc, '['); encode_nl (enc);	542	if (enc->indent >= enc->json.max_depth)
267	++enc->indent;	543	croak (ERR_NESTING_EXCEEDED);
268		544
		545	encode_ch (enc, '[');
		546
		547	if (len >= 0)
		548	{
		549	encode_nl (enc); ++enc->indent;
		550
269	for (i = 0; i <= len; ++i)	551	for (i = 0; i <= len; ++i)
270	{	552	{
		553	SV **svp = av_fetch (av, i, 0);
		554
271	encode_indent (enc);	555	encode_indent (enc);
272	encode_sv (enc, *av_fetch (av, i, 0));
273		556
		557	if (svp)
		558	encode_sv (enc, *svp);
		559	else
		560	encode_str (enc, "null", 4, 0);
		561
274	if (i < len)	562	if (i < len)
275	encode_comma (enc);	563	encode_comma (enc);
276	}	564	}
277		565
		566	encode_nl (enc); --enc->indent; encode_indent (enc);
		567	}
		568
278	encode_nl (enc);	569	encode_ch (enc, ']');
279
280	--enc->indent;
281	encode_indent (enc); encode_ch (enc, ']');
282	}	570	}
283		571
284	static void	572	static void
285	encode_he (enc_t *enc, HE *he)	573	encode_hk (enc_t *enc, HE *he)
286	{	574	{
287	encode_ch (enc, '"');	575	encode_ch (enc, '"');
288		576
289	if (HeKLEN (he) == HEf_SVKEY)	577	if (HeKLEN (he) == HEf_SVKEY)
290	{	578	{
291	SV *sv = HeSVKEY (he);	579	SV *sv = HeSVKEY (he);
292	STRLEN len;	580	STRLEN len;
293	char *str;	581	char *str;
294		582
295	SvGETMAGIC (sv);	583	SvGETMAGIC (sv);
296	str = SvPV (sv, len);	584	str = SvPV (sv, len);
297		585
298	encode_str (enc, str, len, SvUTF8 (sv));	586	encode_str (enc, str, len, SvUTF8 (sv));
299	}	587	}
300	else	588	else
301	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	589	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
302		590
303	encode_ch (enc, '"');	591	encode_ch (enc, '"');
304		592
305	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	593	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
306	encode_ch (enc, ':');	594	encode_ch (enc, ':');
307	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	595	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
308	encode_sv (enc, HeVAL (he));
309	}	596	}
310		597
311	// compare hash entries, used when all keys are bytestrings	598	// compare hash entries, used when all keys are bytestrings
312	static int	599	static int
313	he_cmp_fast (const void *a_, const void *b_)	600	he_cmp_fast (const void *a_, const void *b_)
…		…
318	HE *b = *(HE **)b_;	605	HE *b = *(HE **)b_;
319		606
320	STRLEN la = HeKLEN (a);	607	STRLEN la = HeKLEN (a);
321	STRLEN lb = HeKLEN (b);	608	STRLEN lb = HeKLEN (b);
322		609
323	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	610	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
324	cmp = la - lb;	611	cmp = lb - la;
325		612
326	return cmp;	613	return cmp;
327	}	614	}
328		615
329	// compare hash entries, used when some keys are sv's or utf-x	616	// compare hash entries, used when some keys are sv's or utf-x
330	static int	617	static int
331	he_cmp_slow (const void *a, const void *b)	618	he_cmp_slow (const void *a, const void *b)
332	{	619	{
333	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	620	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
334	}	621	}
335		622
336	static void	623	static void
337	encode_hv (enc_t *enc, HV *hv)	624	encode_hv (enc_t *enc, HV *hv)
338	{	625	{
339	int count, i;	626	HE *he;
340		627
341	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	628	if (enc->indent >= enc->json.max_depth)
		629	croak (ERR_NESTING_EXCEEDED);
342		630
343	if ((count = hv_iterinit (hv)))	631	encode_ch (enc, '{');
344	{	632
345	// for canonical output we have to sort by keys first	633	// for canonical output we have to sort by keys first
346	// actually, this is mostly due to the stupid so-called	634	// actually, this is mostly due to the stupid so-called
347	// security workaround added somewhere in 5.8.x.	635	// security workaround added somewhere in 5.8.x
348	// that randomises hash orderings	636	// that randomises hash orderings
349	if (enc->flags & F_CANONICAL)	637	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
		638	{
		639	int count = hv_iterinit (hv);
		640
		641	if (SvMAGICAL (hv))
350	{	642	{
351	HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode	643	// need to count by iterating. could improve by dynamically building the vector below
		644	// but I don't care for the speed of this special case.
		645	// note also that we will run into undefined behaviour when the two iterations
		646	// do not result in the same count, something I might care for in some later release.
		647
		648	count = 0;
		649	while (hv_iternext (hv))
		650	++count;
		651
		652	hv_iterinit (hv);
		653	}
		654
		655	if (count)
		656	{
352	int fast = 1;	657	int i, fast = 1;
		658	HE *hes_stack [STACK_HES];
		659	HE **hes = hes_stack;
		660
		661	// allocate larger arrays on the heap
		662	if (count > STACK_HES)
		663	{
		664	SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
		665	hes = (HE **)SvPVX (sv);
		666	}
353		667
354	i = 0;	668	i = 0;
355	while ((he = hv_iternext (hv)))	669	while ((he = hv_iternext (hv)))
356	{	670	{
357	hes [i++] = he;	671	hes [i++] = he;
…		…
379		693
380	FREETMPS;	694	FREETMPS;
381	LEAVE;	695	LEAVE;
382	}	696	}
383		697
384	for (i = 0; i < count; ++i)	698	encode_nl (enc); ++enc->indent;
		699
		700	while (count--)
385	{	701	{
386	encode_indent (enc);	702	encode_indent (enc);
		703	he = hes [count];
387	encode_he (enc, hes [i]);	704	encode_hk (enc, he);
		705	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
388		706
389	if (i < count - 1)	707	if (count)
390	encode_comma (enc);	708	encode_comma (enc);
391	}	709	}
392		710
393	encode_nl (enc);	711	encode_nl (enc); --enc->indent; encode_indent (enc);
394	}	712	}
		713	}
		714	else
		715	{
		716	if (hv_iterinit (hv) || SvMAGICAL (hv))
		717	if ((he = hv_iternext (hv)))
		718	{
		719	encode_nl (enc); ++enc->indent;
		720
		721	for (;;)
		722	{
		723	encode_indent (enc);
		724	encode_hk (enc, he);
		725	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		726
		727	if (!(he = hv_iternext (hv)))
		728	break;
		729
		730	encode_comma (enc);
		731	}
		732
		733	encode_nl (enc); --enc->indent; encode_indent (enc);
		734	}
		735	}
		736
		737	encode_ch (enc, '}');
		738	}
		739
		740	// encode objects, arrays and special \0=false and \1=true values.
		741	static void
		742	encode_rv (enc_t *enc, SV *sv)
		743	{
		744	svtype svt;
		745	GV *method;
		746
		747	SvGETMAGIC (sv);
		748	svt = SvTYPE (sv);
		749
		750	if (expect_false (SvOBJECT (sv)))
		751	{
		752	HV *stash = SvSTASH (sv);
		753
		754	if (stash == bool_stash)
		755	{
		756	if (SvIV (sv))
		757	encode_str (enc, "true", 4, 0);
		758	else
		759	encode_str (enc, "false", 5, 0);
		760	}
		761	else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
		762	{
		763	int count;
		764	dSP;
		765
		766	ENTER; SAVETMPS;
		767	SAVESTACK_POS ();
		768	PUSHMARK (SP);
		769	EXTEND (SP, 2);
		770	// we re-bless the reference to get overload and other niceties right
		771	PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		772	PUSHs (sv_json);
		773
		774	PUTBACK;
		775	count = call_sv ((SV *)GvCV (method), G_ARRAY);
		776	SPAGAIN;
		777
		778	// catch this surprisingly common error
		779	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		780	croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		781
		782	encode_ch (enc, '(');
		783	encode_ch (enc, '"');
		784	encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
		785	encode_ch (enc, '"');
		786	encode_ch (enc, ')');
		787	encode_ch (enc, '[');
		788
		789	while (count)
		790	{
		791	encode_sv (enc, SP[1 - count--]);
		792
		793	if (count)
		794	encode_ch (enc, ',');
		795	}
		796
		797	encode_ch (enc, ']');
		798
		799	FREETMPS; LEAVE;
		800	}
		801	else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
		802	{
		803	dSP;
		804
		805	ENTER; SAVETMPS;
		806	PUSHMARK (SP);
		807	// we re-bless the reference to get overload and other niceties right
		808	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		809
		810	// calling with G_SCALAR ensures that we always get a 1 return value
		811	PUTBACK;
		812	call_sv ((SV *)GvCV (method), G_SCALAR);
		813	SPAGAIN;
		814
		815	// catch this surprisingly common error
		816	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		817	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		818
		819	sv = POPs;
		820	PUTBACK;
		821
		822	encode_sv (enc, sv);
		823
		824	FREETMPS; LEAVE;
		825	}
		826	else if (enc->json.flags & F_ALLOW_BLESSED)
		827	encode_str (enc, "null", 4, 0);
395	else	828	else
396	{	829	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;	830	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
398	HE *he = hv_iternext (hv);	831	}
399		832	else if (svt == SVt_PVHV)
400	for (;;)	833	encode_hv (enc, (HV *)sv);
401	{	834	else if (svt == SVt_PVAV)
402	encode_indent (enc);	835	encode_av (enc, (AV *)sv);
403	encode_he (enc, he);	836	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	}	837	{
		838	int bool_type = ref_bool_type (sv);
414		839
415	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	840	if (bool_type == 1)
		841	encode_str (enc, "true", 4, 0);
		842	else if (bool_type == 0)
		843	encode_str (enc, "false", 5, 0);
		844	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		845	encode_str (enc, "null", 4, 0);
		846	else
		847	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		848	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		849	}
		850	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		851	encode_str (enc, "null", 4, 0);
		852	else
		853	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		854	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
416	}	855	}
417		856
418	static void	857	static void
419	encode_sv (enc_t *enc, SV *sv)	858	encode_sv (enc_t *enc, SV *sv)
420	{	859	{
…		…
428	encode_str (enc, str, len, SvUTF8 (sv));	867	encode_str (enc, str, len, SvUTF8 (sv));
429	encode_ch (enc, '"');	868	encode_ch (enc, '"');
430	}	869	}
431	else if (SvNOKp (sv))	870	else if (SvNOKp (sv))
432	{	871	{
		872	// trust that perl will do the right thing w.r.t. JSON syntax.
433	need (enc, NV_DIG + 32);	873	need (enc, NV_DIG + 32);
434	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	874	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
435	enc->cur += strlen (enc->cur);	875	enc->cur += strlen (enc->cur);
436	}	876	}
437	else if (SvIOKp (sv))	877	else if (SvIOKp (sv))
438	{	878	{
		879	// we assume we can always read an IV as a UV and vice versa
		880	// we assume two's complement
		881	// we assume no aliasing issues in the union
		882	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		883	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		884	{
		885	// optimise the "small number case"
		886	// code will likely be branchless and use only a single multiplication
		887	// works for numbers up to 59074
		888	I32 i = SvIVX (sv);
		889	U32 u;
		890	char digit, nz = 0;
		891
439	need (enc, 64);	892	need (enc, 6);
		893
		894	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		895	u = i < 0 ? -i : i;
		896
		897	// convert to 4.28 fixed-point representation
		898	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		899
		900	// now output digit by digit, each time masking out the integer part
		901	// and multiplying by 5 while moving the decimal point one to the right,
		902	// resulting in a net multiplication by 10.
		903	// we always write the digit to memory but conditionally increment
		904	// the pointer, to enable the use of conditional move instructions.
		905	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		906	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		907	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		908	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		909	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		910	}
		911	else
		912	{
		913	// large integer, use the (rather slow) snprintf way.
		914	need (enc, IVUV_MAXCHARS);
440	enc->cur +=	915	enc->cur +=
441	SvIsUV(sv)	916	SvIsUV(sv)
442	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	917	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
443	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	918	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		919	}
444	}	920	}
445	else if (SvROK (sv))	921	else if (SvROK (sv))
446	{	922	encode_rv (enc, SvRV (sv));
447	SV *rv = SvRV (sv);	923	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);	924	encode_str (enc, "null", 4, 0);
464	else	925	else
465	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	926	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
466	SvPV_nolen (sv), SvFLAGS (sv));	927	SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
467	}	928	}
468		929
469	static SV *	930	static SV *
470	encode_json (SV *scalar, UV flags)	931	encode_json (SV *scalar, JSON *json)
471	{	932	{
		933	enc_t enc;
		934
472	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	935	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)");	936	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
474		937
475	enc_t enc;	938	enc.json = *json;
476	enc.flags = flags;
477	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	939	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
478	enc.cur = SvPVX (enc.sv);	940	enc.cur = SvPVX (enc.sv);
479	enc.end = SvEND (enc.sv);	941	enc.end = SvEND (enc.sv);
480	enc.indent = 0;	942	enc.indent = 0;
481	enc.max_depth = 0x7fffffffUL;	943	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		944	: enc.json.flags & F_LATIN1 ? 0x000100UL
		945	: 0x110000UL;
482		946
483	SvPOK_only (enc.sv);	947	SvPOK_only (enc.sv);
484	encode_sv (&enc, scalar);	948	encode_sv (&enc, scalar);
		949	encode_nl (&enc);
485		950
		951	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		952	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		953
486	if (!(flags & (F_ASCII | F_UTF8)))	954	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
487	SvUTF8_on (enc.sv);	955	SvUTF8_on (enc.sv);
488		956
489	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
490
491	if (enc.flags & F_SHRINK)	957	if (enc.json.flags & F_SHRINK)
492	shrink (enc.sv);	958	shrink (enc.sv);
493		959
494	return enc.sv;	960	return enc.sv;
495	}	961	}
496		962
…		…
501	typedef struct	967	typedef struct
502	{	968	{
503	char *cur; // current parser pointer	969	char *cur; // current parser pointer
504	char *end; // end of input string	970	char *end; // end of input string
505	const char *err; // parse error, if != 0	971	const char *err; // parse error, if != 0
506	UV flags; // F_*	972	JSON json;
		973	U32 depth; // recursion depth
		974	U32 maxdepth; // recursion depth limit
507	} dec_t;	975	} dec_t;
508		976
509	static void	977	INLINE void
		978	decode_comment (dec_t *dec)
		979	{
		980	// only '#'-style comments allowed a.t.m.
		981
		982	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		983	++dec->cur;
		984	}
		985
		986	INLINE void
510	decode_ws (dec_t *dec)	987	decode_ws (dec_t *dec)
511	{	988	{
512	for (;;)	989	for (;;)
513	{	990	{
514	char ch = *dec->cur;	991	char ch = *dec->cur;
515		992
516	if (ch > 0x20	993	if (ch > 0x20)
		994	{
		995	if (expect_false (ch == '#'))
		996	{
		997	if (dec->json.flags & F_RELAXED)
		998	decode_comment (dec);
		999	else
		1000	break;
		1001	}
		1002	else
		1003	break;
		1004	}
517	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	1005	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
518	break;	1006	break; // parse error, but let higher level handle it, gives better error messages
519		1007
520	++dec->cur;	1008	++dec->cur;
521	}	1009	}
522	}	1010	}
523		1011
524	#define ERR(reason) SB dec->err = reason; goto fail; SE	1012	#define ERR(reason) SB dec->err = reason; goto fail; SE
		1013
525	#define EXPECT_CH(ch) SB \	1014	#define EXPECT_CH(ch) SB \
526	if (*dec->cur != ch) \	1015	if (*dec->cur != ch) \
527	ERR (# ch " expected"); \	1016	ERR (# ch " expected"); \
528	++dec->cur; \	1017	++dec->cur; \
529	SE	1018	SE
530		1019
		1020	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
		1021	#define DEC_DEC_DEPTH --dec->depth
		1022
531	static SV *decode_sv (dec_t *dec);	1023	static SV *decode_sv (dec_t *dec);
532		1024
533	static signed char decode_hexdigit[256];	1025	static signed char decode_hexdigit[256];
534		1026
535	static UV	1027	static UV
536	decode_4hex (dec_t *dec)	1028	decode_4hex (dec_t *dec)
537	{	1029	{
538	signed char d1, d2, d3, d4;	1030	signed char d1, d2, d3, d4;
539	unsigned char *cur = (unsigned char *)dec->cur;	1031	unsigned char *cur = (unsigned char *)dec->cur;
540		1032
541	d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected");	1033	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");	1034	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");	1035	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");	1036	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
545		1037
546	dec->cur += 4;	1038	dec->cur += 4;
547		1039
548	return ((UV)d1) << 12	1040	return ((UV)d1) << 12
549	| ((UV)d2) << 8	1041	| ((UV)d2) << 8
…		…
557	static SV *	1049	static SV *
558	decode_str (dec_t *dec)	1050	decode_str (dec_t *dec)
559	{	1051	{
560	SV *sv = 0;	1052	SV *sv = 0;
561	int utf8 = 0;	1053	int utf8 = 0;
		1054	char *dec_cur = dec->cur;
562		1055
563	do	1056	do
564	{	1057	{
565	char buf [SHORT_STRING_LEN + UTF8_MAX_LEN];	1058	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
566	char *cur = buf;	1059	char *cur = buf;
567		1060
568	do	1061	do
569	{	1062	{
570	unsigned char ch = *(unsigned char *)dec->cur++;	1063	unsigned char ch = *(unsigned char *)dec_cur++;
571		1064
572	if (ch == '"')	1065	if (expect_false (ch == '"'))
573	{	1066	{
574	--dec->cur;	1067	--dec_cur;
575	break;	1068	break;
576	}	1069	}
577	else if (ch == '\\')	1070	else if (expect_false (ch == '\\'))
578	{	1071	{
579	switch (*dec->cur)	1072	switch (*dec_cur)
580	{	1073	{
581	case '\\':	1074	case '\\':
582	case '/':	1075	case '/':
583	case '"': *cur++ = *dec->cur++; break;	1076	case '"': *cur++ = *dec_cur++; break;
584		1077
585	case 'b': ++dec->cur; *cur++ = '\010'; break;	1078	case 'b': ++dec_cur; *cur++ = '\010'; break;
586	case 't': ++dec->cur; *cur++ = '\011'; break;	1079	case 't': ++dec_cur; *cur++ = '\011'; break;
587	case 'n': ++dec->cur; *cur++ = '\012'; break;	1080	case 'n': ++dec_cur; *cur++ = '\012'; break;
588	case 'f': ++dec->cur; *cur++ = '\014'; break;	1081	case 'f': ++dec_cur; *cur++ = '\014'; break;
589	case 'r': ++dec->cur; *cur++ = '\015'; break;	1082	case 'r': ++dec_cur; *cur++ = '\015'; break;
590		1083
591	case 'u':	1084	case 'u':
592	{	1085	{
593	UV lo, hi;	1086	UV lo, hi;
594	++dec->cur;	1087	++dec_cur;
595		1088
		1089	dec->cur = dec_cur;
596	hi = decode_4hex (dec);	1090	hi = decode_4hex (dec);
		1091	dec_cur = dec->cur;
597	if (hi == (UV)-1)	1092	if (hi == (UV)-1)
598	goto fail;	1093	goto fail;
599		1094
600	// possibly a surrogate pair	1095	// possibly a surrogate pair
601	if (hi >= 0xd800)	1096	if (hi >= 0xd800)
602	if (hi < 0xdc00)	1097	if (hi < 0xdc00)
603	{	1098	{
604	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	1099	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
605	ERR ("missing low surrogate character in surrogate pair");	1100	ERR ("missing low surrogate character in surrogate pair");
606		1101
607	dec->cur += 2;	1102	dec_cur += 2;
608		1103
		1104	dec->cur = dec_cur;
609	lo = decode_4hex (dec);	1105	lo = decode_4hex (dec);
		1106	dec_cur = dec->cur;
610	if (lo == (UV)-1)	1107	if (lo == (UV)-1)
611	goto fail;	1108	goto fail;
612		1109
613	if (lo < 0xdc00 || lo >= 0xe000)	1110	if (lo < 0xdc00 || lo >= 0xe000)
614	ERR ("surrogate pair expected");	1111	ERR ("surrogate pair expected");
…		…
620		1117
621	if (hi >= 0x80)	1118	if (hi >= 0x80)
622	{	1119	{
623	utf8 = 1;	1120	utf8 = 1;
624		1121
625	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	1122	cur = encode_utf8 (cur, hi);
626	}	1123	}
627	else	1124	else
628	*cur++ = hi;	1125	*cur++ = hi;
629	}	1126	}
630	break;	1127	break;
631		1128
632	default:	1129	default:
633	--dec->cur;	1130	--dec_cur;
634	ERR ("illegal backslash escape sequence in string");	1131	ERR ("illegal backslash escape sequence in string");
635	}	1132	}
636	}	1133	}
637	else if (ch >= 0x20 && ch <= 0x7f)	1134	else if (expect_true (ch >= 0x20 && ch < 0x80))
638	*cur++ = ch;	1135	*cur++ = ch;
639	else if (ch >= 0x80)	1136	else if (ch >= 0x80)
640	{	1137	{
641	--dec->cur;
642
643	STRLEN clen;	1138	STRLEN clen;
		1139
		1140	--dec_cur;
		1141
644	UV uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	1142	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
645	if (clen == (STRLEN)-1)	1143	if (clen == (STRLEN)-1)
646	ERR ("malformed UTF-8 character in JSON string");	1144	ERR ("malformed UTF-8 character in JSON string");
647		1145
648	do	1146	do
649	{
650	*cur++ = *dec->cur++;	1147	*cur++ = *dec_cur++;
651	}
652	while (--clen);	1148	while (--clen);
653		1149
654	utf8 = 1;	1150	utf8 = 1;
655	}	1151	}
656	else if (!ch)	1152	else if (ch == '\t' && dec->json.flags & F_RELAXED)
657	ERR ("unexpected end of string while parsing json string");	1153	*cur++ = ch;
658	else	1154	else
		1155	{
		1156	--dec_cur;
		1157
		1158	if (!ch)
		1159	ERR ("unexpected end of string while parsing JSON string");
		1160	else
659	ERR ("invalid character encountered");	1161	ERR ("invalid character encountered while parsing JSON string");
660		1162	}
661	}	1163	}
662	while (cur < buf + SHORT_STRING_LEN);	1164	while (cur < buf + SHORT_STRING_LEN);
663		1165
		1166	{
664	STRLEN len = cur - buf;	1167	STRLEN len = cur - buf;
665		1168
666	if (sv)	1169	if (sv)
667	{	1170	{
668	SvGROW (sv, SvCUR (sv) + len + 1);	1171	STRLEN cur = SvCUR (sv);
		1172
		1173	if (SvLEN (sv) <= cur + len)
		1174	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1175
669	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	1176	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
670	SvCUR_set (sv, SvCUR (sv) + len);	1177	SvCUR_set (sv, SvCUR (sv) + len);
671	}	1178	}
672	else	1179	else
673	sv = newSVpvn (buf, len);	1180	sv = newSVpvn (buf, len);
674	}	1181	}
		1182	}
675	while (*dec->cur != '"');	1183	while (*dec_cur != '"');
676		1184
677	++dec->cur;	1185	++dec_cur;
678		1186
679	if (sv)	1187	if (sv)
680	{	1188	{
681	SvPOK_only (sv);	1189	SvPOK_only (sv);
682	*SvEND (sv) = 0;	1190	*SvEND (sv) = 0;
…		…
685	SvUTF8_on (sv);	1193	SvUTF8_on (sv);
686	}	1194	}
687	else	1195	else
688	sv = newSVpvn ("", 0);	1196	sv = newSVpvn ("", 0);
689		1197
		1198	dec->cur = dec_cur;
690	return sv;	1199	return sv;
691		1200
692	fail:	1201	fail:
		1202	dec->cur = dec_cur;
693	return 0;	1203	return 0;
694	}	1204	}
695		1205
696	static SV *	1206	static SV *
697	decode_num (dec_t *dec)	1207	decode_num (dec_t *dec)
…		…
755	is_nv = 1;	1265	is_nv = 1;
756	}	1266	}
757		1267
758	if (!is_nv)	1268	if (!is_nv)
759	{	1269	{
760	UV uv;	1270	int len = dec->cur - start;
761	int numtype = grok_number (start, dec->cur - start, &uv);	1271
762	if (numtype & IS_NUMBER_IN_UV)	1272	// special case the rather common 1..5-digit-int case
763	if (numtype & IS_NUMBER_NEG)	1273	if (*start == '-')
		1274	switch (len)
764	{	1275	{
765	if (uv < (UV)IV_MIN)	1276	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
766	return newSViv (-(IV)uv);	1277	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
		1278	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1279	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1280	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
767	}	1281	}
		1282	else
		1283	switch (len)
		1284	{
		1285	case 1: return newSViv ( start [0] - '0' * 1);
		1286	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1287	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1288	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1289	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1290	}
		1291
		1292	{
		1293	UV uv;
		1294	int numtype = grok_number (start, len, &uv);
		1295	if (numtype & IS_NUMBER_IN_UV)
		1296	if (numtype & IS_NUMBER_NEG)
		1297	{
		1298	if (uv < (UV)IV_MIN)
		1299	return newSViv (-(IV)uv);
		1300	}
768	else	1301	else
769	return newSVuv (uv);	1302	return newSVuv (uv);
770	}	1303	}
771		1304
		1305	len -= *start == '-' ? 1 : 0;
		1306
		1307	// does not fit into IV or UV, try NV
		1308	if (len <= NV_DIG)
		1309	// fits into NV without loss of precision
		1310	return newSVnv (json_atof (start));
		1311
		1312	// everything else fails, convert it to a string
		1313	return newSVpvn (start, dec->cur - start);
		1314	}
		1315
		1316	// loss of precision here
772	return newSVnv (Atof (start));	1317	return newSVnv (json_atof (start));
773		1318
774	fail:	1319	fail:
775	return 0;	1320	return 0;
776	}	1321	}
777		1322
778	static SV *	1323	static SV *
779	decode_av (dec_t *dec)	1324	decode_av (dec_t *dec)
780	{	1325	{
781	AV *av = newAV ();	1326	AV *av = newAV ();
782		1327
		1328	DEC_INC_DEPTH;
783	decode_ws (dec);	1329	decode_ws (dec);
		1330
784	if (*dec->cur == ']')	1331	if (*dec->cur == ']')
785	++dec->cur;	1332	++dec->cur;
786	else	1333	else
787	for (;;)	1334	for (;;)
788	{	1335	{
…		…
799	if (*dec->cur == ']')	1346	if (*dec->cur == ']')
800	{	1347	{
801	++dec->cur;	1348	++dec->cur;
802	break;	1349	break;
803	}	1350	}
804		1351
805	if (*dec->cur != ',')	1352	if (*dec->cur != ',')
806	ERR (", or ] expected while parsing array");	1353	ERR (", or ] expected while parsing array");
807		1354
808	++dec->cur;	1355	++dec->cur;
		1356
		1357	decode_ws (dec);
		1358
		1359	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1360	{
		1361	++dec->cur;
		1362	break;
		1363	}
809	}	1364	}
810		1365
		1366	DEC_DEC_DEPTH;
811	return newRV_noinc ((SV *)av);	1367	return newRV_noinc ((SV *)av);
812		1368
813	fail:	1369	fail:
814	SvREFCNT_dec (av);	1370	SvREFCNT_dec (av);
		1371	DEC_DEC_DEPTH;
815	return 0;	1372	return 0;
816	}	1373	}
817		1374
818	static SV *	1375	static SV *
819	decode_hv (dec_t *dec)	1376	decode_hv (dec_t *dec)
820	{	1377	{
		1378	SV *sv;
821	HV *hv = newHV ();	1379	HV *hv = newHV ();
822		1380
		1381	DEC_INC_DEPTH;
823	decode_ws (dec);	1382	decode_ws (dec);
		1383
824	if (*dec->cur == '}')	1384	if (*dec->cur == '}')
825	++dec->cur;	1385	++dec->cur;
826	else	1386	else
827	for (;;)	1387	for (;;)
828	{	1388	{
		1389	EXPECT_CH ('"');
		1390
		1391	// heuristic: assume that
		1392	// a) decode_str + hv_store_ent are abysmally slow.
		1393	// b) most hash keys are short, simple ascii text.
		1394	// => try to "fast-match" such strings to avoid
		1395	// the overhead of decode_str + hv_store_ent.
		1396	{
829	SV *key, *value;	1397	SV *value;
		1398	char *p = dec->cur;
		1399	char *e = p + 24; // only try up to 24 bytes
830		1400
831	decode_ws (dec); EXPECT_CH ('"');	1401	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	{	1402	{
		1403	// the >= 0x80 is false on most architectures
		1404	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1405	{
		1406	// slow path, back up and use decode_str
		1407	SV *key = decode_str (dec);
		1408	if (!key)
		1409	goto fail;
		1410
		1411	decode_ws (dec); EXPECT_CH (':');
		1412
		1413	decode_ws (dec);
		1414	value = decode_sv (dec);
		1415	if (!value)
		1416	{
		1417	SvREFCNT_dec (key);
		1418	goto fail;
		1419	}
		1420
		1421	hv_store_ent (hv, key, value, 0);
842	SvREFCNT_dec (key);	1422	SvREFCNT_dec (key);
		1423
		1424	break;
		1425	}
		1426	else if (*p == '"')
		1427	{
		1428	// fast path, got a simple key
		1429	char *key = dec->cur;
		1430	int len = p - key;
		1431	dec->cur = p + 1;
		1432
		1433	decode_ws (dec); EXPECT_CH (':');
		1434
		1435	decode_ws (dec);
		1436	value = decode_sv (dec);
		1437	if (!value)
843	goto fail;	1438	goto fail;
		1439
		1440	hv_store (hv, key, len, value, 0);
		1441
		1442	break;
		1443	}
		1444
		1445	++p;
844	}	1446	}
845		1447	}
846	//TODO: optimise
847	hv_store_ent (hv, key, value, 0);
848		1448
849	decode_ws (dec);	1449	decode_ws (dec);
850		1450
851	if (*dec->cur == '}')	1451	if (*dec->cur == '}')
852	{	1452	{
…		…
856		1456
857	if (*dec->cur != ',')	1457	if (*dec->cur != ',')
858	ERR (", or } expected while parsing object/hash");	1458	ERR (", or } expected while parsing object/hash");
859		1459
860	++dec->cur;	1460	++dec->cur;
		1461
		1462	decode_ws (dec);
		1463
		1464	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1465	{
		1466	++dec->cur;
		1467	break;
		1468	}
861	}	1469	}
862		1470
		1471	DEC_DEC_DEPTH;
863	return newRV_noinc ((SV *)hv);	1472	sv = newRV_noinc ((SV *)hv);
		1473
		1474	// check filter callbacks
		1475	if (dec->json.flags & F_HOOK)
		1476	{
		1477	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1478	{
		1479	HE *cb, *he;
		1480
		1481	hv_iterinit (hv);
		1482	he = hv_iternext (hv);
		1483	hv_iterinit (hv);
		1484
		1485	// the next line creates a mortal sv each time it's called.
		1486	// might want to optimise this for common cases.
		1487	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1488
		1489	if (cb)
		1490	{
		1491	dSP;
		1492	int count;
		1493
		1494	ENTER; SAVETMPS;
		1495	SAVESTACK_POS ();
		1496	PUSHMARK (SP);
		1497	XPUSHs (HeVAL (he));
		1498	sv_2mortal (sv);
		1499
		1500	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1501
		1502	if (count == 1)
		1503	{
		1504	sv = newSVsv (POPs);
		1505	FREETMPS; LEAVE;
		1506	return sv;
		1507	}
		1508
		1509	SvREFCNT_inc (sv);
		1510	FREETMPS; LEAVE;
		1511	}
		1512	}
		1513
		1514	if (dec->json.cb_object)
		1515	{
		1516	dSP;
		1517	int count;
		1518
		1519	ENTER; SAVETMPS;
		1520	SAVESTACK_POS ();
		1521	PUSHMARK (SP);
		1522	XPUSHs (sv_2mortal (sv));
		1523
		1524	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1525
		1526	if (count == 1)
		1527	{
		1528	sv = newSVsv (POPs);
		1529	FREETMPS; LEAVE;
		1530	return sv;
		1531	}
		1532
		1533	SvREFCNT_inc (sv);
		1534	FREETMPS; LEAVE;
		1535	}
		1536	}
		1537
		1538	return sv;
864		1539
865	fail:	1540	fail:
866	SvREFCNT_dec (hv);	1541	SvREFCNT_dec (hv);
		1542	DEC_DEC_DEPTH;
		1543	return 0;
		1544	}
		1545
		1546	static SV *
		1547	decode_tag (dec_t *dec)
		1548	{
		1549	SV *tag = 0;
		1550	SV *val = 0;
		1551
		1552	if (!(dec->json.flags & F_ALLOW_TAGS))
		1553	ERR ("malformed JSON string, neither array, object, number, string or atom");
		1554
		1555	++dec->cur;
		1556
		1557	decode_ws (dec);
		1558
		1559	tag = decode_sv (dec);
		1560	if (!tag)
		1561	goto fail;
		1562
		1563	if (!SvPOK (tag))
		1564	ERR ("malformed JSON string, (tag) must be a string");
		1565
		1566	decode_ws (dec);
		1567
		1568	if (*dec->cur != ')')
		1569	ERR (") expected after tag");
		1570
		1571	++dec->cur;
		1572
		1573	decode_ws (dec);
		1574
		1575	val = decode_sv (dec);
		1576	if (!val)
		1577	goto fail;
		1578
		1579	if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
		1580	ERR ("malformed JSON string, tag value must be an array");
		1581
		1582	{
		1583	AV *av = (AV *)SvRV (val);
		1584	int i, len = av_len (av) + 1;
		1585	HV *stash = gv_stashsv (tag, 0);
		1586	SV *sv;
		1587
		1588	if (!stash)
		1589	ERR ("cannot decode perl-object (package does not exist)");
		1590
		1591	GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
		1592
		1593	if (!method)
		1594	ERR ("cannot decode perl-object (package does not have a THAW method)");
		1595
		1596	dSP;
		1597
		1598	ENTER; SAVETMPS;
		1599	PUSHMARK (SP);
		1600	EXTEND (SP, len + 2);
		1601	// we re-bless the reference to get overload and other niceties right
		1602	PUSHs (tag);
		1603	PUSHs (sv_json);
		1604
		1605	for (i = 0; i < len; ++i)
		1606	PUSHs (*av_fetch (av, i, 1));
		1607
		1608	PUTBACK;
		1609	call_sv ((SV *)GvCV (method), G_SCALAR);
		1610	SPAGAIN;
		1611
		1612	SvREFCNT_dec (tag);
		1613	SvREFCNT_dec (val);
		1614	sv = SvREFCNT_inc (POPs);
		1615
		1616	PUTBACK;
		1617
		1618	FREETMPS; LEAVE;
		1619
		1620	return sv;
		1621	}
		1622
		1623	fail:
		1624	SvREFCNT_dec (tag);
		1625	SvREFCNT_dec (val);
867	return 0;	1626	return 0;
868	}	1627	}
869		1628
870	static SV *	1629	static SV *
871	decode_sv (dec_t *dec)	1630	decode_sv (dec_t *dec)
872	{	1631	{
873	decode_ws (dec);	1632	// the beauty of JSON: you need exactly one character lookahead
		1633	// to parse everything.
874	switch (*dec->cur)	1634	switch (*dec->cur)
875	{	1635	{
876	case '"': ++dec->cur; return decode_str (dec);	1636	case '"': ++dec->cur; return decode_str (dec);
877	case '[': ++dec->cur; return decode_av (dec);	1637	case '[': ++dec->cur; return decode_av (dec);
878	case '{': ++dec->cur; return decode_hv (dec);	1638	case '{': ++dec->cur; return decode_hv (dec);
		1639	case '(': return decode_tag (dec);
879		1640
880	case '-':	1641	case '-':
881	case '0': case '1': case '2': case '3': case '4':	1642	case '0': case '1': case '2': case '3': case '4':
882	case '5': case '6': case '7': case '8': case '9':	1643	case '5': case '6': case '7': case '8': case '9':
883	return decode_num (dec);	1644	return decode_num (dec);
884		1645
885	case 't':	1646	case 't':
886	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1647	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
887	{	1648	{
888	dec->cur += 4;	1649	dec->cur += 4;
		1650	#if JSON_SLOW
		1651	bool_true = get_bool ("Types::Serialiser::true");
		1652	#endif
889	return newSViv (1);	1653	return newSVsv (bool_true);
890	}	1654	}
891	else	1655	else
892	ERR ("'true' expected");	1656	ERR ("'true' expected");
893		1657
894	break;	1658	break;
895		1659
896	case 'f':	1660	case 'f':
897	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1661	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
898	{	1662	{
899	dec->cur += 5;	1663	dec->cur += 5;
		1664	#if JSON_SLOW
		1665	bool_false = get_bool ("Types::Serialiser::false");
		1666	#endif
900	return newSViv (0);	1667	return newSVsv (bool_false);
901	}	1668	}
902	else	1669	else
903	ERR ("'false' expected");	1670	ERR ("'false' expected");
904		1671
905	break;	1672	break;
…		…
914	ERR ("'null' expected");	1681	ERR ("'null' expected");
915		1682
916	break;	1683	break;
917		1684
918	default:	1685	default:
919	ERR ("malformed json string, neither array, object, number, string or atom");	1686	ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
920	break;	1687	break;
921	}	1688	}
922		1689
923	fail:	1690	fail:
924	return 0;	1691	return 0;
925	}	1692	}
926		1693
927	static SV *	1694	static SV *
928	decode_json (SV *string, UV flags)	1695	decode_json (SV *string, JSON *json, STRLEN *offset_return)
929	{	1696	{
		1697	dec_t dec;
930	SV *sv;	1698	SV *sv;
931		1699
932	if (flags & F_UTF8)	1700	/* work around bugs in 5.10 where manipulating magic values
		1701	* makes perl ignore the magic in subsequent accesses.
		1702	* also make a copy of non-PV values, to get them into a clean
		1703	* state (SvPV should do that, but it's buggy, see below).
		1704	*
		1705	* SvIsCOW_shared_hash works around a bug in perl (possibly 5.16),
		1706	* as reported by Reini Urban.
		1707	*/
		1708	/*SvGETMAGIC (string);*/
		1709	if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
		1710	string = sv_2mortal (newSVsv (string));
		1711
		1712	SvUPGRADE (string, SVt_PV);
		1713
		1714	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1715	* assertion with -DDEBUGGING, although SvCUR is documented to
		1716	* return the xpv_cur field which certainly exists after upgrading.
		1717	* according to nicholas clark, calling SvPOK fixes this.
		1718	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1719	* and hope for the best.
		1720	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1721	* assertion business is seriously broken, try yet another workaround
		1722	* for the broken -DDEBUGGING.
		1723	*/
		1724	{
		1725	#ifdef DEBUGGING
		1726	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1727	#else
		1728	STRLEN offset = SvCUR (string);
		1729	#endif
		1730
		1731	if (offset > json->max_size && json->max_size)
		1732	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1733	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1734	}
		1735
		1736	if (DECODE_WANTS_OCTETS (json))
933	sv_utf8_downgrade (string, 0);	1737	sv_utf8_downgrade (string, 0);
934	else	1738	else
935	sv_utf8_upgrade (string);	1739	sv_utf8_upgrade (string);
936		1740
937	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1741	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
938		1742
939	dec_t dec;	1743	dec.json = *json;
940	dec.flags = flags;
941	dec.cur = SvPVX (string);	1744	dec.cur = SvPVX (string);
942	dec.end = SvEND (string);	1745	dec.end = SvEND (string);
943	dec.err = 0;	1746	dec.err = 0;
		1747	dec.depth = 0;
944		1748
		1749	if (dec.json.cb_object || dec.json.cb_sk_object)
		1750	dec.json.flags |= F_HOOK;
		1751
		1752	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1753
		1754	decode_ws (&dec);
945	sv = decode_sv (&dec);	1755	sv = decode_sv (&dec);
946		1756
		1757	if (offset_return)
		1758	*offset_return = dec.cur - SvPVX (string);
		1759	else if (sv)
		1760	{
		1761	// check for trailing garbage
		1762	decode_ws (&dec);
		1763
		1764	if (*dec.cur)
		1765	{
		1766	dec.err = "garbage after JSON object";
		1767	SvREFCNT_dec (sv);
		1768	sv = 0;
		1769	}
		1770	}
		1771
947	if (!sv)	1772	if (!sv)
948	{	1773	{
949	IV offset = dec.flags & F_UTF8
950	? dec.cur - SvPVX (string)
951	: utf8_distance (dec.cur, SvPVX (string));
952	SV *uni = sv_newmortal ();	1774	SV *uni = sv_newmortal ();
953		1775
954	// horrible hack to silence warning inside pv_uni_display	1776	// horrible hack to silence warning inside pv_uni_display
955	COP cop = *PL_curcop;	1777	COP cop = *PL_curcop;
956	cop.cop_warnings = pWARN_NONE;	1778	cop.cop_warnings = pWARN_NONE;
…		…
958	SAVEVPTR (PL_curcop);	1780	SAVEVPTR (PL_curcop);
959	PL_curcop = &cop;	1781	PL_curcop = &cop;
960	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1782	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
961	LEAVE;	1783	LEAVE;
962		1784
963	croak ("%s, at character offset %d (%s)",	1785	croak ("%s, at character offset %d (before \"%s\")",
964	dec.err,	1786	dec.err,
965	(int)offset,	1787	(int)ptr_to_index (string, dec.cur),
966	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1788	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
967	}	1789	}
968		1790
969	sv = sv_2mortal (sv);	1791	sv = sv_2mortal (sv);
970		1792
971	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1793	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)");	1794	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
973		1795
974	return sv;	1796	return sv;
		1797	}
		1798
		1799	/////////////////////////////////////////////////////////////////////////////
		1800	// incremental parser
		1801
		1802	static void
		1803	incr_parse (JSON *self)
		1804	{
		1805	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1806
		1807	// the state machine here is a bit convoluted and could be simplified a lot
		1808	// but this would make it slower, so...
		1809
		1810	for (;;)
		1811	{
		1812	//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
		1813	switch (self->incr_mode)
		1814	{
		1815	// only used for initial whitespace skipping
		1816	case INCR_M_WS:
		1817	for (;;)
		1818	{
		1819	if (*p > 0x20)
		1820	{
		1821	if (*p == '#')
		1822	{
		1823	self->incr_mode = INCR_M_C0;
		1824	goto incr_m_c;
		1825	}
		1826	else
		1827	{
		1828	self->incr_mode = INCR_M_JSON;
		1829	goto incr_m_json;
		1830	}
		1831	}
		1832	else if (!*p)
		1833	goto interrupt;
		1834
		1835	++p;
		1836	}
		1837
		1838	// skip a single char inside a string (for \\-processing)
		1839	case INCR_M_BS:
		1840	if (!*p)
		1841	goto interrupt;
		1842
		1843	++p;
		1844	self->incr_mode = INCR_M_STR;
		1845	goto incr_m_str;
		1846
		1847	// inside #-style comments
		1848	case INCR_M_C0:
		1849	case INCR_M_C1:
		1850	incr_m_c:
		1851	for (;;)
		1852	{
		1853	if (*p == '\n')
		1854	{
		1855	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1856	break;
		1857	}
		1858	else if (!*p)
		1859	goto interrupt;
		1860
		1861	++p;
		1862	}
		1863
		1864	break;
		1865
		1866	// inside a string
		1867	case INCR_M_STR:
		1868	incr_m_str:
		1869	for (;;)
		1870	{
		1871	if (*p == '"')
		1872	{
		1873	++p;
		1874	self->incr_mode = INCR_M_JSON;
		1875
		1876	if (!self->incr_nest)
		1877	goto interrupt;
		1878
		1879	goto incr_m_json;
		1880	}
		1881	else if (*p == '\\')
		1882	{
		1883	++p; // "virtually" consumes character after \
		1884
		1885	if (!*p) // if at end of string we have to switch modes
		1886	{
		1887	self->incr_mode = INCR_M_BS;
		1888	goto interrupt;
		1889	}
		1890	}
		1891	else if (!*p)
		1892	goto interrupt;
		1893
		1894	++p;
		1895	}
		1896
		1897	// after initial ws, outside string
		1898	case INCR_M_JSON:
		1899	incr_m_json:
		1900	for (;;)
		1901	{
		1902	switch (*p++)
		1903	{
		1904	case 0:
		1905	--p;
		1906	goto interrupt;
		1907
		1908	case 0x09:
		1909	case 0x0a:
		1910	case 0x0d:
		1911	case 0x20:
		1912	if (!self->incr_nest)
		1913	{
		1914	--p; // do not eat the whitespace, let the next round do it
		1915	goto interrupt;
		1916	}
		1917	break;
		1918
		1919	case '"':
		1920	self->incr_mode = INCR_M_STR;
		1921	goto incr_m_str;
		1922
		1923	case '[':
		1924	case '{':
		1925	case '(':
		1926	if (++self->incr_nest > self->max_depth)
		1927	croak (ERR_NESTING_EXCEEDED);
		1928	break;
		1929
		1930	case ']':
		1931	case '}':
		1932	if (--self->incr_nest <= 0)
		1933	goto interrupt;
		1934	break;
		1935
		1936	case ')':
		1937	--self->incr_nest;
		1938	break;
		1939
		1940	case '#':
		1941	self->incr_mode = INCR_M_C1;
		1942	goto incr_m_c;
		1943	}
		1944	}
		1945	}
		1946
		1947	modechange:
		1948	;
		1949	}
		1950
		1951	interrupt:
		1952	self->incr_pos = p - SvPVX (self->incr_text);
		1953	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1954	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
975	}	1955	}
976		1956
977	/////////////////////////////////////////////////////////////////////////////	1957	/////////////////////////////////////////////////////////////////////////////
978	// XS interface functions	1958	// XS interface functions
979		1959
…		…
981		1961
982	BOOT:	1962	BOOT:
983	{	1963	{
984	int i;	1964	int i;
985		1965
986	memset (decode_hexdigit, 0xff, 256);
987	for (i = 10; i--; )	1966	for (i = 0; i < 256; ++i)
988	decode_hexdigit ['0' + i] = i;	1967	decode_hexdigit [i] =
		1968	i >= '0' && i <= '9' ? i - '0'
		1969	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1970	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1971	: -1;
989		1972
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);	1973	json_stash = gv_stashpv ("JSON::XS" , 1);
		1974	bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
		1975	bool_true = get_bool ("Types::Serialiser::true");
		1976	bool_false = get_bool ("Types::Serialiser::false");
		1977
		1978	sv_json = newSVpv ("JSON", 0);
		1979	SvREADONLY_on (sv_json);
		1980
		1981	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
997	}	1982	}
998		1983
999	PROTOTYPES: DISABLE	1984	PROTOTYPES: DISABLE
1000		1985
1001	SV *new (char *dummy)	1986	void CLONE (...)
1002	CODE:	1987	CODE:
1003	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1988	json_stash = 0;
		1989	bool_stash = 0;
		1990
		1991	void new (char *klass)
		1992	PPCODE:
		1993	{
		1994	SV *pv = NEWSV (0, sizeof (JSON));
		1995	SvPOK_only (pv);
		1996	json_init ((JSON *)SvPVX (pv));
		1997	XPUSHs (sv_2mortal (sv_bless (
		1998	newRV_noinc (pv),
		1999	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		2000	)));
		2001	}
		2002
		2003	void ascii (JSON *self, int enable = 1)
		2004	ALIAS:
		2005	ascii = F_ASCII
		2006	latin1 = F_LATIN1
		2007	utf8 = F_UTF8
		2008	indent = F_INDENT
		2009	canonical = F_CANONICAL
		2010	space_before = F_SPACE_BEFORE
		2011	space_after = F_SPACE_AFTER
		2012	pretty = F_PRETTY
		2013	allow_nonref = F_ALLOW_NONREF
		2014	shrink = F_SHRINK
		2015	allow_blessed = F_ALLOW_BLESSED
		2016	convert_blessed = F_CONV_BLESSED
		2017	relaxed = F_RELAXED
		2018	allow_unknown = F_ALLOW_UNKNOWN
		2019	allow_tags = F_ALLOW_TAGS
		2020	PPCODE:
		2021	{
		2022	if (enable)
		2023	self->flags |= ix;
		2024	else
		2025	self->flags &= ~ix;
		2026
		2027	XPUSHs (ST (0));
		2028	}
		2029
		2030	void get_ascii (JSON *self)
		2031	ALIAS:
		2032	get_ascii = F_ASCII
		2033	get_latin1 = F_LATIN1
		2034	get_utf8 = F_UTF8
		2035	get_indent = F_INDENT
		2036	get_canonical = F_CANONICAL
		2037	get_space_before = F_SPACE_BEFORE
		2038	get_space_after = F_SPACE_AFTER
		2039	get_allow_nonref = F_ALLOW_NONREF
		2040	get_shrink = F_SHRINK
		2041	get_allow_blessed = F_ALLOW_BLESSED
		2042	get_convert_blessed = F_CONV_BLESSED
		2043	get_relaxed = F_RELAXED
		2044	get_allow_unknown = F_ALLOW_UNKNOWN
		2045	get_allow_tags = F_ALLOW_TAGS
		2046	PPCODE:
		2047	XPUSHs (boolSV (self->flags & ix));
		2048
		2049	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		2050	PPCODE:
		2051	self->max_depth = max_depth;
		2052	XPUSHs (ST (0));
		2053
		2054	U32 get_max_depth (JSON *self)
		2055	CODE:
		2056	RETVAL = self->max_depth;
1004	OUTPUT:	2057	OUTPUT:
1005	RETVAL	2058	RETVAL
1006		2059
1007	SV *ascii (SV *self, int enable = 1)	2060	void max_size (JSON *self, U32 max_size = 0)
1008	ALIAS:	2061	PPCODE:
1009	ascii = F_ASCII	2062	self->max_size = max_size;
1010	utf8 = F_UTF8	2063	XPUSHs (ST (0));
1011	indent = F_INDENT	2064
1012	canonical = F_CANONICAL	2065	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:	2066	CODE:
1019	{	2067	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:	2068	OUTPUT:
1029	RETVAL	2069	RETVAL
1030		2070
1031	void encode (SV *self, SV *scalar)	2071	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1032	PPCODE:	2072	PPCODE:
1033	XPUSHs (encode_json (scalar, *SvJSON (self)));	2073	{
		2074	SvREFCNT_dec (self->cb_object);
		2075	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
1034		2076
1035	void decode (SV *self, SV *jsonstr)	2077	XPUSHs (ST (0));
		2078	}
		2079
		2080	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1036	PPCODE:	2081	PPCODE:
1037	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	2082	{
		2083	if (!self->cb_sk_object)
		2084	self->cb_sk_object = newHV ();
		2085
		2086	if (SvOK (cb))
		2087	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		2088	else
		2089	{
		2090	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		2091
		2092	if (!HvKEYS (self->cb_sk_object))
		2093	{
		2094	SvREFCNT_dec (self->cb_sk_object);
		2095	self->cb_sk_object = 0;
		2096	}
		2097	}
		2098
		2099	XPUSHs (ST (0));
		2100	}
		2101
		2102	void encode (JSON *self, SV *scalar)
		2103	PPCODE:
		2104	PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
		2105	XPUSHs (scalar);
		2106
		2107	void decode (JSON *self, SV *jsonstr)
		2108	PPCODE:
		2109	PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
		2110	XPUSHs (jsonstr);
		2111
		2112	void decode_prefix (JSON *self, SV *jsonstr)
		2113	PPCODE:
		2114	{
		2115	SV *sv;
		2116	STRLEN offset;
		2117	PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
		2118	EXTEND (SP, 2);
		2119	PUSHs (sv);
		2120	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset))));
		2121	}
		2122
		2123	void incr_parse (JSON *self, SV *jsonstr = 0)
		2124	PPCODE:
		2125	{
		2126	if (!self->incr_text)
		2127	self->incr_text = newSVpvn ("", 0);
		2128
		2129	/* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
		2130	if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
		2131	if (DECODE_WANTS_OCTETS (self))
		2132	{
		2133	if (self->incr_pos)
		2134	self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
		2135	(U8 *)SvPVX (self->incr_text) + self->incr_pos);
		2136
		2137	sv_utf8_downgrade (self->incr_text, 0);
		2138	}
		2139	else
		2140	{
		2141	sv_utf8_upgrade (self->incr_text);
		2142
		2143	if (self->incr_pos)
		2144	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		2145	- (U8 *)SvPVX (self->incr_text);
		2146	}
		2147
		2148	// append data, if any
		2149	if (jsonstr)
		2150	{
		2151	/* make sure both strings have same encoding */
		2152	if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
		2153	if (SvUTF8 (jsonstr))
		2154	sv_utf8_downgrade (jsonstr, 0);
		2155	else
		2156	sv_utf8_upgrade (jsonstr);
		2157
		2158	/* and then just blindly append */
		2159	{
		2160	STRLEN len;
		2161	const char *str = SvPV (jsonstr, len);
		2162	STRLEN cur = SvCUR (self->incr_text);
		2163
		2164	if (SvLEN (self->incr_text) <= cur + len)
		2165	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		2166
		2167	Move (str, SvEND (self->incr_text), len, char);
		2168	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		2169	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		2170	}
		2171	}
		2172
		2173	if (GIMME_V != G_VOID)
		2174	do
		2175	{
		2176	SV *sv;
		2177	STRLEN offset;
		2178
		2179	if (!INCR_DONE (self))
		2180	{
		2181	incr_parse (self);
		2182
		2183	if (self->incr_pos > self->max_size && self->max_size)
		2184	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		2185	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		2186
		2187	if (!INCR_DONE (self))
		2188	{
		2189	// as an optimisation, do not accumulate white space in the incr buffer
		2190	if (self->incr_mode == INCR_M_WS && self->incr_pos)
		2191	{
		2192	self->incr_pos = 0;
		2193	SvCUR_set (self->incr_text, 0);
		2194	}
		2195
		2196	break;
		2197	}
		2198	}
		2199
		2200	PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
		2201	XPUSHs (sv);
		2202
		2203	self->incr_pos -= offset;
		2204	self->incr_nest = 0;
		2205	self->incr_mode = 0;
		2206
		2207	sv_chop (self->incr_text, SvPVX (self->incr_text) + offset);
		2208	}
		2209	while (GIMME_V == G_ARRAY);
		2210	}
		2211
		2212	SV *incr_text (JSON *self)
		2213	ATTRS: lvalue
		2214	CODE:
		2215	{
		2216	if (self->incr_pos)
		2217	croak ("incr_text can not be called when the incremental parser already started parsing");
		2218
		2219	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		2220	}
		2221	OUTPUT:
		2222	RETVAL
		2223
		2224	void incr_skip (JSON *self)
		2225	CODE:
		2226	{
		2227	if (self->incr_pos)
		2228	{
		2229	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2230	self->incr_pos = 0;
		2231	self->incr_nest = 0;
		2232	self->incr_mode = 0;
		2233	}
		2234	}
		2235
		2236	void incr_reset (JSON *self)
		2237	CODE:
		2238	{
		2239	SvREFCNT_dec (self->incr_text);
		2240	self->incr_text = 0;
		2241	self->incr_pos = 0;
		2242	self->incr_nest = 0;
		2243	self->incr_mode = 0;
		2244	}
		2245
		2246	void DESTROY (JSON *self)
		2247	CODE:
		2248	SvREFCNT_dec (self->cb_sk_object);
		2249	SvREFCNT_dec (self->cb_object);
		2250	SvREFCNT_dec (self->incr_text);
1038		2251
1039	PROTOTYPES: ENABLE	2252	PROTOTYPES: ENABLE
1040		2253
1041	void to_json (SV *scalar)	2254	void encode_json (SV *scalar)
1042	PPCODE:	2255	PPCODE:
1043	XPUSHs (encode_json (scalar, F_UTF8));	2256	{
		2257	JSON json;
		2258	json_init (&json);
		2259	json.flags |= F_UTF8;
		2260	PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
		2261	XPUSHs (scalar);
		2262	}
1044		2263
1045	void from_json (SV *jsonstr)	2264	void decode_json (SV *jsonstr)
1046	PPCODE:	2265	PPCODE:
1047	XPUSHs (decode_json (jsonstr, F_UTF8));	2266	{
		2267	JSON json;
		2268	json_init (&json);
		2269	json.flags |= F_UTF8;
		2270	PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
		2271	XPUSHs (jsonstr);
		2272	}
1048		2273

Diff Legend

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