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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.24 by root, Tue Apr 3 23:59:04 2007 UTC vs.
Revision 1.46 by root, Mon Jun 25 22:11:39 2007 UTC

…		…
3	#include "XSUB.h"	3	#include "XSUB.h"
4		4
5	#include "assert.h"	5	#include "assert.h"
6	#include "string.h"	6	#include "string.h"
7	#include "stdlib.h"	7	#include "stdlib.h"
		8	#include "stdio.h"
8		9
		10	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
		11	# define snprintf _snprintf // C compilers have this in stdio.h
		12	#endif
		13
		14	// some old perls do not have this, try to make it work, no
		15	// guarentees, though. if it breaks, you get to keep the pieces.
		16	#ifndef UTF8_MAXBYTES
		17	# define UTF8_MAXBYTES 13
		18	#endif
		19
9	#define F_ASCII 0x00000001UL	20	#define F_ASCII 0x00000001UL
		21	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002UL	22	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004UL	23	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008UL	24	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010UL	25	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020UL	26	#define F_SPACE_AFTER 0x00000040UL
15	#define F_ALLOW_NONREF 0x00000080UL	27	#define F_ALLOW_NONREF 0x00000100UL
16	#define F_SHRINK 0x00000100UL	28	#define F_SHRINK 0x00000200UL
		29	#define F_ALLOW_BLESSED 0x00000400UL
		30	#define F_CONV_BLESSED 0x00000800UL // NYI
17	#define F_MAXDEPTH 0xf8000000UL	31	#define F_MAXDEPTH 0xf8000000UL
18	#define S_MAXDEPTH 27	32	#define S_MAXDEPTH 27
		33	#define F_MAXSIZE 0x01f00000UL
		34	#define S_MAXSIZE 20
19		35
20	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))	36	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
21		37	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
22	// F_SELFCONVERT? <=> to_json/toJson
23	// F_BLESSED? <=> { $__class__$ => }
24		38
25	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER	39	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER
26	#define F_DEFAULT (9UL << S_MAXDEPTH)	40	#define F_DEFAULT (9UL << S_MAXDEPTH)
27		41
28	#define INIT_SIZE 32 // initial scalar size to be allocated	42	#define INIT_SIZE 32 // initial scalar size to be allocated
29	#define INDENT_STEP 3 // spaces per indentation level	43	#define INDENT_STEP 3 // spaces per indentation level
30		44
31	#define UTF8_MAX_LEN 11 // for perls UTF-X: max. number of octets per character
32	#define SHORT_STRING_LEN 512 // special-case strings of up to this size	45	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
33		46
34	#define SB do {	47	#define SB do {
35	#define SE } while (0)	48	#define SE } while (0)
36		49
		50	#if __GNUC__ >= 3
		51	# define expect(expr,value) __builtin_expect ((expr),(value))
		52	# define inline inline
		53	#else
		54	# define expect(expr,value) (expr)
		55	# define inline static
		56	#endif
		57
		58	#define expect_false(expr) expect ((expr) != 0, 0)
		59	#define expect_true(expr) expect ((expr) != 0, 1)
		60
37	static HV *json_stash; // JSON::XS::	61	static HV json_stash, json_boolean_stash; // JSON::XS::
		62	static SV json_true, json_false;
38		63
39	/////////////////////////////////////////////////////////////////////////////	64	/////////////////////////////////////////////////////////////////////////////
40	// utility functions	65	// utility functions
41		66
42	static UV *	67	static UV *
…		…
65	// decode an utf-8 character and return it, or (UV)-1 in	90	// decode an utf-8 character and return it, or (UV)-1 in
66	// case of an error.	91	// case of an error.
67	// we special-case "safe" characters from U+80 .. U+7FF,	92	// we special-case "safe" characters from U+80 .. U+7FF,
68	// but use the very good perl function to parse anything else.	93	// but use the very good perl function to parse anything else.
69	// note that we never call this function for a ascii codepoints	94	// note that we never call this function for a ascii codepoints
70	static UV	95	inline UV
71	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	96	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
72	{	97	{
73	if (s[0] > 0xdf \|\| s[0] < 0xc2)	98	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))
74	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	99	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
75	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	100	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
76	{	101	{
77	*clen = 2;	102	*clen = 2;
78	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);	103	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
…		…
96	U32 flags; // F_*	121	U32 flags; // F_*
97	U32 indent; // indentation level	122	U32 indent; // indentation level
98	U32 maxdepth; // max. indentation/recursion level	123	U32 maxdepth; // max. indentation/recursion level
99	} enc_t;	124	} enc_t;
100		125
101	static void	126	inline void
102	need (enc_t *enc, STRLEN len)	127	need (enc_t *enc, STRLEN len)
103	{	128	{
104	if (enc->cur + len >= enc->end)	129	if (expect_false (enc->cur + len >= enc->end))
105	{	130	{
106	STRLEN cur = enc->cur - SvPVX (enc->sv);	131	STRLEN cur = enc->cur - SvPVX (enc->sv);
107	SvGROW (enc->sv, cur + len + 1);	132	SvGROW (enc->sv, cur + len + 1);
108	enc->cur = SvPVX (enc->sv) + cur;	133	enc->cur = SvPVX (enc->sv) + cur;
109	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	134	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
110	}	135	}
111	}	136	}
112		137
113	static void	138	inline void
114	encode_ch (enc_t *enc, char ch)	139	encode_ch (enc_t *enc, char ch)
115	{	140	{
116	need (enc, 1);	141	need (enc, 1);
117	*enc->cur++ = ch;	142	*enc->cur++ = ch;
118	}	143	}
…		…
126		151
127	while (str < end)	152	while (str < end)
128	{	153	{
129	unsigned char ch = (unsigned char )str;	154	unsigned char ch = (unsigned char )str;
130		155
131	if (ch >= 0x20 && ch < 0x80) // most common case	156	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
132	{	157	{
133	if (ch == '"') // but with slow exceptions	158	if (expect_false (ch == '"')) // but with slow exceptions
134	{	159	{
135	need (enc, len += 1);	160	need (enc, len += 1);
136	*enc->cur++ = '\\';	161	*enc->cur++ = '\\';
137	*enc->cur++ = '"';	162	*enc->cur++ = '"';
138	}	163	}
139	else if (ch == '\\')	164	else if (expect_false (ch == '\\'))
140	{	165	{
141	need (enc, len += 1);	166	need (enc, len += 1);
142	*enc->cur++ = '\\';	167	*enc->cur++ = '\\';
143	*enc->cur++ = '\\';	168	*enc->cur++ = '\\';
144	}	169	}
…		…
162	STRLEN clen;	187	STRLEN clen;
163	UV uch;	188	UV uch;
164		189
165	if (is_utf8)	190	if (is_utf8)
166	{	191	{
167	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
168	uch = decode_utf8 (str, end - str, &clen);	192	uch = decode_utf8 (str, end - str, &clen);
169	if (clen == (STRLEN)-1)	193	if (clen == (STRLEN)-1)
170	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	194	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
171	}	195	}
172	else	196	else
…		…
176	}	200	}
177		201
178	if (uch > 0x10FFFFUL)	202	if (uch > 0x10FFFFUL)
179	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);	203	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
180		204
181	if (uch < 0x80 \|\| enc->flags & F_ASCII)	205	if (uch < 0x80 \|\| enc->flags & F_ASCII \|\| (enc->flags & F_LATIN1 && uch > 0xFF))
182	{	206	{
183	if (uch > 0xFFFFUL)	207	if (uch > 0xFFFFUL)
184	{	208	{
185	need (enc, len += 11);	209	need (enc, len += 11);
186	sprintf (enc->cur, "\\u%04x\\u%04x",	210	sprintf (enc->cur, "\\u%04x\\u%04x",
…		…
200	*enc->cur++ = hexdigit [(uch >> 0) & 15];	224	*enc->cur++ = hexdigit [(uch >> 0) & 15];
201	}	225	}
202		226
203	str += clen;	227	str += clen;
204	}	228	}
		229	else if (enc->flags & F_LATIN1)
		230	{
		231	*enc->cur++ = uch;
		232	str += clen;
		233	}
205	else if (is_utf8)	234	else if (is_utf8)
206	{	235	{
207	need (enc, len += clen);	236	need (enc, len += clen);
208	do	237	do
209	{	238	{
…		…
211	}	240	}
212	while (--clen);	241	while (--clen);
213	}	242	}
214	else	243	else
215	{	244	{
216	need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed	245	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
217	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	246	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
218	++str;	247	++str;
219	}	248	}
220	}	249	}
221	}	250	}
…		…
223		252
224	--len;	253	--len;
225	}	254	}
226	}	255	}
227		256
228	static void	257	inline void
229	encode_indent (enc_t *enc)	258	encode_indent (enc_t *enc)
230	{	259	{
231	if (enc->flags & F_INDENT)	260	if (enc->flags & F_INDENT)
232	{	261	{
233	int spaces = enc->indent * INDENT_STEP;	262	int spaces = enc->indent * INDENT_STEP;
…		…
236	memset (enc->cur, ' ', spaces);	265	memset (enc->cur, ' ', spaces);
237	enc->cur += spaces;	266	enc->cur += spaces;
238	}	267	}
239	}	268	}
240		269
241	static void	270	inline void
242	encode_space (enc_t *enc)	271	encode_space (enc_t *enc)
243	{	272	{
244	need (enc, 1);	273	need (enc, 1);
245	encode_ch (enc, ' ');	274	encode_ch (enc, ' ');
246	}	275	}
247		276
248	static void	277	inline void
249	encode_nl (enc_t *enc)	278	encode_nl (enc_t *enc)
250	{	279	{
251	if (enc->flags & F_INDENT)	280	if (enc->flags & F_INDENT)
252	{	281	{
253	need (enc, 1);	282	need (enc, 1);
254	encode_ch (enc, '\n');	283	encode_ch (enc, '\n');
255	}	284	}
256	}	285	}
257		286
258	static void	287	inline void
259	encode_comma (enc_t *enc)	288	encode_comma (enc_t *enc)
260	{	289	{
261	encode_ch (enc, ',');	290	encode_ch (enc, ',');
262		291
263	if (enc->flags & F_INDENT)	292	if (enc->flags & F_INDENT)
…		…
362	// actually, this is mostly due to the stupid so-called	391	// actually, this is mostly due to the stupid so-called
363	// security workaround added somewhere in 5.8.x.	392	// security workaround added somewhere in 5.8.x.
364	// that randomises hash orderings	393	// that randomises hash orderings
365	if (enc->flags & F_CANONICAL)	394	if (enc->flags & F_CANONICAL)
366	{	395	{
367	HE he, hes [count]; // if your compiler dies here, you need to enable C99 mode
368	int fast = 1;	396	int fast = 1;
		397	HE *he;
		398	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
		399	HE *hes = _alloca (count sizeof (HE));
		400	#else
		401	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		402	#endif
369		403
370	i = 0;	404	i = 0;
371	while ((he = hv_iternext (hv)))	405	while ((he = hv_iternext (hv)))
372	{	406	{
373	hes [i++] = he;	407	hes [i++] = he;
…		…
408		442
409	encode_nl (enc);	443	encode_nl (enc);
410	}	444	}
411	else	445	else
412	{	446	{
413	SV *sv;
414	HE *he = hv_iternext (hv);	447	HE *he = hv_iternext (hv);
415		448
416	for (;;)	449	for (;;)
417	{	450	{
418	encode_indent (enc);	451	encode_indent (enc);
…		…
433		466
434	// encode objects, arrays and special \0=false and \1=true values.	467	// encode objects, arrays and special \0=false and \1=true values.
435	static void	468	static void
436	encode_rv (enc_t enc, SV sv)	469	encode_rv (enc_t enc, SV sv)
437	{	470	{
		471	svtype svt;
		472
438	SvGETMAGIC (sv);	473	SvGETMAGIC (sv);
439
440	svtype svt = SvTYPE (sv);	474	svt = SvTYPE (sv);
441		475
		476	if (expect_false (SvOBJECT (sv)))
		477	{
		478	if (SvSTASH (sv) == json_boolean_stash)
		479	{
		480	if (SvIV (sv) == 0)
		481	encode_str (enc, "false", 5, 0);
		482	else
		483	encode_str (enc, "true", 4, 0);
		484	}
		485	else
		486	{
		487	#if 0
		488	if (0 && sv_derived_from (rv, "JSON::Literal"))
		489	{
		490	// not yet
		491	}
		492	#endif
		493	if (enc->flags & F_CONV_BLESSED)
		494	{
		495	// we re-bless the reference to get overload and other niceties right
		496	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1);
		497
		498	if (to_json)
		499	{
		500	dSP;
		501	ENTER;
		502	SAVETMPS;
		503	PUSHMARK (SP);
		504	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		505
		506	// calling with G_SCALAR ensures that we always get a 1 reutrn value
		507	// check anyways.
		508	PUTBACK;
		509	assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR));
		510	SPAGAIN;
		511
		512	encode_sv (enc, POPs);
		513
		514	FREETMPS;
		515	LEAVE;
		516	}
		517	else if (enc->flags & F_ALLOW_BLESSED)
		518	encode_str (enc, "null", 4, 0);
		519	else
		520	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		521	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		522	}
		523	else if (enc->flags & F_ALLOW_BLESSED)
		524	encode_str (enc, "null", 4, 0);
		525	else
		526	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		527	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		528	}
		529	}
442	if (svt == SVt_PVHV)	530	else if (svt == SVt_PVHV)
443	encode_hv (enc, (HV *)sv);	531	encode_hv (enc, (HV *)sv);
444	else if (svt == SVt_PVAV)	532	else if (svt == SVt_PVAV)
445	encode_av (enc, (AV *)sv);	533	encode_av (enc, (AV *)sv);
446	else if (svt < SVt_PVAV)	534	else if (svt < SVt_PVAV)
447	{	535	{
…		…
471	encode_str (enc, str, len, SvUTF8 (sv));	559	encode_str (enc, str, len, SvUTF8 (sv));
472	encode_ch (enc, '"');	560	encode_ch (enc, '"');
473	}	561	}
474	else if (SvNOKp (sv))	562	else if (SvNOKp (sv))
475	{	563	{
		564	// trust that perl will do the right thing w.r.t. JSON syntax.
476	need (enc, NV_DIG + 32);	565	need (enc, NV_DIG + 32);
477	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	566	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
478	enc->cur += strlen (enc->cur);	567	enc->cur += strlen (enc->cur);
479	}	568	}
480	else if (SvIOKp (sv))	569	else if (SvIOKp (sv))
481	{	570	{
482	need (enc, 64);	571	// we assume we can always read an IV as a UV
		572	if (SvUV (sv) & ~(UV)0x7fff)
		573	{
		574	// large integer, use the (rather slow) snprintf way.
		575	need (enc, sizeof (UV) * 3);
483	enc->cur +=	576	enc->cur +=
484	SvIsUV(sv)	577	SvIsUV(sv)
485	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	578	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
486	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	579	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		580	}
		581	else
		582	{
		583	// optimise the "small number case"
		584	// code will likely be branchless and use only a single multiplication
		585	I32 i = SvIV (sv);
		586	U32 u;
		587	char digit, nz = 0;
		588
		589	need (enc, 6);
		590
		591	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		592	u = i < 0 ? -i : i;
		593
		594	// convert to 4.28 fixed-point representation
		595	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		596
		597	// now output digit by digit, each time masking out the integer part
		598	// and multiplying by 5 while moving the decimal point one to the right,
		599	// resulting in a net multiplication by 10.
		600	// we always write the digit to memory but conditionally increment
		601	// the pointer, to ease the usage of conditional move instructions.
		602	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffff) 5;
		603	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffff) 5;
		604	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffff) 5;
		605	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffff) 5;
		606	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		607	}
487	}	608	}
488	else if (SvROK (sv))	609	else if (SvROK (sv))
489	encode_rv (enc, SvRV (sv));	610	encode_rv (enc, SvRV (sv));
490	else if (!SvOK (sv))	611	else if (!SvOK (sv))
491	encode_str (enc, "null", 4, 0);	612	encode_str (enc, "null", 4, 0);
…		…
495	}	616	}
496		617
497	static SV *	618	static SV *
498	encode_json (SV *scalar, U32 flags)	619	encode_json (SV *scalar, U32 flags)
499	{	620	{
		621	enc_t enc;
		622
500	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	623	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
501	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	624	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
502		625
503	enc_t enc;
504	enc.flags = flags;	626	enc.flags = flags;
505	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	627	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
506	enc.cur = SvPVX (enc.sv);	628	enc.cur = SvPVX (enc.sv);
507	enc.end = SvEND (enc.sv);	629	enc.end = SvEND (enc.sv);
508	enc.indent = 0;	630	enc.indent = 0;
509	enc.maxdepth = DEC_DEPTH (flags);	631	enc.maxdepth = DEC_DEPTH (flags);
510		632
511	SvPOK_only (enc.sv);	633	SvPOK_only (enc.sv);
512	encode_sv (&enc, scalar);	634	encode_sv (&enc, scalar);
513		635
		636	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		637	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		638
514	if (!(flags & (F_ASCII \| F_UTF8)))	639	if (!(flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))
515	SvUTF8_on (enc.sv);	640	SvUTF8_on (enc.sv);
516
517	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
518		641
519	if (enc.flags & F_SHRINK)	642	if (enc.flags & F_SHRINK)
520	shrink (enc.sv);	643	shrink (enc.sv);
521		644
522	return enc.sv;	645	return enc.sv;
…		…
534	U32 flags; // F_*	657	U32 flags; // F_*
535	U32 depth; // recursion depth	658	U32 depth; // recursion depth
536	U32 maxdepth; // recursion depth limit	659	U32 maxdepth; // recursion depth limit
537	} dec_t;	660	} dec_t;
538		661
539	static void	662	inline void
540	decode_ws (dec_t *dec)	663	decode_ws (dec_t *dec)
541	{	664	{
542	for (;;)	665	for (;;)
543	{	666	{
544	char ch = *dec->cur;	667	char ch = *dec->cur;
…		…
570	decode_4hex (dec_t *dec)	693	decode_4hex (dec_t *dec)
571	{	694	{
572	signed char d1, d2, d3, d4;	695	signed char d1, d2, d3, d4;
573	unsigned char cur = (unsigned char )dec->cur;	696	unsigned char cur = (unsigned char )dec->cur;
574		697
575	d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected");	698	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
576	d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected");	699	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
577	d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected");	700	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
578	d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected");	701	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
579		702
580	dec->cur += 4;	703	dec->cur += 4;
581		704
582	return ((UV)d1) << 12	705	return ((UV)d1) << 12
583	\| ((UV)d2) << 8	706	\| ((UV)d2) << 8
…		…
591	static SV *	714	static SV *
592	decode_str (dec_t *dec)	715	decode_str (dec_t *dec)
593	{	716	{
594	SV *sv = 0;	717	SV *sv = 0;
595	int utf8 = 0;	718	int utf8 = 0;
		719	char *dec_cur = dec->cur;
596		720
597	do	721	do
598	{	722	{
599	char buf [SHORT_STRING_LEN + UTF8_MAX_LEN];	723	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
600	char *cur = buf;	724	char *cur = buf;
601		725
602	do	726	do
603	{	727	{
604	unsigned char ch = (unsigned char )dec->cur++;	728	unsigned char ch = (unsigned char )dec_cur++;
605		729
606	if (ch == '"')	730	if (expect_false (ch == '"'))
607	{	731	{
608	--dec->cur;	732	--dec_cur;
609	break;	733	break;
610	}	734	}
611	else if (ch == '\\')	735	else if (expect_false (ch == '\\'))
612	{	736	{
613	switch (*dec->cur)	737	switch (*dec_cur)
614	{	738	{
615	case '\\':	739	case '\\':
616	case '/':	740	case '/':
617	case '"': cur++ = dec->cur++; break;	741	case '"': cur++ = dec_cur++; break;
618		742
619	case 'b': ++dec->cur; *cur++ = '\010'; break;	743	case 'b': ++dec_cur; *cur++ = '\010'; break;
620	case 't': ++dec->cur; *cur++ = '\011'; break;	744	case 't': ++dec_cur; *cur++ = '\011'; break;
621	case 'n': ++dec->cur; *cur++ = '\012'; break;	745	case 'n': ++dec_cur; *cur++ = '\012'; break;
622	case 'f': ++dec->cur; *cur++ = '\014'; break;	746	case 'f': ++dec_cur; *cur++ = '\014'; break;
623	case 'r': ++dec->cur; *cur++ = '\015'; break;	747	case 'r': ++dec_cur; *cur++ = '\015'; break;
624		748
625	case 'u':	749	case 'u':
626	{	750	{
627	UV lo, hi;	751	UV lo, hi;
628	++dec->cur;	752	++dec_cur;
629		753
		754	dec->cur = dec_cur;
630	hi = decode_4hex (dec);	755	hi = decode_4hex (dec);
		756	dec_cur = dec->cur;
631	if (hi == (UV)-1)	757	if (hi == (UV)-1)
632	goto fail;	758	goto fail;
633		759
634	// possibly a surrogate pair	760	// possibly a surrogate pair
635	if (hi >= 0xd800)	761	if (hi >= 0xd800)
636	if (hi < 0xdc00)	762	if (hi < 0xdc00)
637	{	763	{
638	if (dec->cur [0] != '\\' \|\| dec->cur [1] != 'u')	764	if (dec_cur [0] != '\\' \|\| dec_cur [1] != 'u')
639	ERR ("missing low surrogate character in surrogate pair");	765	ERR ("missing low surrogate character in surrogate pair");
640		766
641	dec->cur += 2;	767	dec_cur += 2;
642		768
		769	dec->cur = dec_cur;
643	lo = decode_4hex (dec);	770	lo = decode_4hex (dec);
		771	dec_cur = dec->cur;
644	if (lo == (UV)-1)	772	if (lo == (UV)-1)
645	goto fail;	773	goto fail;
646		774
647	if (lo < 0xdc00 \|\| lo >= 0xe000)	775	if (lo < 0xdc00 \|\| lo >= 0xe000)
648	ERR ("surrogate pair expected");	776	ERR ("surrogate pair expected");
…		…
662	*cur++ = hi;	790	*cur++ = hi;
663	}	791	}
664	break;	792	break;
665		793
666	default:	794	default:
667	--dec->cur;	795	--dec_cur;
668	ERR ("illegal backslash escape sequence in string");	796	ERR ("illegal backslash escape sequence in string");
669	}	797	}
670	}	798	}
671	else if (ch >= 0x20 && ch <= 0x7f)	799	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
672	*cur++ = ch;	800	*cur++ = ch;
673	else if (ch >= 0x80)	801	else if (ch >= 0x80)
674	{	802	{
675	--dec->cur;
676
677	STRLEN clen;	803	STRLEN clen;
		804	UV uch;
		805
		806	--dec_cur;
		807
678	UV uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	808	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
679	if (clen == (STRLEN)-1)	809	if (clen == (STRLEN)-1)
680	ERR ("malformed UTF-8 character in JSON string");	810	ERR ("malformed UTF-8 character in JSON string");
681		811
682	do	812	do
683	cur++ = dec->cur++;	813	cur++ = dec_cur++;
684	while (--clen);	814	while (--clen);
685		815
686	utf8 = 1;	816	utf8 = 1;
687	}	817	}
688	else	818	else
689	{	819	{
690	--dec->cur;	820	--dec_cur;
691		821
692	if (!ch)	822	if (!ch)
693	ERR ("unexpected end of string while parsing JSON string");	823	ERR ("unexpected end of string while parsing JSON string");
694	else	824	else
695	ERR ("invalid character encountered while parsing JSON string");	825	ERR ("invalid character encountered while parsing JSON string");
696	}	826	}
697	}	827	}
698	while (cur < buf + SHORT_STRING_LEN);	828	while (cur < buf + SHORT_STRING_LEN);
699		829
		830	{
700	STRLEN len = cur - buf;	831	STRLEN len = cur - buf;
701		832
702	if (sv)	833	if (sv)
703	{	834	{
704	SvGROW (sv, SvCUR (sv) + len + 1);	835	SvGROW (sv, SvCUR (sv) + len + 1);
705	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	836	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
706	SvCUR_set (sv, SvCUR (sv) + len);	837	SvCUR_set (sv, SvCUR (sv) + len);
707	}	838	}
708	else	839	else
709	sv = newSVpvn (buf, len);	840	sv = newSVpvn (buf, len);
710	}	841	}
		842	}
711	while (*dec->cur != '"');	843	while (*dec_cur != '"');
712		844
713	++dec->cur;	845	++dec_cur;
714		846
715	if (sv)	847	if (sv)
716	{	848	{
717	SvPOK_only (sv);	849	SvPOK_only (sv);
718	*SvEND (sv) = 0;	850	*SvEND (sv) = 0;
…		…
721	SvUTF8_on (sv);	853	SvUTF8_on (sv);
722	}	854	}
723	else	855	else
724	sv = newSVpvn ("", 0);	856	sv = newSVpvn ("", 0);
725		857
		858	dec->cur = dec_cur;
726	return sv;	859	return sv;
727		860
728	fail:	861	fail:
		862	dec->cur = dec_cur;
729	return 0;	863	return 0;
730	}	864	}
731		865
732	static SV *	866	static SV *
733	decode_num (dec_t *dec)	867	decode_num (dec_t *dec)
…		…
791	is_nv = 1;	925	is_nv = 1;
792	}	926	}
793		927
794	if (!is_nv)	928	if (!is_nv)
795	{	929	{
796	UV uv;	930	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
797	int numtype = grok_number (start, dec->cur - start, &uv);	931	if (*start == '-')
798	if (numtype & IS_NUMBER_IN_UV)	932	switch (dec->cur - start)
799	if (numtype & IS_NUMBER_NEG)
800	{	933	{
801	if (uv < (UV)IV_MIN)	934	case 2: return newSViv (-( start [1] - '0' * 1));
802	return newSViv (-(IV)uv);	935	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		936	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		937	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
803	}	938	}
		939	else
		940	switch (dec->cur - start)
		941	{
		942	case 1: return newSViv ( start [0] - '0' * 1);
		943	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		944	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		945	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		946	}
		947
		948	{
		949	UV uv;
		950	int numtype = grok_number (start, dec->cur - start, &uv);
		951	if (numtype & IS_NUMBER_IN_UV)
		952	if (numtype & IS_NUMBER_NEG)
		953	{
		954	if (uv < (UV)IV_MIN)
		955	return newSViv (-(IV)uv);
		956	}
804	else	957	else
805	return newSVuv (uv);	958	return newSVuv (uv);
		959
		960	// here would likely be the place for bigint support
806	}	961	}
		962	}
807		963
		964	// if we ever support bigint or bigfloat, this is the place for bigfloat
808	return newSVnv (Atof (start));	965	return newSVnv (Atof (start));
809		966
810	fail:	967	fail:
811	return 0;	968	return 0;
812	}	969	}
…		…
866	if (*dec->cur == '}')	1023	if (*dec->cur == '}')
867	++dec->cur;	1024	++dec->cur;
868	else	1025	else
869	for (;;)	1026	for (;;)
870	{	1027	{
871	SV key, value;
872
873	decode_ws (dec); EXPECT_CH ('"');	1028	decode_ws (dec); EXPECT_CH ('"');
874		1029
875	key = decode_str (dec);	1030	// heuristic: assume that
876	if (!key)	1031	// a) decode_str + hv_store_ent are abysmally slow
877	goto fail;	1032	// b) most hash keys are short, simple ascii text
		1033	// so try to "fast-match" such strings to avoid
		1034	// the overhead of hv_store_ent.
		1035	{
		1036	SV *value;
		1037	char *p = dec->cur;
		1038	char *e = p + 24; // only try up to 24 bytes
878		1039
879	decode_ws (dec); EXPECT_CH (':');	1040	for (;;)
880
881	value = decode_sv (dec);
882	if (!value)
883	{	1041	{
		1042	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
		1043	{
		1044	// slow path, back up and use decode_str
		1045	SV *key = decode_str (dec);
		1046	if (!key)
		1047	goto fail;
		1048
		1049	decode_ws (dec); EXPECT_CH (':');
		1050
		1051	value = decode_sv (dec);
		1052	if (!value)
		1053	{
		1054	SvREFCNT_dec (key);
		1055	goto fail;
		1056	}
		1057
		1058	hv_store_ent (hv, key, value, 0);
884	SvREFCNT_dec (key);	1059	SvREFCNT_dec (key);
		1060
		1061	break;
		1062	}
		1063	else if (*p == '"')
		1064	{
		1065	// fast path, got a simple key
		1066	char *key = dec->cur;
		1067	int len = p - key;
		1068	dec->cur = p + 1;
		1069
		1070	decode_ws (dec); EXPECT_CH (':');
		1071
		1072	value = decode_sv (dec);
		1073	if (!value)
885	goto fail;	1074	goto fail;
		1075
		1076	hv_store (hv, key, len, value, 0);
		1077
		1078	break;
		1079	}
		1080
		1081	++p;
886	}	1082	}
887		1083	}
888	hv_store_ent (hv, key, value, 0);
889	SvREFCNT_dec (key);
890		1084
891	decode_ws (dec);	1085	decode_ws (dec);
892		1086
893	if (*dec->cur == '}')	1087	if (*dec->cur == '}')
894	{	1088	{
…		…
913		1107
914	static SV *	1108	static SV *
915	decode_sv (dec_t *dec)	1109	decode_sv (dec_t *dec)
916	{	1110	{
917	decode_ws (dec);	1111	decode_ws (dec);
		1112
		1113	// the beauty of JSON: you need exactly one character lookahead
		1114	// to parse anything.
918	switch (*dec->cur)	1115	switch (*dec->cur)
919	{	1116	{
920	case '"': ++dec->cur; return decode_str (dec);	1117	case '"': ++dec->cur; return decode_str (dec);
921	case '[': ++dec->cur; return decode_av (dec);	1118	case '[': ++dec->cur; return decode_av (dec);
922	case '{': ++dec->cur; return decode_hv (dec);	1119	case '{': ++dec->cur; return decode_hv (dec);
…		…
928		1125
929	case 't':	1126	case 't':
930	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1127	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
931	{	1128	{
932	dec->cur += 4;	1129	dec->cur += 4;
933	return newSViv (1);	1130	return SvREFCNT_inc (json_true);
934	}	1131	}
935	else	1132	else
936	ERR ("'true' expected");	1133	ERR ("'true' expected");
937		1134
938	break;	1135	break;
939		1136
940	case 'f':	1137	case 'f':
941	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1138	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
942	{	1139	{
943	dec->cur += 5;	1140	dec->cur += 5;
944	return newSViv (0);	1141	return SvREFCNT_inc (json_false);
945	}	1142	}
946	else	1143	else
947	ERR ("'false' expected");	1144	ERR ("'false' expected");
948		1145
949	break;	1146	break;
…		…
967	fail:	1164	fail:
968	return 0;	1165	return 0;
969	}	1166	}
970		1167
971	static SV *	1168	static SV *
972	decode_json (SV *string, U32 flags)	1169	decode_json (SV string, U32 flags, UV offset_return)
973	{	1170	{
		1171	dec_t dec;
		1172	UV offset;
974	SV *sv;	1173	SV *sv;
975		1174
		1175	SvGETMAGIC (string);
976	SvUPGRADE (string, SVt_PV);	1176	SvUPGRADE (string, SVt_PV);
		1177
		1178	if (flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (flags))
		1179	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1180	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (flags));
977		1181
978	if (flags & F_UTF8)	1182	if (flags & F_UTF8)
979	sv_utf8_downgrade (string, 0);	1183	sv_utf8_downgrade (string, 0);
980	else	1184	else
981	sv_utf8_upgrade (string);	1185	sv_utf8_upgrade (string);
982		1186
983	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1187	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
984		1188
985	dec_t dec;
986	dec.flags = flags;	1189	dec.flags = flags;
987	dec.cur = SvPVX (string);	1190	dec.cur = SvPVX (string);
988	dec.end = SvEND (string);	1191	dec.end = SvEND (string);
989	dec.err = 0;	1192	dec.err = 0;
990	dec.depth = 0;	1193	dec.depth = 0;
991	dec.maxdepth = DEC_DEPTH (dec.flags);	1194	dec.maxdepth = DEC_DEPTH (dec.flags);
992		1195
993	*dec.end = 0; // this should basically be a nop, too, but make sure its there	1196	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
994	sv = decode_sv (&dec);	1197	sv = decode_sv (&dec);
995		1198
		1199	if (!(offset_return \|\| !sv))
		1200	{
		1201	// check for trailing garbage
		1202	decode_ws (&dec);
		1203
		1204	if (*dec.cur)
		1205	{
		1206	dec.err = "garbage after JSON object";
		1207	SvREFCNT_dec (sv);
		1208	sv = 0;
		1209	}
		1210	}
		1211
		1212	if (offset_return \|\| !sv)
		1213	{
		1214	offset = dec.flags & F_UTF8
		1215	? dec.cur - SvPVX (string)
		1216	: utf8_distance (dec.cur, SvPVX (string));
		1217
		1218	if (offset_return)
		1219	*offset_return = offset;
		1220	}
		1221
996	if (!sv)	1222	if (!sv)
997	{	1223	{
998	IV offset = dec.flags & F_UTF8
999	? dec.cur - SvPVX (string)
1000	: utf8_distance (dec.cur, SvPVX (string));
1001	SV *uni = sv_newmortal ();	1224	SV *uni = sv_newmortal ();
1002		1225
1003	// horrible hack to silence warning inside pv_uni_display	1226	// horrible hack to silence warning inside pv_uni_display
1004	COP cop = *PL_curcop;	1227	COP cop = *PL_curcop;
1005	cop.cop_warnings = pWARN_NONE;	1228	cop.cop_warnings = pWARN_NONE;
…		…
1030		1253
1031	BOOT:	1254	BOOT:
1032	{	1255	{
1033	int i;	1256	int i;
1034		1257
1035	memset (decode_hexdigit, 0xff, 256);
1036
1037	for (i = 0; i < 256; ++i)	1258	for (i = 0; i < 256; ++i)
1038	decode_hexdigit [i] =	1259	decode_hexdigit [i] =
1039	i >= '0' && i <= '9' ? i - '0'	1260	i >= '0' && i <= '9' ? i - '0'
1040	: i >= 'a' && i <= 'f' ? i - 'a' + 10	1261	: i >= 'a' && i <= 'f' ? i - 'a' + 10
1041	: i >= 'A' && i <= 'F' ? i - 'A' + 10	1262	: i >= 'A' && i <= 'F' ? i - 'A' + 10
1042	: -1;	1263	: -1;
1043		1264
1044	json_stash = gv_stashpv ("JSON::XS", 1);	1265	json_stash = gv_stashpv ("JSON::XS" , 1);
		1266	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1267
		1268	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1269	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
1045	}	1270	}
1046		1271
1047	PROTOTYPES: DISABLE	1272	PROTOTYPES: DISABLE
1048		1273
1049	SV new (char dummy)	1274	SV new (char dummy)
…		…
1052	OUTPUT:	1277	OUTPUT:
1053	RETVAL	1278	RETVAL
1054		1279
1055	SV ascii (SV self, int enable = 1)	1280	SV ascii (SV self, int enable = 1)
1056	ALIAS:	1281	ALIAS:
1057	ascii = F_ASCII	1282	ascii = F_ASCII
		1283	latin1 = F_LATIN1
1058	utf8 = F_UTF8	1284	utf8 = F_UTF8
1059	indent = F_INDENT	1285	indent = F_INDENT
1060	canonical = F_CANONICAL	1286	canonical = F_CANONICAL
1061	space_before = F_SPACE_BEFORE	1287	space_before = F_SPACE_BEFORE
1062	space_after = F_SPACE_AFTER	1288	space_after = F_SPACE_AFTER
1063	pretty = F_PRETTY	1289	pretty = F_PRETTY
1064	allow_nonref = F_ALLOW_NONREF	1290	allow_nonref = F_ALLOW_NONREF
1065	shrink = F_SHRINK	1291	shrink = F_SHRINK
		1292	allow_blessed = F_ALLOW_BLESSED
		1293	convert_blessed = F_CONV_BLESSED
1066	CODE:	1294	CODE:
1067	{	1295	{
1068	UV *uv = SvJSON (self);	1296	UV *uv = SvJSON (self);
1069	if (enable)	1297	if (enable)
1070	*uv \|= ix;	1298	*uv \|= ix;
…		…
1074	RETVAL = newSVsv (self);	1302	RETVAL = newSVsv (self);
1075	}	1303	}
1076	OUTPUT:	1304	OUTPUT:
1077	RETVAL	1305	RETVAL
1078		1306
1079	SV max_depth (SV self, int max_depth = 0x80000000UL)	1307	SV max_depth (SV self, UV max_depth = 0x80000000UL)
1080	CODE:	1308	CODE:
1081	{	1309	{
1082	UV *uv = SvJSON (self);	1310	UV *uv = SvJSON (self);
1083	UV log2 = 0;	1311	UV log2 = 0;
1084		1312
…		…
1086		1314
1087	while ((1UL << log2) < max_depth)	1315	while ((1UL << log2) < max_depth)
1088	++log2;	1316	++log2;
1089		1317
1090	uv = uv & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);	1318	uv = uv & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
		1319
		1320	RETVAL = newSVsv (self);
		1321	}
		1322	OUTPUT:
		1323	RETVAL
		1324
		1325	SV max_size (SV self, UV max_size = 0)
		1326	CODE:
		1327	{
		1328	UV *uv = SvJSON (self);
		1329	UV log2 = 0;
		1330
		1331	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1332	if (max_size == 1) max_size = 2;
		1333
		1334	while ((1UL << log2) < max_size)
		1335	++log2;
		1336
		1337	uv = uv & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
1091		1338
1092	RETVAL = newSVsv (self);	1339	RETVAL = newSVsv (self);
1093	}	1340	}
1094	OUTPUT:	1341	OUTPUT:
1095	RETVAL	1342	RETVAL
…		…
1098	PPCODE:	1345	PPCODE:
1099	XPUSHs (encode_json (scalar, *SvJSON (self)));	1346	XPUSHs (encode_json (scalar, *SvJSON (self)));
1100		1347
1101	void decode (SV self, SV jsonstr)	1348	void decode (SV self, SV jsonstr)
1102	PPCODE:	1349	PPCODE:
1103	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1350	XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));
		1351
		1352	void decode_prefix (SV self, SV jsonstr)
		1353	PPCODE:
		1354	{
		1355	UV offset;
		1356	EXTEND (SP, 2);
		1357	PUSHs (decode_json (jsonstr, *SvJSON (self), &offset));
		1358	PUSHs (sv_2mortal (newSVuv (offset)));
		1359	}
1104		1360
1105	PROTOTYPES: ENABLE	1361	PROTOTYPES: ENABLE
1106		1362
1107	void to_json (SV *scalar)	1363	void to_json (SV *scalar)
1108	ALIAS:	1364	ALIAS:
…		…
1112		1368
1113	void from_json (SV *jsonstr)	1369	void from_json (SV *jsonstr)
1114	ALIAS:	1370	ALIAS:
1115	jsonToObj = 0	1371	jsonToObj = 0
1116	PPCODE:	1372	PPCODE:
1117	XPUSHs (decode_json (jsonstr, F_DEFAULT \| F_UTF8));	1373	XPUSHs (decode_json (jsonstr, F_DEFAULT \| F_UTF8, 0));
1118		1374

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.24 by root, Tue Apr 3 23:59:04 2007 UTC vs. Revision 1.46 by root, Mon Jun 25 22:11:39 2007 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.24 by root, Tue Apr 3 23:59:04 2007 UTC vs.
Revision 1.46 by root, Mon Jun 25 22:11:39 2007 UTC