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 1.83 by root, Wed Mar 26 22:54:38 2008 UTC vs.
Revision 1.106 by root, Tue Jan 19 01:36:34 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
107INLINE SV * 114INLINE SV *
117 124
118INLINE void 125INLINE void
119shrink (SV *sv) 126shrink (SV *sv)
120{ 127{
121 sv_utf8_downgrade (sv, 1); 128 sv_utf8_downgrade (sv, 1);
129
122 if (SvLEN (sv) > SvCUR (sv) + 1) 130 if (SvLEN (sv) > SvCUR (sv) + 1)
123 { 131 {
124#ifdef SvPV_shrink_to_cur 132#ifdef SvPV_shrink_to_cur
125 SvPV_shrink_to_cur (sv); 133 SvPV_shrink_to_cur (sv);
126#elif defined (SvPV_renew) 134#elif defined (SvPV_renew)
170 *s++ = 0x80 | ( ch & 0x3f); 178 *s++ = 0x80 | ( ch & 0x3f);
171 179
172 return s; 180 return s;
173} 181}
174 182
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
192/////////////////////////////////////////////////////////////////////////////
193// fp hell
194
195// scan a group of digits, and a trailing exponent
196static void
197json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
198{
199 UV uaccum = 0;
200 int eaccum = 0;
201
202 // if we recurse too deep, skip all remaining digits
203 // to avoid a stack overflow attack
204 if (expect_false (--maxdepth <= 0))
205 while (((U8)*s - '0') < 10)
206 ++s;
207
208 for (;;)
209 {
210 U8 dig = (U8)*s - '0';
211
212 if (expect_false (dig >= 10))
213 {
214 if (dig == (U8)((U8)'.' - (U8)'0'))
215 {
216 ++s;
217 json_atof_scan1 (s, accum, expo, 1, maxdepth);
218 }
219 else if ((dig | ' ') == 'e' - '0')
220 {
221 int exp2 = 0;
222 int neg = 0;
223
224 ++s;
225
226 if (*s == '-')
227 {
228 ++s;
229 neg = 1;
230 }
231 else if (*s == '+')
232 ++s;
233
234 while ((dig = (U8)*s - '0') < 10)
235 exp2 = exp2 * 10 + *s++ - '0';
236
237 *expo += neg ? -exp2 : exp2;
238 }
239
240 break;
241 }
242
243 ++s;
244
245 uaccum = uaccum * 10 + dig;
246 ++eaccum;
247
248 // if we have too many digits, then recurse for more
249 // we actually do this for rather few digits
250 if (uaccum >= (UV_MAX - 9) / 10)
251 {
252 if (postdp) *expo -= eaccum;
253 json_atof_scan1 (s, accum, expo, postdp, maxdepth);
254 if (postdp) *expo += eaccum;
255
256 break;
257 }
258 }
259
260 // this relies greatly on the quality of the pow ()
261 // implementation of the platform, but a good
262 // implementation is hard to beat.
263 if (postdp) *expo -= eaccum;
264 *accum += uaccum * Perl_pow (10., *expo);
265 *expo += eaccum;
266}
267
268static NV
269json_atof (const char *s)
270{
271 NV accum = 0.;
272 int expo = 0;
273 int neg = 0;
274
275 if (*s == '-')
276 {
277 ++s;
278 neg = 1;
279 }
280
281 // a recursion depth of ten gives us >>500 bits
282 json_atof_scan1 (s, &accum, &expo, 0, 10);
283
284 return neg ? -accum : accum;
285}
175///////////////////////////////////////////////////////////////////////////// 286/////////////////////////////////////////////////////////////////////////////
176// encoder 287// encoder
177 288
178// structure used for encoding JSON 289// structure used for encoding JSON
179typedef struct 290typedef struct
181 char *cur; // SvPVX (sv) + current output position 292 char *cur; // SvPVX (sv) + current output position
182 char *end; // SvEND (sv) 293 char *end; // SvEND (sv)
183 SV *sv; // result scalar 294 SV *sv; // result scalar
184 JSON json; 295 JSON json;
185 U32 indent; // indentation level 296 U32 indent; // indentation level
186 U32 maxdepth; // max. indentation/recursion level
187 UV limit; // escape character values >= this value when encoding 297 UV limit; // escape character values >= this value when encoding
188} enc_t; 298} enc_t;
189 299
190INLINE void 300INLINE void
191need (enc_t *enc, STRLEN len) 301need (enc_t *enc, STRLEN len)
192{ 302{
193 if (expect_false (enc->cur + len >= enc->end)) 303 if (expect_false (enc->cur + len >= enc->end))
194 { 304 {
195 STRLEN cur = enc->cur - SvPVX (enc->sv); 305 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
196 SvGROW (enc->sv, cur + len + 1); 306 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
197 enc->cur = SvPVX (enc->sv) + cur; 307 enc->cur = SvPVX (enc->sv) + cur;
198 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 308 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
199 } 309 }
200} 310}
201 311
276 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 386 (int)((uch - 0x10000) % 0x400 + 0xDC00));
277 enc->cur += 12; 387 enc->cur += 12;
278 } 388 }
279 else 389 else
280 { 390 {
281 static char hexdigit [16] = "0123456789abcdef";
282 need (enc, len += 5); 391 need (enc, len += 5);
283 *enc->cur++ = '\\'; 392 *enc->cur++ = '\\';
284 *enc->cur++ = 'u'; 393 *enc->cur++ = 'u';
285 *enc->cur++ = hexdigit [ uch >> 12 ]; 394 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
286 *enc->cur++ = hexdigit [(uch >> 8) & 15]; 395 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
287 *enc->cur++ = hexdigit [(uch >> 4) & 15]; 396 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
288 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 397 *enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
289 } 398 }
290 399
291 str += clen; 400 str += clen;
292 } 401 }
293 else if (enc->json.flags & F_LATIN1) 402 else if (enc->json.flags & F_LATIN1)
364static void 473static void
365encode_av (enc_t *enc, AV *av) 474encode_av (enc_t *enc, AV *av)
366{ 475{
367 int i, len = av_len (av); 476 int i, len = av_len (av);
368 477
369 if (enc->indent >= enc->maxdepth) 478 if (enc->indent >= enc->json.max_depth)
370 croak ("data structure too deep (hit recursion limit)"); 479 croak (ERR_NESTING_EXCEEDED);
371 480
372 encode_ch (enc, '['); 481 encode_ch (enc, '[');
373 482
374 if (len >= 0) 483 if (len >= 0)
375 { 484 {
450static void 559static void
451encode_hv (enc_t *enc, HV *hv) 560encode_hv (enc_t *enc, HV *hv)
452{ 561{
453 HE *he; 562 HE *he;
454 563
455 if (enc->indent >= enc->maxdepth) 564 if (enc->indent >= enc->json.max_depth)
456 croak ("data structure too deep (hit recursion limit)"); 565 croak (ERR_NESTING_EXCEEDED);
457 566
458 encode_ch (enc, '{'); 567 encode_ch (enc, '{');
459 568
460 // for canonical output we have to sort by keys first 569 // for canonical output we have to sort by keys first
461 // actually, this is mostly due to the stupid so-called 570 // actually, this is mostly due to the stupid so-called
462 // security workaround added somewhere in 5.8.x. 571 // security workaround added somewhere in 5.8.x
463 // that randomises hash orderings 572 // that randomises hash orderings
464 if (enc->json.flags & F_CANONICAL) 573 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
465 { 574 {
466 int count = hv_iterinit (hv); 575 int count = hv_iterinit (hv);
467 576
468 if (SvMAGICAL (hv)) 577 if (SvMAGICAL (hv))
469 { 578 {
642 751
643 if (len == 1 && *pv == '1') 752 if (len == 1 && *pv == '1')
644 encode_str (enc, "true", 4, 0); 753 encode_str (enc, "true", 4, 0);
645 else if (len == 1 && *pv == '0') 754 else if (len == 1 && *pv == '0')
646 encode_str (enc, "false", 5, 0); 755 encode_str (enc, "false", 5, 0);
756 else if (enc->json.flags & F_ALLOW_UNKNOWN)
757 encode_str (enc, "null", 4, 0);
647 else 758 else
648 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", 759 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
649 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 760 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
650 } 761 }
762 else if (enc->json.flags & F_ALLOW_UNKNOWN)
763 encode_str (enc, "null", 4, 0);
651 else 764 else
652 croak ("encountered %s, but JSON can only represent references to arrays or hashes", 765 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
653 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 766 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
654} 767}
655 768
717 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 830 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
718 } 831 }
719 } 832 }
720 else if (SvROK (sv)) 833 else if (SvROK (sv))
721 encode_rv (enc, SvRV (sv)); 834 encode_rv (enc, SvRV (sv));
722 else if (!SvOK (sv)) 835 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
723 encode_str (enc, "null", 4, 0); 836 encode_str (enc, "null", 4, 0);
724 else 837 else
725 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", 838 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
726 SvPV_nolen (sv), SvFLAGS (sv)); 839 SvPV_nolen (sv), SvFLAGS (sv));
727} 840}
737 enc.json = *json; 850 enc.json = *json;
738 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 851 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
739 enc.cur = SvPVX (enc.sv); 852 enc.cur = SvPVX (enc.sv);
740 enc.end = SvEND (enc.sv); 853 enc.end = SvEND (enc.sv);
741 enc.indent = 0; 854 enc.indent = 0;
742 enc.maxdepth = DEC_DEPTH (enc.json.flags);
743 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL 855 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
744 : enc.json.flags & F_LATIN1 ? 0x000100UL 856 : enc.json.flags & F_LATIN1 ? 0x000100UL
745 : 0x110000UL; 857 : 0x110000UL;
746 858
747 SvPOK_only (enc.sv); 859 SvPOK_only (enc.sv);
748 encode_sv (&enc, scalar); 860 encode_sv (&enc, scalar);
861 encode_nl (&enc);
749 862
750 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 863 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
751 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 864 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
752 865
753 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) 866 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
814 if (*dec->cur != ch) \ 927 if (*dec->cur != ch) \
815 ERR (# ch " expected"); \ 928 ERR (# ch " expected"); \
816 ++dec->cur; \ 929 ++dec->cur; \
817 SE 930 SE
818 931
819#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") 932#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
820#define DEC_DEC_DEPTH --dec->depth 933#define DEC_DEC_DEPTH --dec->depth
821 934
822static SV *decode_sv (dec_t *dec); 935static SV *decode_sv (dec_t *dec);
823 936
824static signed char decode_hexdigit[256]; 937static signed char decode_hexdigit[256];
933 else if (expect_true (ch >= 0x20 && ch < 0x80)) 1046 else if (expect_true (ch >= 0x20 && ch < 0x80))
934 *cur++ = ch; 1047 *cur++ = ch;
935 else if (ch >= 0x80) 1048 else if (ch >= 0x80)
936 { 1049 {
937 STRLEN clen; 1050 STRLEN clen;
938 UV uch;
939 1051
940 --dec_cur; 1052 --dec_cur;
941 1053
942 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 1054 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
943 if (clen == (STRLEN)-1) 1055 if (clen == (STRLEN)-1)
944 ERR ("malformed UTF-8 character in JSON string"); 1056 ERR ("malformed UTF-8 character in JSON string");
945 1057
946 do 1058 do
947 *cur++ = *dec_cur++; 1059 *cur++ = *dec_cur++;
964 { 1076 {
965 STRLEN len = cur - buf; 1077 STRLEN len = cur - buf;
966 1078
967 if (sv) 1079 if (sv)
968 { 1080 {
969 SvGROW (sv, SvCUR (sv) + len + 1); 1081 STRLEN cur = SvCUR (sv);
1082
1083 if (SvLEN (sv) <= cur + len)
1084 SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
1085
970 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1086 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
971 SvCUR_set (sv, SvCUR (sv) + len); 1087 SvCUR_set (sv, SvCUR (sv) + len);
972 } 1088 }
973 else 1089 else
974 sv = newSVpvn (buf, len); 1090 sv = newSVpvn (buf, len);
1065 1181
1066 // special case the rather common 1..5-digit-int case 1182 // special case the rather common 1..5-digit-int case
1067 if (*start == '-') 1183 if (*start == '-')
1068 switch (len) 1184 switch (len)
1069 { 1185 {
1070 case 2: return newSViv (-( start [1] - '0' * 1)); 1186 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1071 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1187 case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1072 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); 1188 case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1073 case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); 1189 case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
1074 case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); 1190 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1075 } 1191 }
1076 else 1192 else
1077 switch (len) 1193 switch (len)
1078 { 1194 {
1079 case 1: return newSViv ( start [0] - '0' * 1); 1195 case 1: return newSViv ( start [0] - '0' * 1);
1080 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1196 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1081 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); 1197 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1082 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); 1198 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1083 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); 1199 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1084 } 1200 }
1085 1201
1086 { 1202 {
1087 UV uv; 1203 UV uv;
1088 int numtype = grok_number (start, len, &uv); 1204 int numtype = grok_number (start, len, &uv);
1097 } 1213 }
1098 1214
1099 len -= *start == '-' ? 1 : 0; 1215 len -= *start == '-' ? 1 : 0;
1100 1216
1101 // does not fit into IV or UV, try NV 1217 // does not fit into IV or UV, try NV
1102 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1218 if (len <= NV_DIG)
1103 #if defined (LDBL_DIG)
1104 || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1105 #endif
1106 )
1107 // fits into NV without loss of precision 1219 // fits into NV without loss of precision
1108 return newSVnv (Atof (start)); 1220 return newSVnv (json_atof (start));
1109 1221
1110 // everything else fails, convert it to a string 1222 // everything else fails, convert it to a string
1111 return newSVpvn (start, dec->cur - start); 1223 return newSVpvn (start, dec->cur - start);
1112 } 1224 }
1113 1225
1114 // loss of precision here 1226 // loss of precision here
1115 return newSVnv (Atof (start)); 1227 return newSVnv (json_atof (start));
1116 1228
1117fail: 1229fail:
1118 return 0; 1230 return 0;
1119} 1231}
1120 1232
1398fail: 1510fail:
1399 return 0; 1511 return 0;
1400} 1512}
1401 1513
1402static SV * 1514static SV *
1403decode_json (SV *string, JSON *json, STRLEN *offset_return) 1515decode_json (SV *string, JSON *json, char **offset_return)
1404{ 1516{
1405 dec_t dec; 1517 dec_t dec;
1406 STRLEN offset;
1407 SV *sv; 1518 SV *sv;
1408 1519
1520 /* work around bugs in 5.10 where manipulating magic values
1521 * will perl ignore the magic in subsequent accesses
1522 */
1409 SvGETMAGIC (string); 1523 /*SvGETMAGIC (string);*/
1524 if (SvMAGICAL (string))
1525 string = sv_2mortal (newSVsv (string));
1526
1410 SvUPGRADE (string, SVt_PV); 1527 SvUPGRADE (string, SVt_PV);
1411 1528
1412 if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) 1529 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1530 * assertion with -DDEBUGGING, although SvCUR is documented to
1531 * return the xpv_cur field which certainly exists after upgrading.
1532 * according to nicholas clark, calling SvPOK fixes this.
1533 * But it doesn't fix it, so try another workaround, call SvPV_nolen
1534 * and hope for the best.
1535 * Damnit, SvPV_nolen still trips over yet another assertion. This
1536 * assertion business is seriously broken, try yet another workaround
1537 * for the broken -DDEBUGGING.
1538 */
1539 {
1540#ifdef DEBUGGING
1541 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1542#else
1543 STRLEN offset = SvCUR (string);
1544#endif
1545
1546 if (offset > json->max_size && json->max_size)
1413 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 1547 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1414 (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags)); 1548 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1549 }
1415 1550
1416 if (json->flags & F_UTF8) 1551 if (json->flags & F_UTF8)
1417 sv_utf8_downgrade (string, 0); 1552 sv_utf8_downgrade (string, 0);
1418 else 1553 else
1419 sv_utf8_upgrade (string); 1554 sv_utf8_upgrade (string);
1420 1555
1421 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1556 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1422 1557
1423 dec.json = *json; 1558 dec.json = *json;
1424 dec.cur = SvPVX (string); 1559 dec.cur = SvPVX (string);
1425 dec.end = SvEND (string); 1560 dec.end = SvEND (string);
1426 dec.err = 0; 1561 dec.err = 0;
1427 dec.depth = 0; 1562 dec.depth = 0;
1428 dec.maxdepth = DEC_DEPTH (dec.json.flags);
1429 1563
1430 if (dec.json.cb_object || dec.json.cb_sk_object) 1564 if (dec.json.cb_object || dec.json.cb_sk_object)
1431 dec.json.flags |= F_HOOK; 1565 dec.json.flags |= F_HOOK;
1432 1566
1433 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1567 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1434 1568
1435 decode_ws (&dec); 1569 decode_ws (&dec);
1436 sv = decode_sv (&dec); 1570 sv = decode_sv (&dec);
1571
1572 if (offset_return)
1573 *offset_return = dec.cur;
1437 1574
1438 if (!(offset_return || !sv)) 1575 if (!(offset_return || !sv))
1439 { 1576 {
1440 // check for trailing garbage 1577 // check for trailing garbage
1441 decode_ws (&dec); 1578 decode_ws (&dec);
1444 { 1581 {
1445 dec.err = "garbage after JSON object"; 1582 dec.err = "garbage after JSON object";
1446 SvREFCNT_dec (sv); 1583 SvREFCNT_dec (sv);
1447 sv = 0; 1584 sv = 0;
1448 } 1585 }
1449 }
1450
1451 if (offset_return || !sv)
1452 {
1453 offset = dec.json.flags & F_UTF8
1454 ? dec.cur - SvPVX (string)
1455 : utf8_distance (dec.cur, SvPVX (string));
1456
1457 if (offset_return)
1458 *offset_return = offset;
1459 } 1586 }
1460 1587
1461 if (!sv) 1588 if (!sv)
1462 { 1589 {
1463 SV *uni = sv_newmortal (); 1590 SV *uni = sv_newmortal ();
1469 SAVEVPTR (PL_curcop); 1596 SAVEVPTR (PL_curcop);
1470 PL_curcop = &cop; 1597 PL_curcop = &cop;
1471 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1598 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1472 LEAVE; 1599 LEAVE;
1473 1600
1474 croak ("%s, at character offset %d [\"%s\"]", 1601 croak ("%s, at character offset %d (before \"%s\")",
1475 dec.err, 1602 dec.err,
1476 (int)offset, 1603 ptr_to_index (string, dec.cur),
1477 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1604 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1478 } 1605 }
1479 1606
1480 sv = sv_2mortal (sv); 1607 sv = sv_2mortal (sv);
1481 1608
1490 1617
1491static void 1618static void
1492incr_parse (JSON *self) 1619incr_parse (JSON *self)
1493{ 1620{
1494 const char *p = SvPVX (self->incr_text) + self->incr_pos; 1621 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1622
1623 // the state machine here is a bit convoluted and could be simplified a lot
1624 // but this would make it slower, so...
1495 1625
1496 for (;;) 1626 for (;;)
1497 { 1627 {
1498 //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 1628 //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
1499 switch (self->incr_mode) 1629 switch (self->incr_mode)
1500 { 1630 {
1501 // only used for intiial whitespace skipping 1631 // only used for initial whitespace skipping
1502 case INCR_M_WS: 1632 case INCR_M_WS:
1503 for (;;) 1633 for (;;)
1504 { 1634 {
1505 if (*p > 0x20) 1635 if (*p > 0x20)
1506 { 1636 {
1637 if (*p == '#')
1638 {
1639 self->incr_mode = INCR_M_C0;
1640 goto incr_m_c;
1641 }
1642 else
1643 {
1507 self->incr_mode = INCR_M_JSON; 1644 self->incr_mode = INCR_M_JSON;
1508 goto incr_m_json; 1645 goto incr_m_json;
1646 }
1509 } 1647 }
1510 else if (!*p) 1648 else if (!*p)
1511 goto interrupt; 1649 goto interrupt;
1512 1650
1513 ++p; 1651 ++p;
1519 goto interrupt; 1657 goto interrupt;
1520 1658
1521 ++p; 1659 ++p;
1522 self->incr_mode = INCR_M_STR; 1660 self->incr_mode = INCR_M_STR;
1523 goto incr_m_str; 1661 goto incr_m_str;
1662
1663 // inside #-style comments
1664 case INCR_M_C0:
1665 case INCR_M_C1:
1666 incr_m_c:
1667 for (;;)
1668 {
1669 if (*p == '\n')
1670 {
1671 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1672 break;
1673 }
1674 else if (!*p)
1675 goto interrupt;
1676
1677 ++p;
1678 }
1679
1680 break;
1524 1681
1525 // inside a string 1682 // inside a string
1526 case INCR_M_STR: 1683 case INCR_M_STR:
1527 incr_m_str: 1684 incr_m_str:
1528 for (;;) 1685 for (;;)
1579 self->incr_mode = INCR_M_STR; 1736 self->incr_mode = INCR_M_STR;
1580 goto incr_m_str; 1737 goto incr_m_str;
1581 1738
1582 case '[': 1739 case '[':
1583 case '{': 1740 case '{':
1584 ++self->incr_nest; 1741 if (++self->incr_nest > self->max_depth)
1742 croak (ERR_NESTING_EXCEEDED);
1585 break; 1743 break;
1586 1744
1587 case ']': 1745 case ']':
1588 case '}': 1746 case '}':
1589 if (!--self->incr_nest) 1747 if (--self->incr_nest <= 0)
1590 goto interrupt; 1748 goto interrupt;
1749 break;
1750
1751 case '#':
1752 self->incr_mode = INCR_M_C1;
1753 goto incr_m_c;
1591 } 1754 }
1592 } 1755 }
1593 } 1756 }
1594 1757
1595 modechange: 1758 modechange:
1596 ; 1759 ;
1597 } 1760 }
1598 1761
1599interrupt: 1762interrupt:
1600 self->incr_pos = p - SvPVX (self->incr_text); 1763 self->incr_pos = p - SvPVX (self->incr_text);
1764 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1601 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D 1765 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1602} 1766}
1603 1767
1604///////////////////////////////////////////////////////////////////////////// 1768/////////////////////////////////////////////////////////////////////////////
1605// XS interface functions 1769// XS interface functions
1620 json_stash = gv_stashpv ("JSON::XS" , 1); 1784 json_stash = gv_stashpv ("JSON::XS" , 1);
1621 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 1785 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1622 1786
1623 json_true = get_bool ("JSON::XS::true"); 1787 json_true = get_bool ("JSON::XS::true");
1624 json_false = get_bool ("JSON::XS::false"); 1788 json_false = get_bool ("JSON::XS::false");
1789
1790 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1625} 1791}
1626 1792
1627PROTOTYPES: DISABLE 1793PROTOTYPES: DISABLE
1628 1794
1629void CLONE (...) 1795void CLONE (...)
1634void new (char *klass) 1800void new (char *klass)
1635 PPCODE: 1801 PPCODE:
1636{ 1802{
1637 SV *pv = NEWSV (0, sizeof (JSON)); 1803 SV *pv = NEWSV (0, sizeof (JSON));
1638 SvPOK_only (pv); 1804 SvPOK_only (pv);
1639 Zero (SvPVX (pv), 1, JSON); 1805 json_init ((JSON *)SvPVX (pv));
1640 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1641 XPUSHs (sv_2mortal (sv_bless ( 1806 XPUSHs (sv_2mortal (sv_bless (
1642 newRV_noinc (pv), 1807 newRV_noinc (pv),
1643 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 1808 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1644 ))); 1809 )));
1645} 1810}
1657 allow_nonref = F_ALLOW_NONREF 1822 allow_nonref = F_ALLOW_NONREF
1658 shrink = F_SHRINK 1823 shrink = F_SHRINK
1659 allow_blessed = F_ALLOW_BLESSED 1824 allow_blessed = F_ALLOW_BLESSED
1660 convert_blessed = F_CONV_BLESSED 1825 convert_blessed = F_CONV_BLESSED
1661 relaxed = F_RELAXED 1826 relaxed = F_RELAXED
1827 allow_unknown = F_ALLOW_UNKNOWN
1662 PPCODE: 1828 PPCODE:
1663{ 1829{
1664 if (enable) 1830 if (enable)
1665 self->flags |= ix; 1831 self->flags |= ix;
1666 else 1832 else
1681 get_allow_nonref = F_ALLOW_NONREF 1847 get_allow_nonref = F_ALLOW_NONREF
1682 get_shrink = F_SHRINK 1848 get_shrink = F_SHRINK
1683 get_allow_blessed = F_ALLOW_BLESSED 1849 get_allow_blessed = F_ALLOW_BLESSED
1684 get_convert_blessed = F_CONV_BLESSED 1850 get_convert_blessed = F_CONV_BLESSED
1685 get_relaxed = F_RELAXED 1851 get_relaxed = F_RELAXED
1852 get_allow_unknown = F_ALLOW_UNKNOWN
1686 PPCODE: 1853 PPCODE:
1687 XPUSHs (boolSV (self->flags & ix)); 1854 XPUSHs (boolSV (self->flags & ix));
1688 1855
1689void max_depth (JSON *self, UV max_depth = 0x80000000UL) 1856void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1690 PPCODE: 1857 PPCODE:
1691{ 1858 self->max_depth = max_depth;
1692 UV log2 = 0;
1693
1694 if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1695
1696 while ((1UL << log2) < max_depth)
1697 ++log2;
1698
1699 self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1700
1701 XPUSHs (ST (0)); 1859 XPUSHs (ST (0));
1702}
1703 1860
1704U32 get_max_depth (JSON *self) 1861U32 get_max_depth (JSON *self)
1705 CODE: 1862 CODE:
1706 RETVAL = DEC_DEPTH (self->flags); 1863 RETVAL = self->max_depth;
1707 OUTPUT: 1864 OUTPUT:
1708 RETVAL 1865 RETVAL
1709 1866
1710void max_size (JSON *self, UV max_size = 0) 1867void max_size (JSON *self, U32 max_size = 0)
1711 PPCODE: 1868 PPCODE:
1712{ 1869 self->max_size = max_size;
1713 UV log2 = 0;
1714
1715 if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1716 if (max_size == 1) max_size = 2;
1717
1718 while ((1UL << log2) < max_size)
1719 ++log2;
1720
1721 self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1722
1723 XPUSHs (ST (0)); 1870 XPUSHs (ST (0));
1724}
1725 1871
1726int get_max_size (JSON *self) 1872int get_max_size (JSON *self)
1727 CODE: 1873 CODE:
1728 RETVAL = DEC_SIZE (self->flags); 1874 RETVAL = self->max_size;
1729 OUTPUT: 1875 OUTPUT:
1730 RETVAL 1876 RETVAL
1731 1877
1732void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) 1878void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1733 PPCODE: 1879 PPCODE:
1769 XPUSHs (decode_json (jsonstr, self, 0)); 1915 XPUSHs (decode_json (jsonstr, self, 0));
1770 1916
1771void decode_prefix (JSON *self, SV *jsonstr) 1917void decode_prefix (JSON *self, SV *jsonstr)
1772 PPCODE: 1918 PPCODE:
1773{ 1919{
1774 STRLEN offset; 1920 char *offset;
1775 EXTEND (SP, 2); 1921 EXTEND (SP, 2);
1776 PUSHs (decode_json (jsonstr, self, &offset)); 1922 PUSHs (decode_json (jsonstr, self, &offset));
1777 PUSHs (sv_2mortal (newSVuv (offset))); 1923 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
1778} 1924}
1779 1925
1780void incr_parse (JSON *self, SV *jsonstr = 0) 1926void incr_parse (JSON *self, SV *jsonstr = 0)
1781 PPCODE: 1927 PPCODE:
1782{ 1928{
1784 self->incr_text = newSVpvn ("", 0); 1930 self->incr_text = newSVpvn ("", 0);
1785 1931
1786 // append data, if any 1932 // append data, if any
1787 if (jsonstr) 1933 if (jsonstr)
1788 { 1934 {
1789 if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text)) 1935 if (SvUTF8 (jsonstr))
1790 { 1936 {
1937 if (!SvUTF8 (self->incr_text))
1938 {
1791 /* utf-8-ness differs, need to upgrade */ 1939 /* utf-8-ness differs, need to upgrade */
1792 sv_utf8_upgrade (self->incr_text); 1940 sv_utf8_upgrade (self->incr_text);
1793 1941
1794 if (self->incr_pos) 1942 if (self->incr_pos)
1795 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos) 1943 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1796 - (U8 *)SvPVX (self->incr_text); 1944 - (U8 *)SvPVX (self->incr_text);
1945 }
1797 } 1946 }
1947 else if (SvUTF8 (self->incr_text))
1948 sv_utf8_upgrade (jsonstr);
1798 1949
1799 { 1950 {
1800 STRLEN len; 1951 STRLEN len;
1801 const char *str = SvPV (jsonstr, len); 1952 const char *str = SvPV (jsonstr, len);
1802 SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1); 1953 STRLEN cur = SvCUR (self->incr_text);
1954
1955 if (SvLEN (self->incr_text) <= cur + len)
1956 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
1957
1803 Move (str, SvEND (self->incr_text), len, char); 1958 Move (str, SvEND (self->incr_text), len, char);
1804 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); 1959 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
1805 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there 1960 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1806 } 1961 }
1807 } 1962 }
1808 1963
1809 if (GIMME_V != G_VOID) 1964 if (GIMME_V != G_VOID)
1810 do 1965 do
1811 { 1966 {
1812 STRLEN offset; 1967 char *offset;
1813 1968
1814 if (!INCR_DONE (self)) 1969 if (!INCR_DONE (self))
1815 { 1970 {
1816 incr_parse (self); 1971 incr_parse (self);
1972
1973 if (self->incr_pos > self->max_size && self->max_size)
1974 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1975 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
1976
1817 if (!INCR_DONE (self)) 1977 if (!INCR_DONE (self))
1818 break; 1978 break;
1819 } 1979 }
1820 1980
1821 XPUSHs (decode_json (self->incr_text, self, &offset)); 1981 XPUSHs (decode_json (self->incr_text, self, &offset));
1822 1982
1823 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
1824 self->incr_pos -= offset; 1983 self->incr_pos -= offset - SvPVX (self->incr_text);
1825 self->incr_nest = 0; 1984 self->incr_nest = 0;
1826 self->incr_mode = 0; 1985 self->incr_mode = 0;
1986
1987 sv_chop (self->incr_text, offset);
1827 } 1988 }
1828 while (GIMME_V == G_ARRAY); 1989 while (GIMME_V == G_ARRAY);
1829} 1990}
1830 1991
1831SV *incr_text (JSON *self) 1992SV *incr_text (JSON *self)
1850 self->incr_nest = 0; 2011 self->incr_nest = 0;
1851 self->incr_mode = 0; 2012 self->incr_mode = 0;
1852 } 2013 }
1853} 2014}
1854 2015
2016void incr_reset (JSON *self)
2017 CODE:
2018{
2019 SvREFCNT_dec (self->incr_text);
2020 self->incr_text = 0;
2021 self->incr_pos = 0;
2022 self->incr_nest = 0;
2023 self->incr_mode = 0;
2024}
2025
1855void DESTROY (JSON *self) 2026void DESTROY (JSON *self)
1856 CODE: 2027 CODE:
1857 SvREFCNT_dec (self->cb_sk_object); 2028 SvREFCNT_dec (self->cb_sk_object);
1858 SvREFCNT_dec (self->cb_object); 2029 SvREFCNT_dec (self->cb_object);
1859 SvREFCNT_dec (self->incr_text); 2030 SvREFCNT_dec (self->incr_text);
1864 ALIAS: 2035 ALIAS:
1865 to_json_ = 0 2036 to_json_ = 0
1866 encode_json = F_UTF8 2037 encode_json = F_UTF8
1867 PPCODE: 2038 PPCODE:
1868{ 2039{
1869 JSON json = { F_DEFAULT | ix }; 2040 JSON json;
2041 json_init (&json);
2042 json.flags |= ix;
1870 XPUSHs (encode_json (scalar, &json)); 2043 XPUSHs (encode_json (scalar, &json));
1871} 2044}
1872 2045
1873void decode_json (SV *jsonstr) 2046void decode_json (SV *jsonstr)
1874 ALIAS: 2047 ALIAS:
1875 from_json_ = 0 2048 from_json_ = 0
1876 decode_json = F_UTF8 2049 decode_json = F_UTF8
1877 PPCODE: 2050 PPCODE:
1878{ 2051{
1879 JSON json = { F_DEFAULT | ix }; 2052 JSON json;
2053 json_init (&json);
2054 json.flags |= ix;
1880 XPUSHs (decode_json (jsonstr, &json, 0)); 2055 XPUSHs (decode_json (jsonstr, &json, 0));
1881} 2056}
1882 2057
1883

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines