ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/JSON-XS/XS.xs
(Generate patch)

Comparing JSON-XS/XS.xs (file contents):
Revision 1.77 by root, Tue Mar 25 06:37:38 2008 UTC vs.
Revision 1.103 by root, Wed Jan 6 08:01:39 2010 UTC

12#if defined(__BORLANDC__) || defined(_MSC_VER) 12#if defined(__BORLANDC__) || defined(_MSC_VER)
13# define snprintf _snprintf // C compilers have this in stdio.h 13# define snprintf _snprintf // C compilers have this in stdio.h
14#endif 14#endif
15 15
16// some old perls do not have this, try to make it work, no 16// some old perls do not have this, try to make it work, no
17// guarentees, though. if it breaks, you get to keep the pieces. 17// guarantees, though. if it breaks, you get to keep the pieces.
18#ifndef UTF8_MAXBYTES 18#ifndef UTF8_MAXBYTES
19# define UTF8_MAXBYTES 13 19# define UTF8_MAXBYTES 13
20#endif 20#endif
21 21
22// three extra for rounding, sign, and end of string
22#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2) 23#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
23 24
24#define F_ASCII 0x00000001UL 25#define F_ASCII 0x00000001UL
25#define F_LATIN1 0x00000002UL 26#define F_LATIN1 0x00000002UL
26#define F_UTF8 0x00000004UL 27#define F_UTF8 0x00000004UL
27#define F_INDENT 0x00000008UL 28#define F_INDENT 0x00000008UL
31#define F_ALLOW_NONREF 0x00000100UL 32#define F_ALLOW_NONREF 0x00000100UL
32#define F_SHRINK 0x00000200UL 33#define F_SHRINK 0x00000200UL
33#define F_ALLOW_BLESSED 0x00000400UL 34#define F_ALLOW_BLESSED 0x00000400UL
34#define F_CONV_BLESSED 0x00000800UL 35#define F_CONV_BLESSED 0x00000800UL
35#define F_RELAXED 0x00001000UL 36#define F_RELAXED 0x00001000UL
36 37#define F_ALLOW_UNKNOWN 0x00002000UL
37#define F_MAXDEPTH 0xf8000000UL
38#define S_MAXDEPTH 27
39#define F_MAXSIZE 0x01f00000UL
40#define S_MAXSIZE 20
41#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing 38#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
42 39
43#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
44#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
45
46#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 40#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
47#define F_DEFAULT (9UL << S_MAXDEPTH)
48 41
49#define INIT_SIZE 32 // initial scalar size to be allocated 42#define INIT_SIZE 32 // initial scalar size to be allocated
50#define INDENT_STEP 3 // spaces per indentation level 43#define INDENT_STEP 3 // spaces per indentation level
51 44
52#define SHORT_STRING_LEN 16384 // special-case strings of up to this size 45#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
67 60
68#define IN_RANGE_INC(type,val,beg,end) \ 61#define IN_RANGE_INC(type,val,beg,end) \
69 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ 62 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
70 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) 63 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
71 64
65#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
66
72#ifdef USE_ITHREADS 67#ifdef USE_ITHREADS
73# define JSON_SLOW 1 68# define JSON_SLOW 1
74# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) 69# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
75#else 70#else
76# define JSON_SLOW 0 71# define JSON_SLOW 0
82 77
83enum { 78enum {
84 INCR_M_WS = 0, // initial whitespace skipping, must be 0 79 INCR_M_WS = 0, // initial whitespace skipping, must be 0
85 INCR_M_STR, // inside string 80 INCR_M_STR, // inside string
86 INCR_M_BS, // inside backslash 81 INCR_M_BS, // inside backslash
82 INCR_M_C0, // inside comment in initial whitespace sequence
83 INCR_M_C1, // inside comment in other places
87 INCR_M_JSON // outside anything, count nesting 84 INCR_M_JSON // outside anything, count nesting
88}; 85};
89 86
90#define INCR_DONE(json) (!(json)->incr_nest && (json)->incr_mode == INCR_M_JSON) 87#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
91 88
92typedef struct { 89typedef struct {
93 U32 flags; 90 U32 flags;
91 U32 max_depth;
92 STRLEN max_size;
93
94 SV *cb_object; 94 SV *cb_object;
95 HV *cb_sk_object; 95 HV *cb_sk_object;
96 96
97 // for the incremental parser 97 // for the incremental parser
98 SV *incr_text; // the source text so far 98 SV *incr_text; // the source text so far
99 STRLEN incr_pos; // the current offset into the text 99 STRLEN incr_pos; // the current offset into the text
100 int incr_nest; // {[]}-nesting level 100 int incr_nest; // {[]}-nesting level
101 int incr_mode; 101 unsigned char incr_mode;
102} JSON; 102} JSON;
103
104INLINE void
105json_init (JSON *json)
106{
107 Zero (json, 1, JSON);
108 json->max_depth = 512;
109}
103 110
104///////////////////////////////////////////////////////////////////////////// 111/////////////////////////////////////////////////////////////////////////////
105// utility functions 112// utility functions
106 113
114INLINE SV *
115get_bool (const char *name)
116{
117 SV *sv = get_sv (name, 1);
118
119 SvREADONLY_on (sv);
120 SvREADONLY_on (SvRV (sv));
121
122 return sv;
123}
124
107INLINE void 125INLINE void
108shrink (SV *sv) 126shrink (SV *sv)
109{ 127{
110 sv_utf8_downgrade (sv, 1); 128 sv_utf8_downgrade (sv, 1);
129
111 if (SvLEN (sv) > SvCUR (sv) + 1) 130 if (SvLEN (sv) > SvCUR (sv) + 1)
112 { 131 {
113#ifdef SvPV_shrink_to_cur 132#ifdef SvPV_shrink_to_cur
114 SvPV_shrink_to_cur (sv); 133 SvPV_shrink_to_cur (sv);
115#elif defined (SvPV_renew) 134#elif defined (SvPV_renew)
159 *s++ = 0x80 | ( ch & 0x3f); 178 *s++ = 0x80 | ( ch & 0x3f);
160 179
161 return s; 180 return s;
162} 181}
163 182
183// convert offset pointer to character index, sv must be string
184static STRLEN
185ptr_to_index (SV *sv, char *offset)
186{
187 return SvUTF8 (sv)
188 ? utf8_distance (offset, SvPVX (sv))
189 : offset - SvPVX (sv);
190}
191
164///////////////////////////////////////////////////////////////////////////// 192/////////////////////////////////////////////////////////////////////////////
165// encoder 193// encoder
166 194
167// structure used for encoding JSON 195// structure used for encoding JSON
168typedef struct 196typedef struct
170 char *cur; // SvPVX (sv) + current output position 198 char *cur; // SvPVX (sv) + current output position
171 char *end; // SvEND (sv) 199 char *end; // SvEND (sv)
172 SV *sv; // result scalar 200 SV *sv; // result scalar
173 JSON json; 201 JSON json;
174 U32 indent; // indentation level 202 U32 indent; // indentation level
175 U32 maxdepth; // max. indentation/recursion level
176 UV limit; // escape character values >= this value when encoding 203 UV limit; // escape character values >= this value when encoding
177} enc_t; 204} enc_t;
178 205
179INLINE void 206INLINE void
180need (enc_t *enc, STRLEN len) 207need (enc_t *enc, STRLEN len)
181{ 208{
182 if (expect_false (enc->cur + len >= enc->end)) 209 if (expect_false (enc->cur + len >= enc->end))
183 { 210 {
184 STRLEN cur = enc->cur - SvPVX (enc->sv); 211 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
185 SvGROW (enc->sv, cur + len + 1); 212 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
186 enc->cur = SvPVX (enc->sv) + cur; 213 enc->cur = SvPVX (enc->sv) + cur;
187 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 214 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
188 } 215 }
189} 216}
190 217
265 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 292 (int)((uch - 0x10000) % 0x400 + 0xDC00));
266 enc->cur += 12; 293 enc->cur += 12;
267 } 294 }
268 else 295 else
269 { 296 {
270 static char hexdigit [16] = "0123456789abcdef";
271 need (enc, len += 5); 297 need (enc, len += 5);
272 *enc->cur++ = '\\'; 298 *enc->cur++ = '\\';
273 *enc->cur++ = 'u'; 299 *enc->cur++ = 'u';
274 *enc->cur++ = hexdigit [ uch >> 12 ]; 300 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
275 *enc->cur++ = hexdigit [(uch >> 8) & 15]; 301 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
276 *enc->cur++ = hexdigit [(uch >> 4) & 15]; 302 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
277 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 303 *enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
278 } 304 }
279 305
280 str += clen; 306 str += clen;
281 } 307 }
282 else if (enc->json.flags & F_LATIN1) 308 else if (enc->json.flags & F_LATIN1)
353static void 379static void
354encode_av (enc_t *enc, AV *av) 380encode_av (enc_t *enc, AV *av)
355{ 381{
356 int i, len = av_len (av); 382 int i, len = av_len (av);
357 383
358 if (enc->indent >= enc->maxdepth) 384 if (enc->indent >= enc->json.max_depth)
359 croak ("data structure too deep (hit recursion limit)"); 385 croak (ERR_NESTING_EXCEEDED);
360 386
361 encode_ch (enc, '['); 387 encode_ch (enc, '[');
362 388
363 if (len >= 0) 389 if (len >= 0)
364 { 390 {
438 464
439static void 465static void
440encode_hv (enc_t *enc, HV *hv) 466encode_hv (enc_t *enc, HV *hv)
441{ 467{
442 HE *he; 468 HE *he;
443 int count;
444 469
445 if (enc->indent >= enc->maxdepth) 470 if (enc->indent >= enc->json.max_depth)
446 croak ("data structure too deep (hit recursion limit)"); 471 croak (ERR_NESTING_EXCEEDED);
447 472
448 encode_ch (enc, '{'); 473 encode_ch (enc, '{');
449 474
450 // for canonical output we have to sort by keys first 475 // for canonical output we have to sort by keys first
451 // actually, this is mostly due to the stupid so-called 476 // actually, this is mostly due to the stupid so-called
452 // security workaround added somewhere in 5.8.x. 477 // security workaround added somewhere in 5.8.x
453 // that randomises hash orderings 478 // that randomises hash orderings
454 if (enc->json.flags & F_CANONICAL) 479 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
455 { 480 {
456 int count = hv_iterinit (hv); 481 int count = hv_iterinit (hv);
457 482
458 if (SvMAGICAL (hv)) 483 if (SvMAGICAL (hv))
459 { 484 {
632 657
633 if (len == 1 && *pv == '1') 658 if (len == 1 && *pv == '1')
634 encode_str (enc, "true", 4, 0); 659 encode_str (enc, "true", 4, 0);
635 else if (len == 1 && *pv == '0') 660 else if (len == 1 && *pv == '0')
636 encode_str (enc, "false", 5, 0); 661 encode_str (enc, "false", 5, 0);
662 else if (enc->json.flags & F_ALLOW_UNKNOWN)
663 encode_str (enc, "null", 4, 0);
637 else 664 else
638 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", 665 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
639 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 666 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
640 } 667 }
668 else if (enc->json.flags & F_ALLOW_UNKNOWN)
669 encode_str (enc, "null", 4, 0);
641 else 670 else
642 croak ("encountered %s, but JSON can only represent references to arrays or hashes", 671 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
643 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 672 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
644} 673}
645 674
707 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 736 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
708 } 737 }
709 } 738 }
710 else if (SvROK (sv)) 739 else if (SvROK (sv))
711 encode_rv (enc, SvRV (sv)); 740 encode_rv (enc, SvRV (sv));
712 else if (!SvOK (sv)) 741 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
713 encode_str (enc, "null", 4, 0); 742 encode_str (enc, "null", 4, 0);
714 else 743 else
715 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", 744 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
716 SvPV_nolen (sv), SvFLAGS (sv)); 745 SvPV_nolen (sv), SvFLAGS (sv));
717} 746}
727 enc.json = *json; 756 enc.json = *json;
728 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 757 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
729 enc.cur = SvPVX (enc.sv); 758 enc.cur = SvPVX (enc.sv);
730 enc.end = SvEND (enc.sv); 759 enc.end = SvEND (enc.sv);
731 enc.indent = 0; 760 enc.indent = 0;
732 enc.maxdepth = DEC_DEPTH (enc.json.flags);
733 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL 761 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
734 : enc.json.flags & F_LATIN1 ? 0x000100UL 762 : enc.json.flags & F_LATIN1 ? 0x000100UL
735 : 0x110000UL; 763 : 0x110000UL;
736 764
737 SvPOK_only (enc.sv); 765 SvPOK_only (enc.sv);
738 encode_sv (&enc, scalar); 766 encode_sv (&enc, scalar);
767 encode_nl (&enc);
739 768
740 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 769 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
741 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 770 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
742 771
743 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) 772 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
804 if (*dec->cur != ch) \ 833 if (*dec->cur != ch) \
805 ERR (# ch " expected"); \ 834 ERR (# ch " expected"); \
806 ++dec->cur; \ 835 ++dec->cur; \
807 SE 836 SE
808 837
809#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") 838#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
810#define DEC_DEC_DEPTH --dec->depth 839#define DEC_DEC_DEPTH --dec->depth
811 840
812static SV *decode_sv (dec_t *dec); 841static SV *decode_sv (dec_t *dec);
813 842
814static signed char decode_hexdigit[256]; 843static signed char decode_hexdigit[256];
923 else if (expect_true (ch >= 0x20 && ch < 0x80)) 952 else if (expect_true (ch >= 0x20 && ch < 0x80))
924 *cur++ = ch; 953 *cur++ = ch;
925 else if (ch >= 0x80) 954 else if (ch >= 0x80)
926 { 955 {
927 STRLEN clen; 956 STRLEN clen;
928 UV uch;
929 957
930 --dec_cur; 958 --dec_cur;
931 959
932 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 960 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
933 if (clen == (STRLEN)-1) 961 if (clen == (STRLEN)-1)
934 ERR ("malformed UTF-8 character in JSON string"); 962 ERR ("malformed UTF-8 character in JSON string");
935 963
936 do 964 do
937 *cur++ = *dec_cur++; 965 *cur++ = *dec_cur++;
954 { 982 {
955 STRLEN len = cur - buf; 983 STRLEN len = cur - buf;
956 984
957 if (sv) 985 if (sv)
958 { 986 {
959 SvGROW (sv, SvCUR (sv) + len + 1); 987 STRLEN cur = SvCUR (sv);
988
989 if (SvLEN (sv) <= cur + len)
990 SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
991
960 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 992 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
961 SvCUR_set (sv, SvCUR (sv) + len); 993 SvCUR_set (sv, SvCUR (sv) + len);
962 } 994 }
963 else 995 else
964 sv = newSVpvn (buf, len); 996 sv = newSVpvn (buf, len);
1055 1087
1056 // special case the rather common 1..5-digit-int case 1088 // special case the rather common 1..5-digit-int case
1057 if (*start == '-') 1089 if (*start == '-')
1058 switch (len) 1090 switch (len)
1059 { 1091 {
1060 case 2: return newSViv (-( start [1] - '0' * 1)); 1092 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1061 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1093 case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1062 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); 1094 case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1063 case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); 1095 case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
1064 case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); 1096 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1065 } 1097 }
1066 else 1098 else
1067 switch (len) 1099 switch (len)
1068 { 1100 {
1069 case 1: return newSViv ( start [0] - '0' * 1); 1101 case 1: return newSViv ( start [0] - '0' * 1);
1070 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1102 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1071 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); 1103 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1072 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); 1104 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1073 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); 1105 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1074 } 1106 }
1075 1107
1076 { 1108 {
1077 UV uv; 1109 UV uv;
1078 int numtype = grok_number (start, len, &uv); 1110 int numtype = grok_number (start, len, &uv);
1344 case 't': 1376 case 't':
1345 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1377 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1346 { 1378 {
1347 dec->cur += 4; 1379 dec->cur += 4;
1348#if JSON_SLOW 1380#if JSON_SLOW
1349 json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true); 1381 json_true = get_bool ("JSON::XS::true");
1350#endif 1382#endif
1351 return SvREFCNT_inc (json_true); 1383 return newSVsv (json_true);
1352 } 1384 }
1353 else 1385 else
1354 ERR ("'true' expected"); 1386 ERR ("'true' expected");
1355 1387
1356 break; 1388 break;
1358 case 'f': 1390 case 'f':
1359 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1391 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1360 { 1392 {
1361 dec->cur += 5; 1393 dec->cur += 5;
1362#if JSON_SLOW 1394#if JSON_SLOW
1363 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); 1395 json_false = get_bool ("JSON::XS::false");
1364#endif 1396#endif
1365 return SvREFCNT_inc (json_false); 1397 return newSVsv (json_false);
1366 } 1398 }
1367 else 1399 else
1368 ERR ("'false' expected"); 1400 ERR ("'false' expected");
1369 1401
1370 break; 1402 break;
1388fail: 1420fail:
1389 return 0; 1421 return 0;
1390} 1422}
1391 1423
1392static SV * 1424static SV *
1393decode_json (SV *string, JSON *json, STRLEN *offset_return) 1425decode_json (SV *string, JSON *json, char **offset_return)
1394{ 1426{
1395 dec_t dec; 1427 dec_t dec;
1396 STRLEN offset;
1397 SV *sv; 1428 SV *sv;
1398 1429
1430 /* work around bugs in 5.10 where manipulating magic values
1431 * will perl ignore the magic in subsequent accesses
1432 */
1399 SvGETMAGIC (string); 1433 /*SvGETMAGIC (string);*/
1434 if (SvMAGICAL (string))
1435 string = sv_2mortal (newSVsv (string));
1436
1400 SvUPGRADE (string, SVt_PV); 1437 SvUPGRADE (string, SVt_PV);
1401 1438
1402 if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) 1439 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1440 * assertion with -DDEBUGGING, although SvCUR is documented to
1441 * return the xpv_cur field which certainly exists after upgrading.
1442 * according to nicholas clark, calling SvPOK fixes this.
1443 * But it doesn't fix it, so try another workaround, call SvPV_nolen
1444 * and hope for the best.
1445 * Damnit, SvPV_nolen still trips over yet another assertion. This
1446 * assertion business is seriously broken, try yet another workaround
1447 * for the broken -DDEBUGGING.
1448 */
1449 {
1450#ifdef DEBUGGING
1451 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1452#else
1453 STRLEN offset = SvCUR (string);
1454#endif
1455
1456 if (offset > json->max_size && json->max_size)
1403 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 1457 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1404 (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags)); 1458 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1459 }
1405 1460
1406 if (json->flags & F_UTF8) 1461 if (json->flags & F_UTF8)
1407 sv_utf8_downgrade (string, 0); 1462 sv_utf8_downgrade (string, 0);
1408 else 1463 else
1409 sv_utf8_upgrade (string); 1464 sv_utf8_upgrade (string);
1410 1465
1411 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1466 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1412 1467
1413 dec.json = *json; 1468 dec.json = *json;
1414 dec.cur = SvPVX (string); 1469 dec.cur = SvPVX (string);
1415 dec.end = SvEND (string); 1470 dec.end = SvEND (string);
1416 dec.err = 0; 1471 dec.err = 0;
1417 dec.depth = 0; 1472 dec.depth = 0;
1418 dec.maxdepth = DEC_DEPTH (dec.json.flags);
1419 1473
1420 if (dec.json.cb_object || dec.json.cb_sk_object) 1474 if (dec.json.cb_object || dec.json.cb_sk_object)
1421 dec.json.flags |= F_HOOK; 1475 dec.json.flags |= F_HOOK;
1422 1476
1423 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1477 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1424 1478
1425 decode_ws (&dec); 1479 decode_ws (&dec);
1426 sv = decode_sv (&dec); 1480 sv = decode_sv (&dec);
1481
1482 if (offset_return)
1483 *offset_return = dec.cur;
1427 1484
1428 if (!(offset_return || !sv)) 1485 if (!(offset_return || !sv))
1429 { 1486 {
1430 // check for trailing garbage 1487 // check for trailing garbage
1431 decode_ws (&dec); 1488 decode_ws (&dec);
1434 { 1491 {
1435 dec.err = "garbage after JSON object"; 1492 dec.err = "garbage after JSON object";
1436 SvREFCNT_dec (sv); 1493 SvREFCNT_dec (sv);
1437 sv = 0; 1494 sv = 0;
1438 } 1495 }
1439 }
1440
1441 if (offset_return || !sv)
1442 {
1443 offset = dec.json.flags & F_UTF8
1444 ? dec.cur - SvPVX (string)
1445 : utf8_distance (dec.cur, SvPVX (string));
1446
1447 if (offset_return)
1448 *offset_return = offset;
1449 } 1496 }
1450 1497
1451 if (!sv) 1498 if (!sv)
1452 { 1499 {
1453 SV *uni = sv_newmortal (); 1500 SV *uni = sv_newmortal ();
1459 SAVEVPTR (PL_curcop); 1506 SAVEVPTR (PL_curcop);
1460 PL_curcop = &cop; 1507 PL_curcop = &cop;
1461 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1508 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1462 LEAVE; 1509 LEAVE;
1463 1510
1464 croak ("%s, at character offset %d [\"%s\"]", 1511 croak ("%s, at character offset %d (before \"%s\")",
1465 dec.err, 1512 dec.err,
1466 (int)offset, 1513 ptr_to_index (string, dec.cur),
1467 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1514 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1468 } 1515 }
1469 1516
1470 sv = sv_2mortal (sv); 1517 sv = sv_2mortal (sv);
1471 1518
1480 1527
1481static void 1528static void
1482incr_parse (JSON *self) 1529incr_parse (JSON *self)
1483{ 1530{
1484 const char *p = SvPVX (self->incr_text) + self->incr_pos; 1531 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1532
1533 // the state machine here is a bit convoluted and could be simplified a lot
1534 // but this would make it slower, so...
1485 1535
1486 for (;;) 1536 for (;;)
1487 { 1537 {
1488 //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 1538 //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
1489 switch (self->incr_mode) 1539 switch (self->incr_mode)
1490 { 1540 {
1491 // only used for intiial whitespace skipping 1541 // only used for initial whitespace skipping
1492 case INCR_M_WS: 1542 case INCR_M_WS:
1493 for (;;) 1543 for (;;)
1494 { 1544 {
1495 if (*p > 0x20) 1545 if (*p > 0x20)
1496 { 1546 {
1547 if (*p == '#')
1548 {
1549 self->incr_mode = INCR_M_C0;
1550 goto incr_m_c;
1551 }
1552 else
1553 {
1497 self->incr_mode = INCR_M_JSON; 1554 self->incr_mode = INCR_M_JSON;
1498 goto incr_m_json; 1555 goto incr_m_json;
1556 }
1499 } 1557 }
1500 else if (!*p) 1558 else if (!*p)
1501 goto interrupt; 1559 goto interrupt;
1502 1560
1503 ++p; 1561 ++p;
1509 goto interrupt; 1567 goto interrupt;
1510 1568
1511 ++p; 1569 ++p;
1512 self->incr_mode = INCR_M_STR; 1570 self->incr_mode = INCR_M_STR;
1513 goto incr_m_str; 1571 goto incr_m_str;
1572
1573 // inside #-style comments
1574 case INCR_M_C0:
1575 case INCR_M_C1:
1576 incr_m_c:
1577 for (;;)
1578 {
1579 if (*p == '\n')
1580 {
1581 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1582 break;
1583 }
1584 else if (!*p)
1585 goto interrupt;
1586
1587 ++p;
1588 }
1589
1590 break;
1514 1591
1515 // inside a string 1592 // inside a string
1516 case INCR_M_STR: 1593 case INCR_M_STR:
1517 incr_m_str: 1594 incr_m_str:
1518 for (;;) 1595 for (;;)
1569 self->incr_mode = INCR_M_STR; 1646 self->incr_mode = INCR_M_STR;
1570 goto incr_m_str; 1647 goto incr_m_str;
1571 1648
1572 case '[': 1649 case '[':
1573 case '{': 1650 case '{':
1574 ++self->incr_nest; 1651 if (++self->incr_nest > self->max_depth)
1652 croak (ERR_NESTING_EXCEEDED);
1575 break; 1653 break;
1576 1654
1577 case ']': 1655 case ']':
1578 case '}': 1656 case '}':
1579 if (!--self->incr_nest) 1657 if (--self->incr_nest <= 0)
1580 goto interrupt; 1658 goto interrupt;
1659 break;
1660
1661 case '#':
1662 self->incr_mode = INCR_M_C1;
1663 goto incr_m_c;
1581 } 1664 }
1582 } 1665 }
1583 } 1666 }
1584 1667
1585 modechange: 1668 modechange:
1586 ; 1669 ;
1587 } 1670 }
1588 1671
1589interrupt: 1672interrupt:
1590 self->incr_pos = p - SvPVX (self->incr_text); 1673 self->incr_pos = p - SvPVX (self->incr_text);
1674 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1591 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D 1675 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1592} 1676}
1593 1677
1594///////////////////////////////////////////////////////////////////////////// 1678/////////////////////////////////////////////////////////////////////////////
1595// XS interface functions 1679// XS interface functions
1608 : -1; 1692 : -1;
1609 1693
1610 json_stash = gv_stashpv ("JSON::XS" , 1); 1694 json_stash = gv_stashpv ("JSON::XS" , 1);
1611 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 1695 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1612 1696
1613 json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true ); 1697 json_true = get_bool ("JSON::XS::true");
1614 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); 1698 json_false = get_bool ("JSON::XS::false");
1699
1700 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1615} 1701}
1616 1702
1617PROTOTYPES: DISABLE 1703PROTOTYPES: DISABLE
1618 1704
1619void CLONE (...) 1705void CLONE (...)
1624void new (char *klass) 1710void new (char *klass)
1625 PPCODE: 1711 PPCODE:
1626{ 1712{
1627 SV *pv = NEWSV (0, sizeof (JSON)); 1713 SV *pv = NEWSV (0, sizeof (JSON));
1628 SvPOK_only (pv); 1714 SvPOK_only (pv);
1629 Zero (SvPVX (pv), 1, JSON); 1715 json_init ((JSON *)SvPVX (pv));
1630 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1631 XPUSHs (sv_2mortal (sv_bless ( 1716 XPUSHs (sv_2mortal (sv_bless (
1632 newRV_noinc (pv), 1717 newRV_noinc (pv),
1633 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 1718 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1634 ))); 1719 )));
1635} 1720}
1647 allow_nonref = F_ALLOW_NONREF 1732 allow_nonref = F_ALLOW_NONREF
1648 shrink = F_SHRINK 1733 shrink = F_SHRINK
1649 allow_blessed = F_ALLOW_BLESSED 1734 allow_blessed = F_ALLOW_BLESSED
1650 convert_blessed = F_CONV_BLESSED 1735 convert_blessed = F_CONV_BLESSED
1651 relaxed = F_RELAXED 1736 relaxed = F_RELAXED
1737 allow_unknown = F_ALLOW_UNKNOWN
1652 PPCODE: 1738 PPCODE:
1653{ 1739{
1654 if (enable) 1740 if (enable)
1655 self->flags |= ix; 1741 self->flags |= ix;
1656 else 1742 else
1671 get_allow_nonref = F_ALLOW_NONREF 1757 get_allow_nonref = F_ALLOW_NONREF
1672 get_shrink = F_SHRINK 1758 get_shrink = F_SHRINK
1673 get_allow_blessed = F_ALLOW_BLESSED 1759 get_allow_blessed = F_ALLOW_BLESSED
1674 get_convert_blessed = F_CONV_BLESSED 1760 get_convert_blessed = F_CONV_BLESSED
1675 get_relaxed = F_RELAXED 1761 get_relaxed = F_RELAXED
1762 get_allow_unknown = F_ALLOW_UNKNOWN
1676 PPCODE: 1763 PPCODE:
1677 XPUSHs (boolSV (self->flags & ix)); 1764 XPUSHs (boolSV (self->flags & ix));
1678 1765
1679void max_depth (JSON *self, UV max_depth = 0x80000000UL) 1766void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1680 PPCODE: 1767 PPCODE:
1681{ 1768 self->max_depth = max_depth;
1682 UV log2 = 0;
1683
1684 if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1685
1686 while ((1UL << log2) < max_depth)
1687 ++log2;
1688
1689 self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1690
1691 XPUSHs (ST (0)); 1769 XPUSHs (ST (0));
1692}
1693 1770
1694U32 get_max_depth (JSON *self) 1771U32 get_max_depth (JSON *self)
1695 CODE: 1772 CODE:
1696 RETVAL = DEC_DEPTH (self->flags); 1773 RETVAL = self->max_depth;
1697 OUTPUT: 1774 OUTPUT:
1698 RETVAL 1775 RETVAL
1699 1776
1700void max_size (JSON *self, UV max_size = 0) 1777void max_size (JSON *self, U32 max_size = 0)
1701 PPCODE: 1778 PPCODE:
1702{ 1779 self->max_size = max_size;
1703 UV log2 = 0;
1704
1705 if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1706 if (max_size == 1) max_size = 2;
1707
1708 while ((1UL << log2) < max_size)
1709 ++log2;
1710
1711 self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1712
1713 XPUSHs (ST (0)); 1780 XPUSHs (ST (0));
1714}
1715 1781
1716int get_max_size (JSON *self) 1782int get_max_size (JSON *self)
1717 CODE: 1783 CODE:
1718 RETVAL = DEC_SIZE (self->flags); 1784 RETVAL = self->max_size;
1719 OUTPUT: 1785 OUTPUT:
1720 RETVAL 1786 RETVAL
1721 1787
1722void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) 1788void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1723 PPCODE: 1789 PPCODE:
1759 XPUSHs (decode_json (jsonstr, self, 0)); 1825 XPUSHs (decode_json (jsonstr, self, 0));
1760 1826
1761void decode_prefix (JSON *self, SV *jsonstr) 1827void decode_prefix (JSON *self, SV *jsonstr)
1762 PPCODE: 1828 PPCODE:
1763{ 1829{
1764 STRLEN offset; 1830 char *offset;
1765 EXTEND (SP, 2); 1831 EXTEND (SP, 2);
1766 PUSHs (decode_json (jsonstr, self, &offset)); 1832 PUSHs (decode_json (jsonstr, self, &offset));
1767 PUSHs (sv_2mortal (newSVuv (offset))); 1833 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
1768} 1834}
1769 1835
1770void incr_parse (JSON *self, SV *jsonstr = 0) 1836void incr_parse (JSON *self, SV *jsonstr = 0)
1771 PPCODE: 1837 PPCODE:
1772{ 1838{
1774 self->incr_text = newSVpvn ("", 0); 1840 self->incr_text = newSVpvn ("", 0);
1775 1841
1776 // append data, if any 1842 // append data, if any
1777 if (jsonstr) 1843 if (jsonstr)
1778 { 1844 {
1779 if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text)) 1845 if (SvUTF8 (jsonstr))
1780 { 1846 {
1847 if (!SvUTF8 (self->incr_text))
1848 {
1781 /* utf-8-ness differs, need to upgrade */ 1849 /* utf-8-ness differs, need to upgrade */
1782 sv_utf8_upgrade (self->incr_text); 1850 sv_utf8_upgrade (self->incr_text);
1783 1851
1784 if (self->incr_pos) 1852 if (self->incr_pos)
1785 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos) 1853 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1786 - (U8 *)SvPVX (self->incr_text); 1854 - (U8 *)SvPVX (self->incr_text);
1855 }
1787 } 1856 }
1857 else if (SvUTF8 (self->incr_text))
1858 sv_utf8_upgrade (jsonstr);
1788 1859
1789 { 1860 {
1790 STRLEN len; 1861 STRLEN len;
1791 const char *str = SvPV (jsonstr, len); 1862 const char *str = SvPV (jsonstr, len);
1792 SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1); 1863 STRLEN cur = SvCUR (self->incr_text);
1864
1865 if (SvLEN (self->incr_text) <= cur + len)
1866 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
1867
1793 Move (str, SvEND (self->incr_text), len, char); 1868 Move (str, SvEND (self->incr_text), len, char);
1794 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); 1869 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
1795 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there 1870 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1796 } 1871 }
1797 } 1872 }
1798 1873
1799 if (GIMME_V != G_VOID) 1874 if (GIMME_V != G_VOID)
1800 do 1875 do
1801 { 1876 {
1802 STRLEN offset; 1877 char *offset;
1803
1804 incr_parse (self);
1805 1878
1806 if (!INCR_DONE (self)) 1879 if (!INCR_DONE (self))
1880 {
1881 incr_parse (self);
1882
1883 if (self->incr_pos > self->max_size && self->max_size)
1884 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1885 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
1886
1887 if (!INCR_DONE (self))
1807 break; 1888 break;
1889 }
1808 1890
1809 XPUSHs (decode_json (self->incr_text, self, &offset)); 1891 XPUSHs (decode_json (self->incr_text, self, &offset));
1810 1892
1811 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
1812 self->incr_pos -= offset; 1893 self->incr_pos -= offset - SvPVX (self->incr_text);
1813 self->incr_nest = 0; 1894 self->incr_nest = 0;
1814 self->incr_mode = 0; 1895 self->incr_mode = 0;
1896
1897 sv_chop (self->incr_text, offset);
1815 } 1898 }
1816 while (GIMME_V == G_ARRAY); 1899 while (GIMME_V == G_ARRAY);
1817} 1900}
1818 1901
1819SV *incr_text (JSON *self) 1902SV *incr_text (JSON *self)
1820 ATTRS: lvalue 1903 ATTRS: lvalue
1821 CODE: 1904 CODE:
1822{ 1905{
1823 if (self->incr_pos) 1906 if (self->incr_pos)
1824 croak ("incr_text can only be called after a successful incr_parse call in scalar context %d", self->incr_pos);//D 1907 croak ("incr_text can not be called when the incremental parser already started parsing");
1825 1908
1826 RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef; 1909 RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
1827} 1910}
1828 OUTPUT: 1911 OUTPUT:
1829 RETVAL 1912 RETVAL
1913
1914void incr_skip (JSON *self)
1915 CODE:
1916{
1917 if (self->incr_pos)
1918 {
1919 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
1920 self->incr_pos = 0;
1921 self->incr_nest = 0;
1922 self->incr_mode = 0;
1923 }
1924}
1925
1926void incr_reset (JSON *self)
1927 CODE:
1928{
1929 SvREFCNT_dec (self->incr_text);
1930 self->incr_text = 0;
1931 self->incr_pos = 0;
1932 self->incr_nest = 0;
1933 self->incr_mode = 0;
1934}
1830 1935
1831void DESTROY (JSON *self) 1936void DESTROY (JSON *self)
1832 CODE: 1937 CODE:
1833 SvREFCNT_dec (self->cb_sk_object); 1938 SvREFCNT_dec (self->cb_sk_object);
1834 SvREFCNT_dec (self->cb_object); 1939 SvREFCNT_dec (self->cb_object);
1840 ALIAS: 1945 ALIAS:
1841 to_json_ = 0 1946 to_json_ = 0
1842 encode_json = F_UTF8 1947 encode_json = F_UTF8
1843 PPCODE: 1948 PPCODE:
1844{ 1949{
1845 JSON json = { F_DEFAULT | ix }; 1950 JSON json;
1951 json_init (&json);
1952 json.flags |= ix;
1846 XPUSHs (encode_json (scalar, &json)); 1953 XPUSHs (encode_json (scalar, &json));
1847} 1954}
1848 1955
1849void decode_json (SV *jsonstr) 1956void decode_json (SV *jsonstr)
1850 ALIAS: 1957 ALIAS:
1851 from_json_ = 0 1958 from_json_ = 0
1852 decode_json = F_UTF8 1959 decode_json = F_UTF8
1853 PPCODE: 1960 PPCODE:
1854{ 1961{
1855 JSON json = { F_DEFAULT | ix }; 1962 JSON json;
1963 json_init (&json);
1964 json.flags |= ix;
1856 XPUSHs (decode_json (jsonstr, &json, 0)); 1965 XPUSHs (decode_json (jsonstr, &json, 0));
1857} 1966}
1858 1967
1859 1968

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines