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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.57 by root, Mon Aug 13 16:05:42 2007 UTC vs.
Revision 1.74 by root, Wed Mar 19 15:17:53 2008 UTC

…		…
27	#define F_SPACE_AFTER 0x00000040UL	27	#define F_SPACE_AFTER 0x00000040UL
28	#define F_ALLOW_NONREF 0x00000100UL	28	#define F_ALLOW_NONREF 0x00000100UL
29	#define F_SHRINK 0x00000200UL	29	#define F_SHRINK 0x00000200UL
30	#define F_ALLOW_BLESSED 0x00000400UL	30	#define F_ALLOW_BLESSED 0x00000400UL
31	#define F_CONV_BLESSED 0x00000800UL	31	#define F_CONV_BLESSED 0x00000800UL
		32	#define F_RELAXED 0x00001000UL
		33
32	#define F_MAXDEPTH 0xf8000000UL	34	#define F_MAXDEPTH 0xf8000000UL
33	#define S_MAXDEPTH 27	35	#define S_MAXDEPTH 27
34	#define F_MAXSIZE 0x01f00000UL	36	#define F_MAXSIZE 0x01f00000UL
35	#define S_MAXSIZE 20	37	#define S_MAXSIZE 20
36	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing	38	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
…		…
48		50
49	#define SB do {	51	#define SB do {
50	#define SE } while (0)	52	#define SE } while (0)
51		53
52	#if __GNUC__ >= 3	54	#if __GNUC__ >= 3
53	# define expect(expr,value) __builtin_expect ((expr),(value))	55	# define expect(expr,value) __builtin_expect ((expr), (value))
54	# define inline inline	56	# define INLINE static inline
55	#else	57	#else
56	# define expect(expr,value) (expr)	58	# define expect(expr,value) (expr)
57	# define inline static	59	# define INLINE static
58	#endif	60	#endif
59		61
60	#define expect_false(expr) expect ((expr) != 0, 0)	62	#define expect_false(expr) expect ((expr) != 0, 0)
61	#define expect_true(expr) expect ((expr) != 0, 1)	63	#define expect_true(expr) expect ((expr) != 0, 1)
62		64
		65	#define IN_RANGE_INC(type,val,beg,end) \
		66	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		67	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		68
63	#ifdef USE_ITHREADS	69	#ifdef USE_ITHREADS
64	# define JSON_SLOW 1	70	# define JSON_SLOW 1
		71	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
65	#else	72	#else
66	# define JSON_SLOW 1	73	# define JSON_SLOW 0
		74	# define JSON_STASH json_stash
67	#endif	75	#endif
68		76
69	static HV json_stash, json_boolean_stash; // JSON::XS::	77	static HV json_stash, json_boolean_stash; // JSON::XS::
70	static SV json_true, json_false;	78	static SV json_true, json_false;
71		79
…		…
76	} JSON;	84	} JSON;
77		85
78	/////////////////////////////////////////////////////////////////////////////	86	/////////////////////////////////////////////////////////////////////////////
79	// utility functions	87	// utility functions
80		88
81	inline void	89	INLINE void
82	shrink (SV *sv)	90	shrink (SV *sv)
83	{	91	{
84	sv_utf8_downgrade (sv, 1);	92	sv_utf8_downgrade (sv, 1);
85	if (SvLEN (sv) > SvCUR (sv) + 1)	93	if (SvLEN (sv) > SvCUR (sv) + 1)
86	{	94	{
…		…
95	// decode an utf-8 character and return it, or (UV)-1 in	103	// decode an utf-8 character and return it, or (UV)-1 in
96	// case of an error.	104	// case of an error.
97	// we special-case "safe" characters from U+80 .. U+7FF,	105	// we special-case "safe" characters from U+80 .. U+7FF,
98	// but use the very good perl function to parse anything else.	106	// but use the very good perl function to parse anything else.
99	// note that we never call this function for a ascii codepoints	107	// note that we never call this function for a ascii codepoints
100	inline UV	108	INLINE UV
101	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	109	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
102	{	110	{
103	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))	111	if (expect_true (len >= 2
104	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	112	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
105	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	113	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
106	{	114	{
107	*clen = 2;	115	*clen = 2;
108	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);	116	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
109	}	117	}
110	else	118	else
111	{	119	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
112	*clen = (STRLEN)-1;	120	}
113	return (UV)-1;	121
114	}	122	// likewise for encoding, also never called for ascii codepoints
		123	// this function takes advantage of this fact, although current gccs
		124	// seem to optimise the check for >= 0x80 away anyways
		125	INLINE unsigned char *
		126	encode_utf8 (unsigned char *s, UV ch)
		127	{
		128	if (expect_false (ch < 0x000080))
		129	*s++ = ch;
		130	else if (expect_true (ch < 0x000800))
		131	*s++ = 0xc0 \| ( ch >> 6),
		132	*s++ = 0x80 \| ( ch & 0x3f);
		133	else if ( ch < 0x010000)
		134	*s++ = 0xe0 \| ( ch >> 12),
		135	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		136	*s++ = 0x80 \| ( ch & 0x3f);
		137	else if ( ch < 0x110000)
		138	*s++ = 0xf0 \| ( ch >> 18),
		139	*s++ = 0x80 \| ((ch >> 12) & 0x3f),
		140	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		141	*s++ = 0x80 \| ( ch & 0x3f);
		142
		143	return s;
115	}	144	}
116		145
117	/////////////////////////////////////////////////////////////////////////////	146	/////////////////////////////////////////////////////////////////////////////
118	// encoder	147	// encoder
119		148
…		…
124	char *end; // SvEND (sv)	153	char *end; // SvEND (sv)
125	SV *sv; // result scalar	154	SV *sv; // result scalar
126	JSON json;	155	JSON json;
127	U32 indent; // indentation level	156	U32 indent; // indentation level
128	U32 maxdepth; // max. indentation/recursion level	157	U32 maxdepth; // max. indentation/recursion level
		158	UV limit; // escape character values >= this value when encoding
129	} enc_t;	159	} enc_t;
130		160
131	inline void	161	INLINE void
132	need (enc_t *enc, STRLEN len)	162	need (enc_t *enc, STRLEN len)
133	{	163	{
134	if (expect_false (enc->cur + len >= enc->end))	164	if (expect_false (enc->cur + len >= enc->end))
135	{	165	{
136	STRLEN cur = enc->cur - SvPVX (enc->sv);	166	STRLEN cur = enc->cur - SvPVX (enc->sv);
…		…
138	enc->cur = SvPVX (enc->sv) + cur;	168	enc->cur = SvPVX (enc->sv) + cur;
139	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	169	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
140	}	170	}
141	}	171	}
142		172
143	inline void	173	INLINE void
144	encode_ch (enc_t *enc, char ch)	174	encode_ch (enc_t *enc, char ch)
145	{	175	{
146	need (enc, 1);	176	need (enc, 1);
147	*enc->cur++ = ch;	177	*enc->cur++ = ch;
148	}	178	}
…		…
202	{	232	{
203	uch = ch;	233	uch = ch;
204	clen = 1;	234	clen = 1;
205	}	235	}
206		236
207	if (uch > 0x10FFFFUL)	237	if (uch < 0x80/0x20/ \|\| uch >= enc->limit)
208	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
209
210	if (uch < 0x80 \|\| enc->json.flags & F_ASCII \|\| (enc->json.flags & F_LATIN1 && uch > 0xFF))
211	{	238	{
212	if (uch > 0xFFFFUL)	239	if (uch >= 0x10000UL)
213	{	240	{
		241	if (uch >= 0x110000UL)
		242	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		243
214	need (enc, len += 11);	244	need (enc, len += 11);
215	sprintf (enc->cur, "\\u%04x\\u%04x",	245	sprintf (enc->cur, "\\u%04x\\u%04x",
216	(int)((uch - 0x10000) / 0x400 + 0xD800),	246	(int)((uch - 0x10000) / 0x400 + 0xD800),
217	(int)((uch - 0x10000) % 0x400 + 0xDC00));	247	(int)((uch - 0x10000) % 0x400 + 0xDC00));
218	enc->cur += 12;	248	enc->cur += 12;
…		…
246	while (--clen);	276	while (--clen);
247	}	277	}
248	else	278	else
249	{	279	{
250	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	280	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
251	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	281	enc->cur = encode_utf8 (enc->cur, uch);
252	++str;	282	++str;
253	}	283	}
254	}	284	}
255	}	285	}
256	}	286	}
257		287
258	--len;	288	--len;
259	}	289	}
260	}	290	}
261		291
262	inline void	292	INLINE void
263	encode_indent (enc_t *enc)	293	encode_indent (enc_t *enc)
264	{	294	{
265	if (enc->json.flags & F_INDENT)	295	if (enc->json.flags & F_INDENT)
266	{	296	{
267	int spaces = enc->indent * INDENT_STEP;	297	int spaces = enc->indent * INDENT_STEP;
…		…
270	memset (enc->cur, ' ', spaces);	300	memset (enc->cur, ' ', spaces);
271	enc->cur += spaces;	301	enc->cur += spaces;
272	}	302	}
273	}	303	}
274		304
275	inline void	305	INLINE void
276	encode_space (enc_t *enc)	306	encode_space (enc_t *enc)
277	{	307	{
278	need (enc, 1);	308	need (enc, 1);
279	encode_ch (enc, ' ');	309	encode_ch (enc, ' ');
280	}	310	}
281		311
282	inline void	312	INLINE void
283	encode_nl (enc_t *enc)	313	encode_nl (enc_t *enc)
284	{	314	{
285	if (enc->json.flags & F_INDENT)	315	if (enc->json.flags & F_INDENT)
286	{	316	{
287	need (enc, 1);	317	need (enc, 1);
288	encode_ch (enc, '\n');	318	encode_ch (enc, '\n');
289	}	319	}
290	}	320	}
291		321
292	inline void	322	INLINE void
293	encode_comma (enc_t *enc)	323	encode_comma (enc_t *enc)
294	{	324	{
295	encode_ch (enc, ',');	325	encode_ch (enc, ',');
296		326
297	if (enc->json.flags & F_INDENT)	327	if (enc->json.flags & F_INDENT)
…		…
308	int i, len = av_len (av);	338	int i, len = av_len (av);
309		339
310	if (enc->indent >= enc->maxdepth)	340	if (enc->indent >= enc->maxdepth)
311	croak ("data structure too deep (hit recursion limit)");	341	croak ("data structure too deep (hit recursion limit)");
312		342
313	encode_ch (enc, '['); encode_nl (enc);	343	encode_ch (enc, '[');
314	++enc->indent;	344
		345	if (len >= 0)
		346	{
		347	encode_nl (enc); ++enc->indent;
315		348
316	for (i = 0; i <= len; ++i)	349	for (i = 0; i <= len; ++i)
317	{	350	{
318	SV **svp = av_fetch (av, i, 0);	351	SV **svp = av_fetch (av, i, 0);
319		352
320	encode_indent (enc);	353	encode_indent (enc);
321		354
322	if (svp)	355	if (svp)
323	encode_sv (enc, *svp);	356	encode_sv (enc, *svp);
324	else	357	else
325	encode_str (enc, "null", 4, 0);	358	encode_str (enc, "null", 4, 0);
326		359
327	if (i < len)	360	if (i < len)
328	encode_comma (enc);	361	encode_comma (enc);
329	}	362	}
330		363
		364	encode_nl (enc); --enc->indent; encode_indent (enc);
		365	}
		366
331	encode_nl (enc);	367	encode_ch (enc, ']');
332
333	--enc->indent;
334	encode_indent (enc); encode_ch (enc, ']');
335	}	368	}
336		369
337	static void	370	static void
338	encode_he (enc_t enc, HE he)	371	encode_hk (enc_t enc, HE he)
339	{	372	{
340	encode_ch (enc, '"');	373	encode_ch (enc, '"');
341		374
342	if (HeKLEN (he) == HEf_SVKEY)	375	if (HeKLEN (he) == HEf_SVKEY)
343	{	376	{
…		…
356	encode_ch (enc, '"');	389	encode_ch (enc, '"');
357		390
358	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);	391	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
359	encode_ch (enc, ':');	392	encode_ch (enc, ':');
360	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);	393	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
361	encode_sv (enc, HeVAL (he));
362	}	394	}
363		395
364	// compare hash entries, used when all keys are bytestrings	396	// compare hash entries, used when all keys are bytestrings
365	static int	397	static int
366	he_cmp_fast (const void a_, const void b_)	398	he_cmp_fast (const void a_, const void b_)
…		…
371	HE b = (HE **)b_;	403	HE b = (HE **)b_;
372		404
373	STRLEN la = HeKLEN (a);	405	STRLEN la = HeKLEN (a);
374	STRLEN lb = HeKLEN (b);	406	STRLEN lb = HeKLEN (b);
375		407
376	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	408	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
377	cmp = la - lb;	409	cmp = lb - la;
378		410
379	return cmp;	411	return cmp;
380	}	412	}
381		413
382	// compare hash entries, used when some keys are sv's or utf-x	414	// compare hash entries, used when some keys are sv's or utf-x
383	static int	415	static int
384	he_cmp_slow (const void a, const void b)	416	he_cmp_slow (const void a, const void b)
385	{	417	{
386	return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));	418	return sv_cmp (HeSVKEY_force ((HE )b), HeSVKEY_force ((HE **)a));
387	}	419	}
388		420
389	static void	421	static void
390	encode_hv (enc_t enc, HV hv)	422	encode_hv (enc_t enc, HV hv)
391	{	423	{
		424	HE *he;
392	int count, i;	425	int count;
393		426
394	if (enc->indent >= enc->maxdepth)	427	if (enc->indent >= enc->maxdepth)
395	croak ("data structure too deep (hit recursion limit)");	428	croak ("data structure too deep (hit recursion limit)");
396		429
397	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	430	encode_ch (enc, '{');
398		431
399	if ((count = hv_iterinit (hv)))
400	{
401	// for canonical output we have to sort by keys first	432	// for canonical output we have to sort by keys first
402	// actually, this is mostly due to the stupid so-called	433	// actually, this is mostly due to the stupid so-called
403	// security workaround added somewhere in 5.8.x.	434	// security workaround added somewhere in 5.8.x.
404	// that randomises hash orderings	435	// that randomises hash orderings
405	if (enc->json.flags & F_CANONICAL)	436	if (enc->json.flags & F_CANONICAL)
		437	{
		438	int count = hv_iterinit (hv);
		439
		440	if (SvMAGICAL (hv))
406	{	441	{
		442	// need to count by iterating. could improve by dynamically building the vector below
		443	// but I don't care for the speed of this special case.
		444	// note also that we will run into undefined behaviour when the two iterations
		445	// do not result in the same count, something I might care for in some later release.
		446
		447	count = 0;
		448	while (hv_iternext (hv))
		449	++count;
		450
		451	hv_iterinit (hv);
		452	}
		453
		454	if (count)
		455	{
407	int fast = 1;	456	int i, fast = 1;
408	HE *he;
409	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	457	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
410	HE *hes = _alloca (count sizeof (HE));	458	HE *hes = _alloca (count sizeof (HE));
411	#else	459	#else
412	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode	460	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
413	#endif	461	#endif
…		…
440		488
441	FREETMPS;	489	FREETMPS;
442	LEAVE;	490	LEAVE;
443	}	491	}
444		492
445	for (i = 0; i < count; ++i)	493	encode_nl (enc); ++enc->indent;
		494
		495	while (count--)
446	{	496	{
447	encode_indent (enc);	497	encode_indent (enc);
		498	he = hes [count];
448	encode_he (enc, hes [i]);	499	encode_hk (enc, he);
		500	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
449		501
450	if (i < count - 1)	502	if (count)
451	encode_comma (enc);	503	encode_comma (enc);
452	}	504	}
453		505
454	encode_nl (enc);	506	encode_nl (enc); --enc->indent; encode_indent (enc);
455	}	507	}
		508	}
456	else	509	else
		510	{
		511	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))
		512	if ((he = hv_iternext (hv)))
457	{	513	{
458	HE *he = hv_iternext (hv);	514	encode_nl (enc); ++enc->indent;
459		515
460	for (;;)	516	for (;;)
461	{	517	{
462	encode_indent (enc);	518	encode_indent (enc);
463	encode_he (enc, he);	519	encode_hk (enc, he);
		520	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
464		521
465	if (!(he = hv_iternext (hv)))	522	if (!(he = hv_iternext (hv)))
466	break;	523	break;
467		524
468	encode_comma (enc);	525	encode_comma (enc);
469	}	526	}
470		527
471	encode_nl (enc);	528	encode_nl (enc); --enc->indent; encode_indent (enc);
472	}	529	}
473	}	530	}
474		531
475	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	532	encode_ch (enc, '}');
476	}	533	}
477		534
478	// encode objects, arrays and special \0=false and \1=true values.	535	// encode objects, arrays and special \0=false and \1=true values.
479	static void	536	static void
480	encode_rv (enc_t enc, SV sv)	537	encode_rv (enc_t enc, SV sv)
…		…
510	// we re-bless the reference to get overload and other niceties right	567	// we re-bless the reference to get overload and other niceties right
511	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);	568	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
512		569
513	if (to_json)	570	if (to_json)
514	{	571	{
515	int count;
516	dSP;	572	dSP;
517		573
518	ENTER; SAVETMPS; PUSHMARK (SP);	574	ENTER; SAVETMPS; PUSHMARK (SP);
519	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));	575	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
520		576
…		…
589	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	645	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
590	enc->cur += strlen (enc->cur);	646	enc->cur += strlen (enc->cur);
591	}	647	}
592	else if (SvIOKp (sv))	648	else if (SvIOKp (sv))
593	{	649	{
594	// we assume we can always read an IV as a UV	650	// we assume we can always read an IV as a UV and vice versa
595	if (SvUV (sv) & ~(UV)0x7fff)	651	// we assume two's complement
		652	// we assume no aliasing issues in the union
		653	if (SvIsUV (sv) ? SvUVX (sv) > 59000
		654	: SvIVX (sv) > 59000 \|\| SvIVX (sv) < -59000)
596	{	655	{
597	// large integer, use the (rather slow) snprintf way.	656	// large integer, use the (rather slow) snprintf way.
598	need (enc, sizeof (UV) * 3);	657	need (enc, sizeof (UV) * 5 / 2 + 1); // CHAR_BIT is at least 8
599	enc->cur +=	658	enc->cur +=
600	SvIsUV(sv)	659	SvIsUV(sv)
601	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))	660	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
602	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));	661	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
603	}	662	}
604	else	663	else
605	{	664	{
606	// optimise the "small number case"	665	// optimise the "small number case"
607	// code will likely be branchless and use only a single multiplication	666	// code will likely be branchless and use only a single multiplication
		667	// works for numbers up to 59074
608	I32 i = SvIV (sv);	668	I32 i = SvIVX (sv);
609	U32 u;	669	U32 u;
610	char digit, nz = 0;	670	char digit, nz = 0;
611		671
612	need (enc, 6);	672	need (enc, 6);
613		673
…		…
650	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	710	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
651	enc.cur = SvPVX (enc.sv);	711	enc.cur = SvPVX (enc.sv);
652	enc.end = SvEND (enc.sv);	712	enc.end = SvEND (enc.sv);
653	enc.indent = 0;	713	enc.indent = 0;
654	enc.maxdepth = DEC_DEPTH (enc.json.flags);	714	enc.maxdepth = DEC_DEPTH (enc.json.flags);
		715	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		716	: enc.json.flags & F_LATIN1 ? 0x000100UL
		717	: 0x110000UL;
655		718
656	SvPOK_only (enc.sv);	719	SvPOK_only (enc.sv);
657	encode_sv (&enc, scalar);	720	encode_sv (&enc, scalar);
658		721
659	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	722	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
…		…
680	JSON json;	743	JSON json;
681	U32 depth; // recursion depth	744	U32 depth; // recursion depth
682	U32 maxdepth; // recursion depth limit	745	U32 maxdepth; // recursion depth limit
683	} dec_t;	746	} dec_t;
684		747
685	inline void	748	INLINE void
		749	decode_comment (dec_t *dec)
		750	{
		751	// only '#'-style comments allowed a.t.m.
		752
		753	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)
		754	++dec->cur;
		755	}
		756
		757	INLINE void
686	decode_ws (dec_t *dec)	758	decode_ws (dec_t *dec)
687	{	759	{
688	for (;;)	760	for (;;)
689	{	761	{
690	char ch = *dec->cur;	762	char ch = *dec->cur;
691		763
692	if (ch > 0x20	764	if (ch > 0x20)
		765	{
		766	if (expect_false (ch == '#'))
		767	{
		768	if (dec->json.flags & F_RELAXED)
		769	decode_comment (dec);
		770	else
		771	break;
		772	}
		773	else
		774	break;
		775	}
693	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	776	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
694	break;	777	break; // parse error, but let higher level handle it, gives better error messages
695		778
696	++dec->cur;	779	++dec->cur;
697	}	780	}
698	}	781	}
699		782
…		…
805		888
806	if (hi >= 0x80)	889	if (hi >= 0x80)
807	{	890	{
808	utf8 = 1;	891	utf8 = 1;
809		892
810	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	893	cur = encode_utf8 (cur, hi);
811	}	894	}
812	else	895	else
813	*cur++ = hi;	896	*cur++ = hi;
814	}	897	}
815	break;	898	break;
…		…
817	default:	900	default:
818	--dec_cur;	901	--dec_cur;
819	ERR ("illegal backslash escape sequence in string");	902	ERR ("illegal backslash escape sequence in string");
820	}	903	}
821	}	904	}
822	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	905	else if (expect_true (ch >= 0x20 && ch < 0x80))
823	*cur++ = ch;	906	*cur++ = ch;
824	else if (ch >= 0x80)	907	else if (ch >= 0x80)
825	{	908	{
826	STRLEN clen;	909	STRLEN clen;
827	UV uch;	910	UV uch;
…		…
950		1033
951	if (!is_nv)	1034	if (!is_nv)
952	{	1035	{
953	int len = dec->cur - start;	1036	int len = dec->cur - start;
954		1037
955	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1038	// special case the rather common 1..5-digit-int case
956	if (*start == '-')	1039	if (*start == '-')
957	switch (len)	1040	switch (len)
958	{	1041	{
959	case 2: return newSViv (-( start [1] - '0' * 1));	1042	case 2: return newSViv (-( start [1] - '0' * 1));
960	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1043	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
961	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));	1044	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
962	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));	1045	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1046	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
963	}	1047	}
964	else	1048	else
965	switch (len)	1049	switch (len)
966	{	1050	{
967	case 1: return newSViv ( start [0] - '0' * 1);	1051	case 1: return newSViv ( start [0] - '0' * 1);
968	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1052	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
969	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);	1053	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
970	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);	1054	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1055	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
971	}	1056	}
972		1057
973	{	1058	{
974	UV uv;	1059	UV uv;
975	int numtype = grok_number (start, len, &uv);	1060	int numtype = grok_number (start, len, &uv);
…		…
1036		1121
1037	if (*dec->cur != ',')	1122	if (*dec->cur != ',')
1038	ERR (", or ] expected while parsing array");	1123	ERR (", or ] expected while parsing array");
1039		1124
1040	++dec->cur;	1125	++dec->cur;
		1126
		1127	decode_ws (dec);
		1128
		1129	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1130	{
		1131	++dec->cur;
		1132	break;
		1133	}
1041	}	1134	}
1042		1135
1043	DEC_DEC_DEPTH;	1136	DEC_DEC_DEPTH;
1044	return newRV_noinc ((SV *)av);	1137	return newRV_noinc ((SV *)av);
1045		1138
…		…
1061	if (*dec->cur == '}')	1154	if (*dec->cur == '}')
1062	++dec->cur;	1155	++dec->cur;
1063	else	1156	else
1064	for (;;)	1157	for (;;)
1065	{	1158	{
1066	decode_ws (dec); EXPECT_CH ('"');	1159	EXPECT_CH ('"');
1067		1160
1068	// heuristic: assume that	1161	// heuristic: assume that
1069	// a) decode_str + hv_store_ent are abysmally slow.	1162	// a) decode_str + hv_store_ent are abysmally slow.
1070	// b) most hash keys are short, simple ascii text.	1163	// b) most hash keys are short, simple ascii text.
1071	// => try to "fast-match" such strings to avoid	1164	// => try to "fast-match" such strings to avoid
…		…
1085	if (!key)	1178	if (!key)
1086	goto fail;	1179	goto fail;
1087		1180
1088	decode_ws (dec); EXPECT_CH (':');	1181	decode_ws (dec); EXPECT_CH (':');
1089		1182
		1183	decode_ws (dec);
1090	value = decode_sv (dec);	1184	value = decode_sv (dec);
1091	if (!value)	1185	if (!value)
1092	{	1186	{
1093	SvREFCNT_dec (key);	1187	SvREFCNT_dec (key);
1094	goto fail;	1188	goto fail;
…		…
1106	int len = p - key;	1200	int len = p - key;
1107	dec->cur = p + 1;	1201	dec->cur = p + 1;
1108		1202
1109	decode_ws (dec); EXPECT_CH (':');	1203	decode_ws (dec); EXPECT_CH (':');
1110		1204
		1205	decode_ws (dec);
1111	value = decode_sv (dec);	1206	value = decode_sv (dec);
1112	if (!value)	1207	if (!value)
1113	goto fail;	1208	goto fail;
1114		1209
1115	hv_store (hv, key, len, value, 0);	1210	hv_store (hv, key, len, value, 0);
…		…
1131		1226
1132	if (*dec->cur != ',')	1227	if (*dec->cur != ',')
1133	ERR (", or } expected while parsing object/hash");	1228	ERR (", or } expected while parsing object/hash");
1134		1229
1135	++dec->cur;	1230	++dec->cur;
		1231
		1232	decode_ws (dec);
		1233
		1234	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1235	{
		1236	++dec->cur;
		1237	break;
		1238	}
1136	}	1239	}
1137		1240
1138	DEC_DEC_DEPTH;	1241	DEC_DEC_DEPTH;
1139	sv = newRV_noinc ((SV *)hv);	1242	sv = newRV_noinc ((SV *)hv);
1140		1243
…		…
1205	}	1308	}
1206		1309
1207	static SV *	1310	static SV *
1208	decode_sv (dec_t *dec)	1311	decode_sv (dec_t *dec)
1209	{	1312	{
1210	decode_ws (dec);
1211
1212	// the beauty of JSON: you need exactly one character lookahead	1313	// the beauty of JSON: you need exactly one character lookahead
1213	// to parse anything.	1314	// to parse everything.
1214	switch (*dec->cur)	1315	switch (*dec->cur)
1215	{	1316	{
1216	case '"': ++dec->cur; return decode_str (dec);	1317	case '"': ++dec->cur; return decode_str (dec);
1217	case '[': ++dec->cur; return decode_av (dec);	1318	case '[': ++dec->cur; return decode_av (dec);
1218	case '{': ++dec->cur; return decode_hv (dec);	1319	case '{': ++dec->cur; return decode_hv (dec);
1219		1320
1220	case '-':	1321	case '-':
1221	case '0': case '1': case '2': case '3': case '4':	1322	case '0': case '1': case '2': case '3': case '4':
1222	case '5': case '6': case '7': case '8': case '9':	1323	case '5': case '6': case '7': case '8': case '9':
1223	return decode_num (dec);	1324	return decode_num (dec);
1224		1325
1225	case 't':	1326	case 't':
1226	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1327	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1227	{	1328	{
1228	dec->cur += 4;	1329	dec->cur += 4;
		1330	#if JSON_SLOW
		1331	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);
		1332	#endif
1229	return SvREFCNT_inc (json_true);	1333	return SvREFCNT_inc (json_true);
1230	}	1334	}
1231	else	1335	else
1232	ERR ("'true' expected");	1336	ERR ("'true' expected");
1233		1337
…		…
1235		1339
1236	case 'f':	1340	case 'f':
1237	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1341	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1238	{	1342	{
1239	dec->cur += 5;	1343	dec->cur += 5;
		1344	#if JSON_SLOW
		1345	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1346	#endif
1240	return SvREFCNT_inc (json_false);	1347	return SvREFCNT_inc (json_false);
1241	}	1348	}
1242	else	1349	else
1243	ERR ("'false' expected");	1350	ERR ("'false' expected");
1244		1351
…		…
1294		1401
1295	if (dec.json.cb_object \|\| dec.json.cb_sk_object)	1402	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
1296	dec.json.flags \|= F_HOOK;	1403	dec.json.flags \|= F_HOOK;
1297		1404
1298	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1405	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1406
		1407	decode_ws (&dec);
1299	sv = decode_sv (&dec);	1408	sv = decode_sv (&dec);
1300		1409
1301	if (!(offset_return \|\| !sv))	1410	if (!(offset_return \|\| !sv))
1302	{	1411	{
1303	// check for trailing garbage	1412	// check for trailing garbage
…		…
1379	json_boolean_stash = 0;	1488	json_boolean_stash = 0;
1380		1489
1381	void new (char *klass)	1490	void new (char *klass)
1382	PPCODE:	1491	PPCODE:
1383	{	1492	{
1384	HV *stash = !JSON_SLOW \|\| json_stash
1385	? json_stash
1386	: gv_stashpv ("JSON::XS", 1);
1387	SV *pv = NEWSV (0, sizeof (JSON));	1493	SV *pv = NEWSV (0, sizeof (JSON));
1388	SvPOK_only (pv);	1494	SvPOK_only (pv);
1389	Zero (SvPVX (pv), 1, JSON);	1495	Zero (SvPVX (pv), 1, JSON);
1390	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	1496	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1391	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), stash)));	1497	XPUSHs (sv_2mortal (sv_bless (
		1498	newRV_noinc (pv),
		1499	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1500	)));
1392	}	1501	}
1393		1502
1394	void ascii (JSON *self, int enable = 1)	1503	void ascii (JSON *self, int enable = 1)
1395	ALIAS:	1504	ALIAS:
1396	ascii = F_ASCII	1505	ascii = F_ASCII
…		…
1403	pretty = F_PRETTY	1512	pretty = F_PRETTY
1404	allow_nonref = F_ALLOW_NONREF	1513	allow_nonref = F_ALLOW_NONREF
1405	shrink = F_SHRINK	1514	shrink = F_SHRINK
1406	allow_blessed = F_ALLOW_BLESSED	1515	allow_blessed = F_ALLOW_BLESSED
1407	convert_blessed = F_CONV_BLESSED	1516	convert_blessed = F_CONV_BLESSED
		1517	relaxed = F_RELAXED
1408	PPCODE:	1518	PPCODE:
1409	{	1519	{
1410	if (enable)	1520	if (enable)
1411	self->flags \|= ix;	1521	self->flags \|= ix;
1412	else	1522	else
1413	self->flags &= ~ix;	1523	self->flags &= ~ix;
1414		1524
1415	XPUSHs (ST (0));	1525	XPUSHs (ST (0));
1416	}	1526	}
1417		1527
		1528	void get_ascii (JSON *self)
		1529	ALIAS:
		1530	get_ascii = F_ASCII
		1531	get_latin1 = F_LATIN1
		1532	get_utf8 = F_UTF8
		1533	get_indent = F_INDENT
		1534	get_canonical = F_CANONICAL
		1535	get_space_before = F_SPACE_BEFORE
		1536	get_space_after = F_SPACE_AFTER
		1537	get_allow_nonref = F_ALLOW_NONREF
		1538	get_shrink = F_SHRINK
		1539	get_allow_blessed = F_ALLOW_BLESSED
		1540	get_convert_blessed = F_CONV_BLESSED
		1541	get_relaxed = F_RELAXED
		1542	PPCODE:
		1543	XPUSHs (boolSV (self->flags & ix));
		1544
1418	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1545	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
1419	PPCODE:	1546	PPCODE:
1420	{	1547	{
1421	UV log2 = 0;	1548	UV log2 = 0;
1422		1549
…		…
1428	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);	1555	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1429		1556
1430	XPUSHs (ST (0));	1557	XPUSHs (ST (0));
1431	}	1558	}
1432		1559
		1560	U32 get_max_depth (JSON *self)
		1561	CODE:
		1562	RETVAL = DEC_DEPTH (self->flags);
		1563	OUTPUT:
		1564	RETVAL
		1565
1433	void max_size (JSON *self, UV max_size = 0)	1566	void max_size (JSON *self, UV max_size = 0)
1434	PPCODE:	1567	PPCODE:
1435	{	1568	{
1436	UV log2 = 0;	1569	UV log2 = 0;
1437		1570
…		…
1443		1576
1444	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);	1577	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
1445		1578
1446	XPUSHs (ST (0));	1579	XPUSHs (ST (0));
1447	}	1580	}
		1581
		1582	int get_max_size (JSON *self)
		1583	CODE:
		1584	RETVAL = DEC_SIZE (self->flags);
		1585	OUTPUT:
		1586	RETVAL
1448		1587
1449	void filter_json_object (JSON self, SV cb = &PL_sv_undef)	1588	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1450	PPCODE:	1589	PPCODE:
1451	{	1590	{
1452	SvREFCNT_dec (self->cb_object);	1591	SvREFCNT_dec (self->cb_object);
…		…
1499	SvREFCNT_dec (self->cb_sk_object);	1638	SvREFCNT_dec (self->cb_sk_object);
1500	SvREFCNT_dec (self->cb_object);	1639	SvREFCNT_dec (self->cb_object);
1501		1640
1502	PROTOTYPES: ENABLE	1641	PROTOTYPES: ENABLE
1503		1642
1504	void to_json (SV *scalar)	1643	void encode_json (SV *scalar)
1505	PPCODE:	1644	PPCODE:
1506	{	1645	{
1507	JSON json = { F_DEFAULT \| F_UTF8 };	1646	JSON json = { F_DEFAULT \| F_UTF8 };
1508	XPUSHs (encode_json (scalar, &json));	1647	XPUSHs (encode_json (scalar, &json));
1509	}	1648	}
1510		1649
1511	void from_json (SV *jsonstr)	1650	void decode_json (SV *jsonstr)
1512	PPCODE:	1651	PPCODE:
1513	{	1652	{
1514	JSON json = { F_DEFAULT \| F_UTF8 };	1653	JSON json = { F_DEFAULT \| F_UTF8 };
1515	XPUSHs (decode_json (jsonstr, &json, 0));	1654	XPUSHs (decode_json (jsonstr, &json, 0));
1516	}	1655	}

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.57 by root, Mon Aug 13 16:05:42 2007 UTC vs. Revision 1.74 by root, Wed Mar 19 15:17:53 2008 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.57 by root, Mon Aug 13 16:05:42 2007 UTC vs.
Revision 1.74 by root, Wed Mar 19 15:17:53 2008 UTC