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 1.105 by root, Tue Jan 19 01:07:27 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)
198{
199 UV uaccum = 0;
200 int eaccum = 0;
201
202 for (;;)
203 {
204 U8 dig = (U8)*s - '0';
205
206 if (expect_false (dig >= 10))
207 {
208 if (dig == (U8)((U8)'.' - (U8)'0'))
209 {
210 ++s;
211 json_atof_scan1 (s, accum, expo, 1);
212 }
213 else if ((dig | ' ') == 'e' - '0')
214 {
215 int exp2 = 0;
216 int neg = 0;
217
218 ++s;
219
220 if (*s == '-')
221 {
222 ++s;
223 neg = 1;
224 }
225 else if (*s == '+')
226 ++s;
227
228 while ((dig = (U8)*s - '0') < 10)
229 exp2 = exp2 * 10 + *s++ - '0';
230
231 *expo += neg ? -exp2 : exp2;
232 }
233
234 break;
235 }
236
237 ++s;
238
239 uaccum = uaccum * 10 + dig;
240 ++eaccum;
241
242 // if we have too many digits, then recurse for more
243 // we actually do this for rather few digits
244 if (uaccum >= (UV_MAX - 9) / 10)
245 {
246 if (postdp) *expo -= eaccum;
247 json_atof_scan1 (s, accum, expo, postdp);
248 if (postdp) *expo += eaccum;
249
250 break;
251 }
252 }
253
254 // this relies greatly on the quality of the pow ()
255 // implementation of the platform, but a good
256 // implementation is hard to beat.
257 if (postdp) *expo -= eaccum;
258 *accum += uaccum * Perl_pow (10., *expo);
259 *expo += eaccum;
260}
261
262static NV
263json_atof (const char *s)
264{
265 NV accum = 0.;
266 int expo = 0;
267 int neg = 0;
268
269 if (*s == '-')
270 {
271 ++s;
272 neg = 1;
273 }
274
275 json_atof_scan1 (s, &accum, &expo, 0);
276
277 return neg ? -accum : accum;
278}
175///////////////////////////////////////////////////////////////////////////// 279/////////////////////////////////////////////////////////////////////////////
176// encoder 280// encoder
177 281
178// structure used for encoding JSON 282// structure used for encoding JSON
179typedef struct 283typedef struct
181 char *cur; // SvPVX (sv) + current output position 285 char *cur; // SvPVX (sv) + current output position
182 char *end; // SvEND (sv) 286 char *end; // SvEND (sv)
183 SV *sv; // result scalar 287 SV *sv; // result scalar
184 JSON json; 288 JSON json;
185 U32 indent; // indentation level 289 U32 indent; // indentation level
186 U32 maxdepth; // max. indentation/recursion level
187 UV limit; // escape character values >= this value when encoding 290 UV limit; // escape character values >= this value when encoding
188} enc_t; 291} enc_t;
189 292
190INLINE void 293INLINE void
191need (enc_t *enc, STRLEN len) 294need (enc_t *enc, STRLEN len)
192{ 295{
193 if (expect_false (enc->cur + len >= enc->end)) 296 if (expect_false (enc->cur + len >= enc->end))
194 { 297 {
195 STRLEN cur = enc->cur - SvPVX (enc->sv); 298 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
196 SvGROW (enc->sv, cur + len + 1); 299 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
197 enc->cur = SvPVX (enc->sv) + cur; 300 enc->cur = SvPVX (enc->sv) + cur;
198 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 301 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
199 } 302 }
200} 303}
201 304
276 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 379 (int)((uch - 0x10000) % 0x400 + 0xDC00));
277 enc->cur += 12; 380 enc->cur += 12;
278 } 381 }
279 else 382 else
280 { 383 {
281 static char hexdigit [16] = "0123456789abcdef";
282 need (enc, len += 5); 384 need (enc, len += 5);
283 *enc->cur++ = '\\'; 385 *enc->cur++ = '\\';
284 *enc->cur++ = 'u'; 386 *enc->cur++ = 'u';
285 *enc->cur++ = hexdigit [ uch >> 12 ]; 387 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
286 *enc->cur++ = hexdigit [(uch >> 8) & 15]; 388 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
287 *enc->cur++ = hexdigit [(uch >> 4) & 15]; 389 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
288 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 390 *enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
289 } 391 }
290 392
291 str += clen; 393 str += clen;
292 } 394 }
293 else if (enc->json.flags & F_LATIN1) 395 else if (enc->json.flags & F_LATIN1)
364static void 466static void
365encode_av (enc_t *enc, AV *av) 467encode_av (enc_t *enc, AV *av)
366{ 468{
367 int i, len = av_len (av); 469 int i, len = av_len (av);
368 470
369 if (enc->indent >= enc->maxdepth) 471 if (enc->indent >= enc->json.max_depth)
370 croak ("data structure too deep (hit recursion limit)"); 472 croak (ERR_NESTING_EXCEEDED);
371 473
372 encode_ch (enc, '['); 474 encode_ch (enc, '[');
373 475
374 if (len >= 0) 476 if (len >= 0)
375 { 477 {
450static void 552static void
451encode_hv (enc_t *enc, HV *hv) 553encode_hv (enc_t *enc, HV *hv)
452{ 554{
453 HE *he; 555 HE *he;
454 556
455 if (enc->indent >= enc->maxdepth) 557 if (enc->indent >= enc->json.max_depth)
456 croak ("data structure too deep (hit recursion limit)"); 558 croak (ERR_NESTING_EXCEEDED);
457 559
458 encode_ch (enc, '{'); 560 encode_ch (enc, '{');
459 561
460 // for canonical output we have to sort by keys first 562 // for canonical output we have to sort by keys first
461 // actually, this is mostly due to the stupid so-called 563 // actually, this is mostly due to the stupid so-called
462 // security workaround added somewhere in 5.8.x. 564 // security workaround added somewhere in 5.8.x
463 // that randomises hash orderings 565 // that randomises hash orderings
464 if (enc->json.flags & F_CANONICAL) 566 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
465 { 567 {
466 int count = hv_iterinit (hv); 568 int count = hv_iterinit (hv);
467 569
468 if (SvMAGICAL (hv)) 570 if (SvMAGICAL (hv))
469 { 571 {
642 744
643 if (len == 1 && *pv == '1') 745 if (len == 1 && *pv == '1')
644 encode_str (enc, "true", 4, 0); 746 encode_str (enc, "true", 4, 0);
645 else if (len == 1 && *pv == '0') 747 else if (len == 1 && *pv == '0')
646 encode_str (enc, "false", 5, 0); 748 encode_str (enc, "false", 5, 0);
749 else if (enc->json.flags & F_ALLOW_UNKNOWN)
750 encode_str (enc, "null", 4, 0);
647 else 751 else
648 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", 752 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
649 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 753 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
650 } 754 }
755 else if (enc->json.flags & F_ALLOW_UNKNOWN)
756 encode_str (enc, "null", 4, 0);
651 else 757 else
652 croak ("encountered %s, but JSON can only represent references to arrays or hashes", 758 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
653 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 759 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
654} 760}
655 761
717 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 823 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
718 } 824 }
719 } 825 }
720 else if (SvROK (sv)) 826 else if (SvROK (sv))
721 encode_rv (enc, SvRV (sv)); 827 encode_rv (enc, SvRV (sv));
722 else if (!SvOK (sv)) 828 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
723 encode_str (enc, "null", 4, 0); 829 encode_str (enc, "null", 4, 0);
724 else 830 else
725 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", 831 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
726 SvPV_nolen (sv), SvFLAGS (sv)); 832 SvPV_nolen (sv), SvFLAGS (sv));
727} 833}
737 enc.json = *json; 843 enc.json = *json;
738 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 844 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
739 enc.cur = SvPVX (enc.sv); 845 enc.cur = SvPVX (enc.sv);
740 enc.end = SvEND (enc.sv); 846 enc.end = SvEND (enc.sv);
741 enc.indent = 0; 847 enc.indent = 0;
742 enc.maxdepth = DEC_DEPTH (enc.json.flags);
743 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL 848 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
744 : enc.json.flags & F_LATIN1 ? 0x000100UL 849 : enc.json.flags & F_LATIN1 ? 0x000100UL
745 : 0x110000UL; 850 : 0x110000UL;
746 851
747 SvPOK_only (enc.sv); 852 SvPOK_only (enc.sv);
748 encode_sv (&enc, scalar); 853 encode_sv (&enc, scalar);
854 encode_nl (&enc);
749 855
750 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 856 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
751 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 857 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
752 858
753 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) 859 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
814 if (*dec->cur != ch) \ 920 if (*dec->cur != ch) \
815 ERR (# ch " expected"); \ 921 ERR (# ch " expected"); \
816 ++dec->cur; \ 922 ++dec->cur; \
817 SE 923 SE
818 924
819#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") 925#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
820#define DEC_DEC_DEPTH --dec->depth 926#define DEC_DEC_DEPTH --dec->depth
821 927
822static SV *decode_sv (dec_t *dec); 928static SV *decode_sv (dec_t *dec);
823 929
824static signed char decode_hexdigit[256]; 930static signed char decode_hexdigit[256];
933 else if (expect_true (ch >= 0x20 && ch < 0x80)) 1039 else if (expect_true (ch >= 0x20 && ch < 0x80))
934 *cur++ = ch; 1040 *cur++ = ch;
935 else if (ch >= 0x80) 1041 else if (ch >= 0x80)
936 { 1042 {
937 STRLEN clen; 1043 STRLEN clen;
938 UV uch;
939 1044
940 --dec_cur; 1045 --dec_cur;
941 1046
942 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 1047 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
943 if (clen == (STRLEN)-1) 1048 if (clen == (STRLEN)-1)
944 ERR ("malformed UTF-8 character in JSON string"); 1049 ERR ("malformed UTF-8 character in JSON string");
945 1050
946 do 1051 do
947 *cur++ = *dec_cur++; 1052 *cur++ = *dec_cur++;
964 { 1069 {
965 STRLEN len = cur - buf; 1070 STRLEN len = cur - buf;
966 1071
967 if (sv) 1072 if (sv)
968 { 1073 {
969 SvGROW (sv, SvCUR (sv) + len + 1); 1074 STRLEN cur = SvCUR (sv);
1075
1076 if (SvLEN (sv) <= cur + len)
1077 SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
1078
970 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1079 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
971 SvCUR_set (sv, SvCUR (sv) + len); 1080 SvCUR_set (sv, SvCUR (sv) + len);
972 } 1081 }
973 else 1082 else
974 sv = newSVpvn (buf, len); 1083 sv = newSVpvn (buf, len);
1065 1174
1066 // special case the rather common 1..5-digit-int case 1175 // special case the rather common 1..5-digit-int case
1067 if (*start == '-') 1176 if (*start == '-')
1068 switch (len) 1177 switch (len)
1069 { 1178 {
1070 case 2: return newSViv (-( start [1] - '0' * 1)); 1179 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1071 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1180 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)); 1181 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)); 1182 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)); 1183 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1075 } 1184 }
1076 else 1185 else
1077 switch (len) 1186 switch (len)
1078 { 1187 {
1079 case 1: return newSViv ( start [0] - '0' * 1); 1188 case 1: return newSViv ( start [0] - '0' * 1);
1080 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1189 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); 1190 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); 1191 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); 1192 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1084 } 1193 }
1085 1194
1086 { 1195 {
1087 UV uv; 1196 UV uv;
1088 int numtype = grok_number (start, len, &uv); 1197 int numtype = grok_number (start, len, &uv);
1097 } 1206 }
1098 1207
1099 len -= *start == '-' ? 1 : 0; 1208 len -= *start == '-' ? 1 : 0;
1100 1209
1101 // does not fit into IV or UV, try NV 1210 // does not fit into IV or UV, try NV
1102 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1211 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 1212 // fits into NV without loss of precision
1108 return newSVnv (Atof (start)); 1213 return newSVnv (json_atof (start));
1109 1214
1110 // everything else fails, convert it to a string 1215 // everything else fails, convert it to a string
1111 return newSVpvn (start, dec->cur - start); 1216 return newSVpvn (start, dec->cur - start);
1112 } 1217 }
1113 1218
1114 // loss of precision here 1219 // loss of precision here
1115 return newSVnv (Atof (start)); 1220 return newSVnv (json_atof (start));
1116 1221
1117fail: 1222fail:
1118 return 0; 1223 return 0;
1119} 1224}
1120 1225
1398fail: 1503fail:
1399 return 0; 1504 return 0;
1400} 1505}
1401 1506
1402static SV * 1507static SV *
1403decode_json (SV *string, JSON *json, STRLEN *offset_return) 1508decode_json (SV *string, JSON *json, char **offset_return)
1404{ 1509{
1405 dec_t dec; 1510 dec_t dec;
1406 STRLEN offset;
1407 SV *sv; 1511 SV *sv;
1408 1512
1513 /* work around bugs in 5.10 where manipulating magic values
1514 * will perl ignore the magic in subsequent accesses
1515 */
1409 SvGETMAGIC (string); 1516 /*SvGETMAGIC (string);*/
1517 if (SvMAGICAL (string))
1518 string = sv_2mortal (newSVsv (string));
1519
1410 SvUPGRADE (string, SVt_PV); 1520 SvUPGRADE (string, SVt_PV);
1411 1521
1412 if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) 1522 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1523 * assertion with -DDEBUGGING, although SvCUR is documented to
1524 * return the xpv_cur field which certainly exists after upgrading.
1525 * according to nicholas clark, calling SvPOK fixes this.
1526 * But it doesn't fix it, so try another workaround, call SvPV_nolen
1527 * and hope for the best.
1528 * Damnit, SvPV_nolen still trips over yet another assertion. This
1529 * assertion business is seriously broken, try yet another workaround
1530 * for the broken -DDEBUGGING.
1531 */
1532 {
1533#ifdef DEBUGGING
1534 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1535#else
1536 STRLEN offset = SvCUR (string);
1537#endif
1538
1539 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", 1540 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)); 1541 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1542 }
1415 1543
1416 if (json->flags & F_UTF8) 1544 if (json->flags & F_UTF8)
1417 sv_utf8_downgrade (string, 0); 1545 sv_utf8_downgrade (string, 0);
1418 else 1546 else
1419 sv_utf8_upgrade (string); 1547 sv_utf8_upgrade (string);
1420 1548
1421 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1549 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1422 1550
1423 dec.json = *json; 1551 dec.json = *json;
1424 dec.cur = SvPVX (string); 1552 dec.cur = SvPVX (string);
1425 dec.end = SvEND (string); 1553 dec.end = SvEND (string);
1426 dec.err = 0; 1554 dec.err = 0;
1427 dec.depth = 0; 1555 dec.depth = 0;
1428 dec.maxdepth = DEC_DEPTH (dec.json.flags);
1429 1556
1430 if (dec.json.cb_object || dec.json.cb_sk_object) 1557 if (dec.json.cb_object || dec.json.cb_sk_object)
1431 dec.json.flags |= F_HOOK; 1558 dec.json.flags |= F_HOOK;
1432 1559
1433 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1560 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1434 1561
1435 decode_ws (&dec); 1562 decode_ws (&dec);
1436 sv = decode_sv (&dec); 1563 sv = decode_sv (&dec);
1564
1565 if (offset_return)
1566 *offset_return = dec.cur;
1437 1567
1438 if (!(offset_return || !sv)) 1568 if (!(offset_return || !sv))
1439 { 1569 {
1440 // check for trailing garbage 1570 // check for trailing garbage
1441 decode_ws (&dec); 1571 decode_ws (&dec);
1444 { 1574 {
1445 dec.err = "garbage after JSON object"; 1575 dec.err = "garbage after JSON object";
1446 SvREFCNT_dec (sv); 1576 SvREFCNT_dec (sv);
1447 sv = 0; 1577 sv = 0;
1448 } 1578 }
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 } 1579 }
1460 1580
1461 if (!sv) 1581 if (!sv)
1462 { 1582 {
1463 SV *uni = sv_newmortal (); 1583 SV *uni = sv_newmortal ();
1469 SAVEVPTR (PL_curcop); 1589 SAVEVPTR (PL_curcop);
1470 PL_curcop = &cop; 1590 PL_curcop = &cop;
1471 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1591 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1472 LEAVE; 1592 LEAVE;
1473 1593
1474 croak ("%s, at character offset %d [\"%s\"]", 1594 croak ("%s, at character offset %d (before \"%s\")",
1475 dec.err, 1595 dec.err,
1476 (int)offset, 1596 ptr_to_index (string, dec.cur),
1477 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1597 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1478 } 1598 }
1479 1599
1480 sv = sv_2mortal (sv); 1600 sv = sv_2mortal (sv);
1481 1601
1490 1610
1491static void 1611static void
1492incr_parse (JSON *self) 1612incr_parse (JSON *self)
1493{ 1613{
1494 const char *p = SvPVX (self->incr_text) + self->incr_pos; 1614 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1615
1616 // the state machine here is a bit convoluted and could be simplified a lot
1617 // but this would make it slower, so...
1495 1618
1496 for (;;) 1619 for (;;)
1497 { 1620 {
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 1621 //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) 1622 switch (self->incr_mode)
1500 { 1623 {
1501 // only used for intiial whitespace skipping 1624 // only used for initial whitespace skipping
1502 case INCR_M_WS: 1625 case INCR_M_WS:
1503 for (;;) 1626 for (;;)
1504 { 1627 {
1505 if (*p > 0x20) 1628 if (*p > 0x20)
1506 { 1629 {
1630 if (*p == '#')
1631 {
1632 self->incr_mode = INCR_M_C0;
1633 goto incr_m_c;
1634 }
1635 else
1636 {
1507 self->incr_mode = INCR_M_JSON; 1637 self->incr_mode = INCR_M_JSON;
1508 goto incr_m_json; 1638 goto incr_m_json;
1639 }
1509 } 1640 }
1510 else if (!*p) 1641 else if (!*p)
1511 goto interrupt; 1642 goto interrupt;
1512 1643
1513 ++p; 1644 ++p;
1519 goto interrupt; 1650 goto interrupt;
1520 1651
1521 ++p; 1652 ++p;
1522 self->incr_mode = INCR_M_STR; 1653 self->incr_mode = INCR_M_STR;
1523 goto incr_m_str; 1654 goto incr_m_str;
1655
1656 // inside #-style comments
1657 case INCR_M_C0:
1658 case INCR_M_C1:
1659 incr_m_c:
1660 for (;;)
1661 {
1662 if (*p == '\n')
1663 {
1664 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1665 break;
1666 }
1667 else if (!*p)
1668 goto interrupt;
1669
1670 ++p;
1671 }
1672
1673 break;
1524 1674
1525 // inside a string 1675 // inside a string
1526 case INCR_M_STR: 1676 case INCR_M_STR:
1527 incr_m_str: 1677 incr_m_str:
1528 for (;;) 1678 for (;;)
1579 self->incr_mode = INCR_M_STR; 1729 self->incr_mode = INCR_M_STR;
1580 goto incr_m_str; 1730 goto incr_m_str;
1581 1731
1582 case '[': 1732 case '[':
1583 case '{': 1733 case '{':
1584 ++self->incr_nest; 1734 if (++self->incr_nest > self->max_depth)
1735 croak (ERR_NESTING_EXCEEDED);
1585 break; 1736 break;
1586 1737
1587 case ']': 1738 case ']':
1588 case '}': 1739 case '}':
1589 if (!--self->incr_nest) 1740 if (--self->incr_nest <= 0)
1590 goto interrupt; 1741 goto interrupt;
1742 break;
1743
1744 case '#':
1745 self->incr_mode = INCR_M_C1;
1746 goto incr_m_c;
1591 } 1747 }
1592 } 1748 }
1593 } 1749 }
1594 1750
1595 modechange: 1751 modechange:
1596 ; 1752 ;
1597 } 1753 }
1598 1754
1599interrupt: 1755interrupt:
1600 self->incr_pos = p - SvPVX (self->incr_text); 1756 self->incr_pos = p - SvPVX (self->incr_text);
1757 //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 1758 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1602} 1759}
1603 1760
1604///////////////////////////////////////////////////////////////////////////// 1761/////////////////////////////////////////////////////////////////////////////
1605// XS interface functions 1762// XS interface functions
1620 json_stash = gv_stashpv ("JSON::XS" , 1); 1777 json_stash = gv_stashpv ("JSON::XS" , 1);
1621 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 1778 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1622 1779
1623 json_true = get_bool ("JSON::XS::true"); 1780 json_true = get_bool ("JSON::XS::true");
1624 json_false = get_bool ("JSON::XS::false"); 1781 json_false = get_bool ("JSON::XS::false");
1782
1783 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1625} 1784}
1626 1785
1627PROTOTYPES: DISABLE 1786PROTOTYPES: DISABLE
1628 1787
1629void CLONE (...) 1788void CLONE (...)
1634void new (char *klass) 1793void new (char *klass)
1635 PPCODE: 1794 PPCODE:
1636{ 1795{
1637 SV *pv = NEWSV (0, sizeof (JSON)); 1796 SV *pv = NEWSV (0, sizeof (JSON));
1638 SvPOK_only (pv); 1797 SvPOK_only (pv);
1639 Zero (SvPVX (pv), 1, JSON); 1798 json_init ((JSON *)SvPVX (pv));
1640 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1641 XPUSHs (sv_2mortal (sv_bless ( 1799 XPUSHs (sv_2mortal (sv_bless (
1642 newRV_noinc (pv), 1800 newRV_noinc (pv),
1643 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 1801 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1644 ))); 1802 )));
1645} 1803}
1657 allow_nonref = F_ALLOW_NONREF 1815 allow_nonref = F_ALLOW_NONREF
1658 shrink = F_SHRINK 1816 shrink = F_SHRINK
1659 allow_blessed = F_ALLOW_BLESSED 1817 allow_blessed = F_ALLOW_BLESSED
1660 convert_blessed = F_CONV_BLESSED 1818 convert_blessed = F_CONV_BLESSED
1661 relaxed = F_RELAXED 1819 relaxed = F_RELAXED
1820 allow_unknown = F_ALLOW_UNKNOWN
1662 PPCODE: 1821 PPCODE:
1663{ 1822{
1664 if (enable) 1823 if (enable)
1665 self->flags |= ix; 1824 self->flags |= ix;
1666 else 1825 else
1681 get_allow_nonref = F_ALLOW_NONREF 1840 get_allow_nonref = F_ALLOW_NONREF
1682 get_shrink = F_SHRINK 1841 get_shrink = F_SHRINK
1683 get_allow_blessed = F_ALLOW_BLESSED 1842 get_allow_blessed = F_ALLOW_BLESSED
1684 get_convert_blessed = F_CONV_BLESSED 1843 get_convert_blessed = F_CONV_BLESSED
1685 get_relaxed = F_RELAXED 1844 get_relaxed = F_RELAXED
1845 get_allow_unknown = F_ALLOW_UNKNOWN
1686 PPCODE: 1846 PPCODE:
1687 XPUSHs (boolSV (self->flags & ix)); 1847 XPUSHs (boolSV (self->flags & ix));
1688 1848
1689void max_depth (JSON *self, UV max_depth = 0x80000000UL) 1849void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1690 PPCODE: 1850 PPCODE:
1691{ 1851 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)); 1852 XPUSHs (ST (0));
1702}
1703 1853
1704U32 get_max_depth (JSON *self) 1854U32 get_max_depth (JSON *self)
1705 CODE: 1855 CODE:
1706 RETVAL = DEC_DEPTH (self->flags); 1856 RETVAL = self->max_depth;
1707 OUTPUT: 1857 OUTPUT:
1708 RETVAL 1858 RETVAL
1709 1859
1710void max_size (JSON *self, UV max_size = 0) 1860void max_size (JSON *self, U32 max_size = 0)
1711 PPCODE: 1861 PPCODE:
1712{ 1862 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)); 1863 XPUSHs (ST (0));
1724}
1725 1864
1726int get_max_size (JSON *self) 1865int get_max_size (JSON *self)
1727 CODE: 1866 CODE:
1728 RETVAL = DEC_SIZE (self->flags); 1867 RETVAL = self->max_size;
1729 OUTPUT: 1868 OUTPUT:
1730 RETVAL 1869 RETVAL
1731 1870
1732void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) 1871void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1733 PPCODE: 1872 PPCODE:
1769 XPUSHs (decode_json (jsonstr, self, 0)); 1908 XPUSHs (decode_json (jsonstr, self, 0));
1770 1909
1771void decode_prefix (JSON *self, SV *jsonstr) 1910void decode_prefix (JSON *self, SV *jsonstr)
1772 PPCODE: 1911 PPCODE:
1773{ 1912{
1774 STRLEN offset; 1913 char *offset;
1775 EXTEND (SP, 2); 1914 EXTEND (SP, 2);
1776 PUSHs (decode_json (jsonstr, self, &offset)); 1915 PUSHs (decode_json (jsonstr, self, &offset));
1777 PUSHs (sv_2mortal (newSVuv (offset))); 1916 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
1778} 1917}
1779 1918
1780void incr_parse (JSON *self, SV *jsonstr = 0) 1919void incr_parse (JSON *self, SV *jsonstr = 0)
1781 PPCODE: 1920 PPCODE:
1782{ 1921{
1784 self->incr_text = newSVpvn ("", 0); 1923 self->incr_text = newSVpvn ("", 0);
1785 1924
1786 // append data, if any 1925 // append data, if any
1787 if (jsonstr) 1926 if (jsonstr)
1788 { 1927 {
1789 if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text)) 1928 if (SvUTF8 (jsonstr))
1790 { 1929 {
1930 if (!SvUTF8 (self->incr_text))
1931 {
1791 /* utf-8-ness differs, need to upgrade */ 1932 /* utf-8-ness differs, need to upgrade */
1792 sv_utf8_upgrade (self->incr_text); 1933 sv_utf8_upgrade (self->incr_text);
1793 1934
1794 if (self->incr_pos) 1935 if (self->incr_pos)
1795 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos) 1936 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1796 - (U8 *)SvPVX (self->incr_text); 1937 - (U8 *)SvPVX (self->incr_text);
1938 }
1797 } 1939 }
1940 else if (SvUTF8 (self->incr_text))
1941 sv_utf8_upgrade (jsonstr);
1798 1942
1799 { 1943 {
1800 STRLEN len; 1944 STRLEN len;
1801 const char *str = SvPV (jsonstr, len); 1945 const char *str = SvPV (jsonstr, len);
1802 SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1); 1946 STRLEN cur = SvCUR (self->incr_text);
1947
1948 if (SvLEN (self->incr_text) <= cur + len)
1949 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
1950
1803 Move (str, SvEND (self->incr_text), len, char); 1951 Move (str, SvEND (self->incr_text), len, char);
1804 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); 1952 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 1953 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1806 } 1954 }
1807 } 1955 }
1808 1956
1809 if (GIMME_V != G_VOID) 1957 if (GIMME_V != G_VOID)
1810 do 1958 do
1811 { 1959 {
1812 STRLEN offset; 1960 char *offset;
1813 1961
1814 if (!INCR_DONE (self)) 1962 if (!INCR_DONE (self))
1815 { 1963 {
1816 incr_parse (self); 1964 incr_parse (self);
1965
1966 if (self->incr_pos > self->max_size && self->max_size)
1967 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1968 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
1969
1817 if (!INCR_DONE (self)) 1970 if (!INCR_DONE (self))
1818 break; 1971 break;
1819 } 1972 }
1820 1973
1821 XPUSHs (decode_json (self->incr_text, self, &offset)); 1974 XPUSHs (decode_json (self->incr_text, self, &offset));
1822 1975
1823 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
1824 self->incr_pos -= offset; 1976 self->incr_pos -= offset - SvPVX (self->incr_text);
1825 self->incr_nest = 0; 1977 self->incr_nest = 0;
1826 self->incr_mode = 0; 1978 self->incr_mode = 0;
1979
1980 sv_chop (self->incr_text, offset);
1827 } 1981 }
1828 while (GIMME_V == G_ARRAY); 1982 while (GIMME_V == G_ARRAY);
1829} 1983}
1830 1984
1831SV *incr_text (JSON *self) 1985SV *incr_text (JSON *self)
1850 self->incr_nest = 0; 2004 self->incr_nest = 0;
1851 self->incr_mode = 0; 2005 self->incr_mode = 0;
1852 } 2006 }
1853} 2007}
1854 2008
2009void incr_reset (JSON *self)
2010 CODE:
2011{
2012 SvREFCNT_dec (self->incr_text);
2013 self->incr_text = 0;
2014 self->incr_pos = 0;
2015 self->incr_nest = 0;
2016 self->incr_mode = 0;
2017}
2018
1855void DESTROY (JSON *self) 2019void DESTROY (JSON *self)
1856 CODE: 2020 CODE:
1857 SvREFCNT_dec (self->cb_sk_object); 2021 SvREFCNT_dec (self->cb_sk_object);
1858 SvREFCNT_dec (self->cb_object); 2022 SvREFCNT_dec (self->cb_object);
1859 SvREFCNT_dec (self->incr_text); 2023 SvREFCNT_dec (self->incr_text);
1864 ALIAS: 2028 ALIAS:
1865 to_json_ = 0 2029 to_json_ = 0
1866 encode_json = F_UTF8 2030 encode_json = F_UTF8
1867 PPCODE: 2031 PPCODE:
1868{ 2032{
1869 JSON json = { F_DEFAULT | ix }; 2033 JSON json;
2034 json_init (&json);
2035 json.flags |= ix;
1870 XPUSHs (encode_json (scalar, &json)); 2036 XPUSHs (encode_json (scalar, &json));
1871} 2037}
1872 2038
1873void decode_json (SV *jsonstr) 2039void decode_json (SV *jsonstr)
1874 ALIAS: 2040 ALIAS:
1875 from_json_ = 0 2041 from_json_ = 0
1876 decode_json = F_UTF8 2042 decode_json = F_UTF8
1877 PPCODE: 2043 PPCODE:
1878{ 2044{
1879 JSON json = { F_DEFAULT | ix }; 2045 JSON json;
2046 json_init (&json);
2047 json.flags |= ix;
1880 XPUSHs (decode_json (jsonstr, &json, 0)); 2048 XPUSHs (decode_json (jsonstr, &json, 0));
1881} 2049}
1882 2050
1883

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines