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 1.84 by root, Thu Mar 27 06:37:35 2008 UTC vs.
Revision 1.109 by root, Tue Mar 8 10:18:50 2011 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
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#define F_ALLOW_UNKNOWN 0x00002000UL 37#define F_ALLOW_UNKNOWN 0x00002000UL
37
38#define F_MAXDEPTH 0xf8000000UL
39#define S_MAXDEPTH 27
40#define F_MAXSIZE 0x01f00000UL
41#define S_MAXSIZE 20
42#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing 38#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
43 39
44#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
45#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
46
47#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 40#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
48#define F_DEFAULT (9UL << S_MAXDEPTH)
49 41
50#define INIT_SIZE 32 // initial scalar size to be allocated 42#define INIT_SIZE 32 // initial scalar size to be allocated
51#define INDENT_STEP 3 // spaces per indentation level 43#define INDENT_STEP 3 // spaces per indentation level
52 44
53#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
46
47#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
54 48
55#define SB do { 49#define SB do {
56#define SE } while (0) 50#define SE } while (0)
57 51
58#if __GNUC__ >= 3 52#if __GNUC__ >= 3
68 62
69#define IN_RANGE_INC(type,val,beg,end) \ 63#define IN_RANGE_INC(type,val,beg,end) \
70 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ 64 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
71 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) 65 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
72 66
67#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
68
73#ifdef USE_ITHREADS 69#ifdef USE_ITHREADS
74# define JSON_SLOW 1 70# define JSON_SLOW 1
75# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) 71# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
76#else 72#else
77# define JSON_SLOW 0 73# define JSON_SLOW 0
83 79
84enum { 80enum {
85 INCR_M_WS = 0, // initial whitespace skipping, must be 0 81 INCR_M_WS = 0, // initial whitespace skipping, must be 0
86 INCR_M_STR, // inside string 82 INCR_M_STR, // inside string
87 INCR_M_BS, // inside backslash 83 INCR_M_BS, // inside backslash
84 INCR_M_C0, // inside comment in initial whitespace sequence
85 INCR_M_C1, // inside comment in other places
88 INCR_M_JSON // outside anything, count nesting 86 INCR_M_JSON // outside anything, count nesting
89}; 87};
90 88
91#define INCR_DONE(json) (!(json)->incr_nest && (json)->incr_mode == INCR_M_JSON) 89#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
92 90
93typedef struct { 91typedef struct {
94 U32 flags; 92 U32 flags;
93 U32 max_depth;
94 STRLEN max_size;
95
95 SV *cb_object; 96 SV *cb_object;
96 HV *cb_sk_object; 97 HV *cb_sk_object;
97 98
98 // for the incremental parser 99 // for the incremental parser
99 SV *incr_text; // the source text so far 100 SV *incr_text; // the source text so far
100 STRLEN incr_pos; // the current offset into the text 101 STRLEN incr_pos; // the current offset into the text
101 int incr_nest; // {[]}-nesting level 102 int incr_nest; // {[]}-nesting level
102 int incr_mode; 103 unsigned char incr_mode;
103} JSON; 104} JSON;
105
106INLINE void
107json_init (JSON *json)
108{
109 Zero (json, 1, JSON);
110 json->max_depth = 512;
111}
104 112
105///////////////////////////////////////////////////////////////////////////// 113/////////////////////////////////////////////////////////////////////////////
106// utility functions 114// utility functions
107 115
108INLINE SV * 116INLINE SV *
118 126
119INLINE void 127INLINE void
120shrink (SV *sv) 128shrink (SV *sv)
121{ 129{
122 sv_utf8_downgrade (sv, 1); 130 sv_utf8_downgrade (sv, 1);
131
123 if (SvLEN (sv) > SvCUR (sv) + 1) 132 if (SvLEN (sv) > SvCUR (sv) + 1)
124 { 133 {
125#ifdef SvPV_shrink_to_cur 134#ifdef SvPV_shrink_to_cur
126 SvPV_shrink_to_cur (sv); 135 SvPV_shrink_to_cur (sv);
127#elif defined (SvPV_renew) 136#elif defined (SvPV_renew)
171 *s++ = 0x80 | ( ch & 0x3f); 180 *s++ = 0x80 | ( ch & 0x3f);
172 181
173 return s; 182 return s;
174} 183}
175 184
185// convert offset pointer to character index, sv must be string
186static STRLEN
187ptr_to_index (SV *sv, char *offset)
188{
189 return SvUTF8 (sv)
190 ? utf8_distance (offset, SvPVX (sv))
191 : offset - SvPVX (sv);
192}
193
194/////////////////////////////////////////////////////////////////////////////
195// fp hell
196
197// scan a group of digits, and a trailing exponent
198static void
199json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
200{
201 UV uaccum = 0;
202 int eaccum = 0;
203
204 // if we recurse too deep, skip all remaining digits
205 // to avoid a stack overflow attack
206 if (expect_false (--maxdepth <= 0))
207 while (((U8)*s - '0') < 10)
208 ++s;
209
210 for (;;)
211 {
212 U8 dig = (U8)*s - '0';
213
214 if (expect_false (dig >= 10))
215 {
216 if (dig == (U8)((U8)'.' - (U8)'0'))
217 {
218 ++s;
219 json_atof_scan1 (s, accum, expo, 1, maxdepth);
220 }
221 else if ((dig | ' ') == 'e' - '0')
222 {
223 int exp2 = 0;
224 int neg = 0;
225
226 ++s;
227
228 if (*s == '-')
229 {
230 ++s;
231 neg = 1;
232 }
233 else if (*s == '+')
234 ++s;
235
236 while ((dig = (U8)*s - '0') < 10)
237 exp2 = exp2 * 10 + *s++ - '0';
238
239 *expo += neg ? -exp2 : exp2;
240 }
241
242 break;
243 }
244
245 ++s;
246
247 uaccum = uaccum * 10 + dig;
248 ++eaccum;
249
250 // if we have too many digits, then recurse for more
251 // we actually do this for rather few digits
252 if (uaccum >= (UV_MAX - 9) / 10)
253 {
254 if (postdp) *expo -= eaccum;
255 json_atof_scan1 (s, accum, expo, postdp, maxdepth);
256 if (postdp) *expo += eaccum;
257
258 break;
259 }
260 }
261
262 // this relies greatly on the quality of the pow ()
263 // implementation of the platform, but a good
264 // implementation is hard to beat.
265 if (postdp) *expo -= eaccum;
266 *accum += uaccum * Perl_pow (10., *expo);
267 *expo += eaccum;
268}
269
270static NV
271json_atof (const char *s)
272{
273 NV accum = 0.;
274 int expo = 0;
275 int neg = 0;
276
277 if (*s == '-')
278 {
279 ++s;
280 neg = 1;
281 }
282
283 // a recursion depth of ten gives us >>500 bits
284 json_atof_scan1 (s, &accum, &expo, 0, 10);
285
286 return neg ? -accum : accum;
287}
176///////////////////////////////////////////////////////////////////////////// 288/////////////////////////////////////////////////////////////////////////////
177// encoder 289// encoder
178 290
179// structure used for encoding JSON 291// structure used for encoding JSON
180typedef struct 292typedef struct
182 char *cur; // SvPVX (sv) + current output position 294 char *cur; // SvPVX (sv) + current output position
183 char *end; // SvEND (sv) 295 char *end; // SvEND (sv)
184 SV *sv; // result scalar 296 SV *sv; // result scalar
185 JSON json; 297 JSON json;
186 U32 indent; // indentation level 298 U32 indent; // indentation level
187 U32 maxdepth; // max. indentation/recursion level
188 UV limit; // escape character values >= this value when encoding 299 UV limit; // escape character values >= this value when encoding
189} enc_t; 300} enc_t;
190 301
191INLINE void 302INLINE void
192need (enc_t *enc, STRLEN len) 303need (enc_t *enc, STRLEN len)
193{ 304{
194 if (expect_false (enc->cur + len >= enc->end)) 305 if (expect_false (enc->cur + len >= enc->end))
195 { 306 {
196 STRLEN cur = enc->cur - SvPVX (enc->sv); 307 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
197 SvGROW (enc->sv, cur + len + 1); 308 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
198 enc->cur = SvPVX (enc->sv) + cur; 309 enc->cur = SvPVX (enc->sv) + cur;
199 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 310 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
200 } 311 }
201} 312}
202 313
277 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 388 (int)((uch - 0x10000) % 0x400 + 0xDC00));
278 enc->cur += 12; 389 enc->cur += 12;
279 } 390 }
280 else 391 else
281 { 392 {
282 static char hexdigit [16] = "0123456789abcdef";
283 need (enc, len += 5); 393 need (enc, len += 5);
284 *enc->cur++ = '\\'; 394 *enc->cur++ = '\\';
285 *enc->cur++ = 'u'; 395 *enc->cur++ = 'u';
286 *enc->cur++ = hexdigit [ uch >> 12 ]; 396 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
287 *enc->cur++ = hexdigit [(uch >> 8) & 15]; 397 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
288 *enc->cur++ = hexdigit [(uch >> 4) & 15]; 398 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
289 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 399 *enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
290 } 400 }
291 401
292 str += clen; 402 str += clen;
293 } 403 }
294 else if (enc->json.flags & F_LATIN1) 404 else if (enc->json.flags & F_LATIN1)
365static void 475static void
366encode_av (enc_t *enc, AV *av) 476encode_av (enc_t *enc, AV *av)
367{ 477{
368 int i, len = av_len (av); 478 int i, len = av_len (av);
369 479
370 if (enc->indent >= enc->maxdepth) 480 if (enc->indent >= enc->json.max_depth)
371 croak ("data structure too deep (hit recursion limit)"); 481 croak (ERR_NESTING_EXCEEDED);
372 482
373 encode_ch (enc, '['); 483 encode_ch (enc, '[');
374 484
375 if (len >= 0) 485 if (len >= 0)
376 { 486 {
451static void 561static void
452encode_hv (enc_t *enc, HV *hv) 562encode_hv (enc_t *enc, HV *hv)
453{ 563{
454 HE *he; 564 HE *he;
455 565
456 if (enc->indent >= enc->maxdepth) 566 if (enc->indent >= enc->json.max_depth)
457 croak ("data structure too deep (hit recursion limit)"); 567 croak (ERR_NESTING_EXCEEDED);
458 568
459 encode_ch (enc, '{'); 569 encode_ch (enc, '{');
460 570
461 // for canonical output we have to sort by keys first 571 // for canonical output we have to sort by keys first
462 // actually, this is mostly due to the stupid so-called 572 // actually, this is mostly due to the stupid so-called
463 // security workaround added somewhere in 5.8.x. 573 // security workaround added somewhere in 5.8.x
464 // that randomises hash orderings 574 // that randomises hash orderings
465 if (enc->json.flags & F_CANONICAL) 575 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
466 { 576 {
467 int count = hv_iterinit (hv); 577 int count = hv_iterinit (hv);
468 578
469 if (SvMAGICAL (hv)) 579 if (SvMAGICAL (hv))
470 { 580 {
742 enc.json = *json; 852 enc.json = *json;
743 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 853 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
744 enc.cur = SvPVX (enc.sv); 854 enc.cur = SvPVX (enc.sv);
745 enc.end = SvEND (enc.sv); 855 enc.end = SvEND (enc.sv);
746 enc.indent = 0; 856 enc.indent = 0;
747 enc.maxdepth = DEC_DEPTH (enc.json.flags);
748 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL 857 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
749 : enc.json.flags & F_LATIN1 ? 0x000100UL 858 : enc.json.flags & F_LATIN1 ? 0x000100UL
750 : 0x110000UL; 859 : 0x110000UL;
751 860
752 SvPOK_only (enc.sv); 861 SvPOK_only (enc.sv);
753 encode_sv (&enc, scalar); 862 encode_sv (&enc, scalar);
863 encode_nl (&enc);
754 864
755 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 865 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
756 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 866 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
757 867
758 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) 868 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
819 if (*dec->cur != ch) \ 929 if (*dec->cur != ch) \
820 ERR (# ch " expected"); \ 930 ERR (# ch " expected"); \
821 ++dec->cur; \ 931 ++dec->cur; \
822 SE 932 SE
823 933
824#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") 934#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
825#define DEC_DEC_DEPTH --dec->depth 935#define DEC_DEC_DEPTH --dec->depth
826 936
827static SV *decode_sv (dec_t *dec); 937static SV *decode_sv (dec_t *dec);
828 938
829static signed char decode_hexdigit[256]; 939static signed char decode_hexdigit[256];
938 else if (expect_true (ch >= 0x20 && ch < 0x80)) 1048 else if (expect_true (ch >= 0x20 && ch < 0x80))
939 *cur++ = ch; 1049 *cur++ = ch;
940 else if (ch >= 0x80) 1050 else if (ch >= 0x80)
941 { 1051 {
942 STRLEN clen; 1052 STRLEN clen;
943 UV uch;
944 1053
945 --dec_cur; 1054 --dec_cur;
946 1055
947 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 1056 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
948 if (clen == (STRLEN)-1) 1057 if (clen == (STRLEN)-1)
949 ERR ("malformed UTF-8 character in JSON string"); 1058 ERR ("malformed UTF-8 character in JSON string");
950 1059
951 do 1060 do
952 *cur++ = *dec_cur++; 1061 *cur++ = *dec_cur++;
969 { 1078 {
970 STRLEN len = cur - buf; 1079 STRLEN len = cur - buf;
971 1080
972 if (sv) 1081 if (sv)
973 { 1082 {
974 SvGROW (sv, SvCUR (sv) + len + 1); 1083 STRLEN cur = SvCUR (sv);
1084
1085 if (SvLEN (sv) <= cur + len)
1086 SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
1087
975 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1088 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
976 SvCUR_set (sv, SvCUR (sv) + len); 1089 SvCUR_set (sv, SvCUR (sv) + len);
977 } 1090 }
978 else 1091 else
979 sv = newSVpvn (buf, len); 1092 sv = newSVpvn (buf, len);
1070 1183
1071 // special case the rather common 1..5-digit-int case 1184 // special case the rather common 1..5-digit-int case
1072 if (*start == '-') 1185 if (*start == '-')
1073 switch (len) 1186 switch (len)
1074 { 1187 {
1075 case 2: return newSViv (-( start [1] - '0' * 1)); 1188 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1076 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1189 case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1077 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); 1190 case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1078 case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); 1191 case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
1079 case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); 1192 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1080 } 1193 }
1081 else 1194 else
1082 switch (len) 1195 switch (len)
1083 { 1196 {
1084 case 1: return newSViv ( start [0] - '0' * 1); 1197 case 1: return newSViv ( start [0] - '0' * 1);
1085 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1198 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1086 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); 1199 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1087 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); 1200 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1088 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); 1201 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1089 } 1202 }
1090 1203
1091 { 1204 {
1092 UV uv; 1205 UV uv;
1093 int numtype = grok_number (start, len, &uv); 1206 int numtype = grok_number (start, len, &uv);
1102 } 1215 }
1103 1216
1104 len -= *start == '-' ? 1 : 0; 1217 len -= *start == '-' ? 1 : 0;
1105 1218
1106 // does not fit into IV or UV, try NV 1219 // does not fit into IV or UV, try NV
1107 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1220 if (len <= NV_DIG)
1108 #if defined (LDBL_DIG)
1109 || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1110 #endif
1111 )
1112 // fits into NV without loss of precision 1221 // fits into NV without loss of precision
1113 return newSVnv (Atof (start)); 1222 return newSVnv (json_atof (start));
1114 1223
1115 // everything else fails, convert it to a string 1224 // everything else fails, convert it to a string
1116 return newSVpvn (start, dec->cur - start); 1225 return newSVpvn (start, dec->cur - start);
1117 } 1226 }
1118 1227
1119 // loss of precision here 1228 // loss of precision here
1120 return newSVnv (Atof (start)); 1229 return newSVnv (json_atof (start));
1121 1230
1122fail: 1231fail:
1123 return 0; 1232 return 0;
1124} 1233}
1125 1234
1294 dSP; 1403 dSP;
1295 int count; 1404 int count;
1296 1405
1297 ENTER; SAVETMPS; PUSHMARK (SP); 1406 ENTER; SAVETMPS; PUSHMARK (SP);
1298 XPUSHs (HeVAL (he)); 1407 XPUSHs (HeVAL (he));
1408 sv_2mortal (sv);
1299 1409
1300 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1410 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1301 1411
1302 if (count == 1) 1412 if (count == 1)
1303 { 1413 {
1304 sv = newSVsv (POPs); 1414 sv = newSVsv (POPs);
1305 FREETMPS; LEAVE; 1415 FREETMPS; LEAVE;
1306 return sv; 1416 return sv;
1307 } 1417 }
1308 1418
1419 SvREFCNT_inc (sv);
1309 FREETMPS; LEAVE; 1420 FREETMPS; LEAVE;
1310 } 1421 }
1311 } 1422 }
1312 1423
1313 if (dec->json.cb_object) 1424 if (dec->json.cb_object)
1403fail: 1514fail:
1404 return 0; 1515 return 0;
1405} 1516}
1406 1517
1407static SV * 1518static SV *
1408decode_json (SV *string, JSON *json, STRLEN *offset_return) 1519decode_json (SV *string, JSON *json, char **offset_return)
1409{ 1520{
1410 dec_t dec; 1521 dec_t dec;
1411 STRLEN offset;
1412 SV *sv; 1522 SV *sv;
1413 1523
1524 /* work around bugs in 5.10 where manipulating magic values
1525 * will perl ignore the magic in subsequent accesses.
1526 * also make a copy of non-PV values, to get them into a clean
1527 * state (SvPV should do that, but it's buggy, see below).
1528 */
1414 SvGETMAGIC (string); 1529 /*SvGETMAGIC (string);*/
1530 if (SvMAGICAL (string) || !SvPOK (string))
1531 string = sv_2mortal (newSVsv (string));
1532
1415 SvUPGRADE (string, SVt_PV); 1533 SvUPGRADE (string, SVt_PV);
1416 1534
1417 if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) 1535 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1536 * assertion with -DDEBUGGING, although SvCUR is documented to
1537 * return the xpv_cur field which certainly exists after upgrading.
1538 * according to nicholas clark, calling SvPOK fixes this.
1539 * But it doesn't fix it, so try another workaround, call SvPV_nolen
1540 * and hope for the best.
1541 * Damnit, SvPV_nolen still trips over yet another assertion. This
1542 * assertion business is seriously broken, try yet another workaround
1543 * for the broken -DDEBUGGING.
1544 */
1545 {
1546#ifdef DEBUGGING
1547 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1548#else
1549 STRLEN offset = SvCUR (string);
1550#endif
1551
1552 if (offset > json->max_size && json->max_size)
1418 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 1553 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1419 (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags)); 1554 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1555 }
1420 1556
1421 if (json->flags & F_UTF8) 1557 if (DECODE_WANTS_OCTETS (json))
1422 sv_utf8_downgrade (string, 0); 1558 sv_utf8_downgrade (string, 0);
1423 else 1559 else
1424 sv_utf8_upgrade (string); 1560 sv_utf8_upgrade (string);
1425 1561
1426 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1562 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1427 1563
1428 dec.json = *json; 1564 dec.json = *json;
1429 dec.cur = SvPVX (string); 1565 dec.cur = SvPVX (string);
1430 dec.end = SvEND (string); 1566 dec.end = SvEND (string);
1431 dec.err = 0; 1567 dec.err = 0;
1432 dec.depth = 0; 1568 dec.depth = 0;
1433 dec.maxdepth = DEC_DEPTH (dec.json.flags);
1434 1569
1435 if (dec.json.cb_object || dec.json.cb_sk_object) 1570 if (dec.json.cb_object || dec.json.cb_sk_object)
1436 dec.json.flags |= F_HOOK; 1571 dec.json.flags |= F_HOOK;
1437 1572
1438 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1573 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1439 1574
1440 decode_ws (&dec); 1575 decode_ws (&dec);
1441 sv = decode_sv (&dec); 1576 sv = decode_sv (&dec);
1577
1578 if (offset_return)
1579 *offset_return = dec.cur;
1442 1580
1443 if (!(offset_return || !sv)) 1581 if (!(offset_return || !sv))
1444 { 1582 {
1445 // check for trailing garbage 1583 // check for trailing garbage
1446 decode_ws (&dec); 1584 decode_ws (&dec);
1449 { 1587 {
1450 dec.err = "garbage after JSON object"; 1588 dec.err = "garbage after JSON object";
1451 SvREFCNT_dec (sv); 1589 SvREFCNT_dec (sv);
1452 sv = 0; 1590 sv = 0;
1453 } 1591 }
1454 }
1455
1456 if (offset_return || !sv)
1457 {
1458 offset = dec.json.flags & F_UTF8
1459 ? dec.cur - SvPVX (string)
1460 : utf8_distance (dec.cur, SvPVX (string));
1461
1462 if (offset_return)
1463 *offset_return = offset;
1464 } 1592 }
1465 1593
1466 if (!sv) 1594 if (!sv)
1467 { 1595 {
1468 SV *uni = sv_newmortal (); 1596 SV *uni = sv_newmortal ();
1474 SAVEVPTR (PL_curcop); 1602 SAVEVPTR (PL_curcop);
1475 PL_curcop = &cop; 1603 PL_curcop = &cop;
1476 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1604 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1477 LEAVE; 1605 LEAVE;
1478 1606
1479 croak ("%s, at character offset %d [\"%s\"]", 1607 croak ("%s, at character offset %d (before \"%s\")",
1480 dec.err, 1608 dec.err,
1481 (int)offset, 1609 ptr_to_index (string, dec.cur),
1482 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1610 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1483 } 1611 }
1484 1612
1485 sv = sv_2mortal (sv); 1613 sv = sv_2mortal (sv);
1486 1614
1495 1623
1496static void 1624static void
1497incr_parse (JSON *self) 1625incr_parse (JSON *self)
1498{ 1626{
1499 const char *p = SvPVX (self->incr_text) + self->incr_pos; 1627 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1628
1629 // the state machine here is a bit convoluted and could be simplified a lot
1630 // but this would make it slower, so...
1500 1631
1501 for (;;) 1632 for (;;)
1502 { 1633 {
1503 //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 1634 //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
1504 switch (self->incr_mode) 1635 switch (self->incr_mode)
1505 { 1636 {
1506 // only used for intiial whitespace skipping 1637 // only used for initial whitespace skipping
1507 case INCR_M_WS: 1638 case INCR_M_WS:
1508 for (;;) 1639 for (;;)
1509 { 1640 {
1510 if (*p > 0x20) 1641 if (*p > 0x20)
1511 { 1642 {
1643 if (*p == '#')
1644 {
1645 self->incr_mode = INCR_M_C0;
1646 goto incr_m_c;
1647 }
1648 else
1649 {
1512 self->incr_mode = INCR_M_JSON; 1650 self->incr_mode = INCR_M_JSON;
1513 goto incr_m_json; 1651 goto incr_m_json;
1652 }
1514 } 1653 }
1515 else if (!*p) 1654 else if (!*p)
1516 goto interrupt; 1655 goto interrupt;
1517 1656
1518 ++p; 1657 ++p;
1524 goto interrupt; 1663 goto interrupt;
1525 1664
1526 ++p; 1665 ++p;
1527 self->incr_mode = INCR_M_STR; 1666 self->incr_mode = INCR_M_STR;
1528 goto incr_m_str; 1667 goto incr_m_str;
1668
1669 // inside #-style comments
1670 case INCR_M_C0:
1671 case INCR_M_C1:
1672 incr_m_c:
1673 for (;;)
1674 {
1675 if (*p == '\n')
1676 {
1677 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1678 break;
1679 }
1680 else if (!*p)
1681 goto interrupt;
1682
1683 ++p;
1684 }
1685
1686 break;
1529 1687
1530 // inside a string 1688 // inside a string
1531 case INCR_M_STR: 1689 case INCR_M_STR:
1532 incr_m_str: 1690 incr_m_str:
1533 for (;;) 1691 for (;;)
1584 self->incr_mode = INCR_M_STR; 1742 self->incr_mode = INCR_M_STR;
1585 goto incr_m_str; 1743 goto incr_m_str;
1586 1744
1587 case '[': 1745 case '[':
1588 case '{': 1746 case '{':
1589 ++self->incr_nest; 1747 if (++self->incr_nest > self->max_depth)
1748 croak (ERR_NESTING_EXCEEDED);
1590 break; 1749 break;
1591 1750
1592 case ']': 1751 case ']':
1593 case '}': 1752 case '}':
1594 if (!--self->incr_nest) 1753 if (--self->incr_nest <= 0)
1595 goto interrupt; 1754 goto interrupt;
1755 break;
1756
1757 case '#':
1758 self->incr_mode = INCR_M_C1;
1759 goto incr_m_c;
1596 } 1760 }
1597 } 1761 }
1598 } 1762 }
1599 1763
1600 modechange: 1764 modechange:
1601 ; 1765 ;
1602 } 1766 }
1603 1767
1604interrupt: 1768interrupt:
1605 self->incr_pos = p - SvPVX (self->incr_text); 1769 self->incr_pos = p - SvPVX (self->incr_text);
1770 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1606 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D 1771 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1607} 1772}
1608 1773
1609///////////////////////////////////////////////////////////////////////////// 1774/////////////////////////////////////////////////////////////////////////////
1610// XS interface functions 1775// XS interface functions
1625 json_stash = gv_stashpv ("JSON::XS" , 1); 1790 json_stash = gv_stashpv ("JSON::XS" , 1);
1626 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 1791 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1627 1792
1628 json_true = get_bool ("JSON::XS::true"); 1793 json_true = get_bool ("JSON::XS::true");
1629 json_false = get_bool ("JSON::XS::false"); 1794 json_false = get_bool ("JSON::XS::false");
1795
1796 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1630} 1797}
1631 1798
1632PROTOTYPES: DISABLE 1799PROTOTYPES: DISABLE
1633 1800
1634void CLONE (...) 1801void CLONE (...)
1639void new (char *klass) 1806void new (char *klass)
1640 PPCODE: 1807 PPCODE:
1641{ 1808{
1642 SV *pv = NEWSV (0, sizeof (JSON)); 1809 SV *pv = NEWSV (0, sizeof (JSON));
1643 SvPOK_only (pv); 1810 SvPOK_only (pv);
1644 Zero (SvPVX (pv), 1, JSON); 1811 json_init ((JSON *)SvPVX (pv));
1645 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1646 XPUSHs (sv_2mortal (sv_bless ( 1812 XPUSHs (sv_2mortal (sv_bless (
1647 newRV_noinc (pv), 1813 newRV_noinc (pv),
1648 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 1814 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1649 ))); 1815 )));
1650} 1816}
1691 get_relaxed = F_RELAXED 1857 get_relaxed = F_RELAXED
1692 get_allow_unknown = F_ALLOW_UNKNOWN 1858 get_allow_unknown = F_ALLOW_UNKNOWN
1693 PPCODE: 1859 PPCODE:
1694 XPUSHs (boolSV (self->flags & ix)); 1860 XPUSHs (boolSV (self->flags & ix));
1695 1861
1696void max_depth (JSON *self, UV max_depth = 0x80000000UL) 1862void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1697 PPCODE: 1863 PPCODE:
1698{ 1864 self->max_depth = max_depth;
1699 UV log2 = 0;
1700
1701 if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1702
1703 while ((1UL << log2) < max_depth)
1704 ++log2;
1705
1706 self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1707
1708 XPUSHs (ST (0)); 1865 XPUSHs (ST (0));
1709}
1710 1866
1711U32 get_max_depth (JSON *self) 1867U32 get_max_depth (JSON *self)
1712 CODE: 1868 CODE:
1713 RETVAL = DEC_DEPTH (self->flags); 1869 RETVAL = self->max_depth;
1714 OUTPUT: 1870 OUTPUT:
1715 RETVAL 1871 RETVAL
1716 1872
1717void max_size (JSON *self, UV max_size = 0) 1873void max_size (JSON *self, U32 max_size = 0)
1718 PPCODE: 1874 PPCODE:
1719{ 1875 self->max_size = max_size;
1720 UV log2 = 0;
1721
1722 if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1723 if (max_size == 1) max_size = 2;
1724
1725 while ((1UL << log2) < max_size)
1726 ++log2;
1727
1728 self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1729
1730 XPUSHs (ST (0)); 1876 XPUSHs (ST (0));
1731}
1732 1877
1733int get_max_size (JSON *self) 1878int get_max_size (JSON *self)
1734 CODE: 1879 CODE:
1735 RETVAL = DEC_SIZE (self->flags); 1880 RETVAL = self->max_size;
1736 OUTPUT: 1881 OUTPUT:
1737 RETVAL 1882 RETVAL
1738 1883
1739void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) 1884void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1740 PPCODE: 1885 PPCODE:
1776 XPUSHs (decode_json (jsonstr, self, 0)); 1921 XPUSHs (decode_json (jsonstr, self, 0));
1777 1922
1778void decode_prefix (JSON *self, SV *jsonstr) 1923void decode_prefix (JSON *self, SV *jsonstr)
1779 PPCODE: 1924 PPCODE:
1780{ 1925{
1781 STRLEN offset; 1926 char *offset;
1782 EXTEND (SP, 2); 1927 EXTEND (SP, 2);
1783 PUSHs (decode_json (jsonstr, self, &offset)); 1928 PUSHs (decode_json (jsonstr, self, &offset));
1784 PUSHs (sv_2mortal (newSVuv (offset))); 1929 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
1785} 1930}
1786 1931
1787void incr_parse (JSON *self, SV *jsonstr = 0) 1932void incr_parse (JSON *self, SV *jsonstr = 0)
1788 PPCODE: 1933 PPCODE:
1789{ 1934{
1790 if (!self->incr_text) 1935 if (!self->incr_text)
1791 self->incr_text = newSVpvn ("", 0); 1936 self->incr_text = newSVpvn ("", 0);
1937
1938 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
1939 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
1940 if (DECODE_WANTS_OCTETS (self))
1941 {
1942 if (self->incr_pos)
1943 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
1944 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
1945
1946 sv_utf8_downgrade (self->incr_text, 0);
1947 }
1948 else
1949 {
1950 sv_utf8_upgrade (self->incr_text);
1951
1952 if (self->incr_pos)
1953 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1954 - (U8 *)SvPVX (self->incr_text);
1955 }
1792 1956
1793 // append data, if any 1957 // append data, if any
1794 if (jsonstr) 1958 if (jsonstr)
1795 { 1959 {
1960 /* make sure both strings have same encoding */
1796 if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text)) 1961 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
1962 if (SvUTF8 (jsonstr))
1963 sv_utf8_downgrade (jsonstr, 0);
1797 { 1964 else
1798 /* utf-8-ness differs, need to upgrade */
1799 sv_utf8_upgrade (self->incr_text); 1965 sv_utf8_upgrade (jsonstr);
1800 1966
1801 if (self->incr_pos) 1967 /* and then just blindly append */
1802 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1803 - (U8 *)SvPVX (self->incr_text);
1804 }
1805
1806 { 1968 {
1807 STRLEN len; 1969 STRLEN len;
1808 const char *str = SvPV (jsonstr, len); 1970 const char *str = SvPV (jsonstr, len);
1809 SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1); 1971 STRLEN cur = SvCUR (self->incr_text);
1972
1973 if (SvLEN (self->incr_text) <= cur + len)
1974 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
1975
1810 Move (str, SvEND (self->incr_text), len, char); 1976 Move (str, SvEND (self->incr_text), len, char);
1811 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); 1977 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
1812 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there 1978 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1813 } 1979 }
1814 } 1980 }
1815 1981
1816 if (GIMME_V != G_VOID) 1982 if (GIMME_V != G_VOID)
1817 do 1983 do
1818 { 1984 {
1819 STRLEN offset; 1985 char *offset;
1820 1986
1821 if (!INCR_DONE (self)) 1987 if (!INCR_DONE (self))
1822 { 1988 {
1823 incr_parse (self); 1989 incr_parse (self);
1990
1991 if (self->incr_pos > self->max_size && self->max_size)
1992 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1993 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
1994
1824 if (!INCR_DONE (self)) 1995 if (!INCR_DONE (self))
1825 break; 1996 break;
1826 } 1997 }
1827 1998
1828 XPUSHs (decode_json (self->incr_text, self, &offset)); 1999 XPUSHs (decode_json (self->incr_text, self, &offset));
1829 2000
1830 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
1831 self->incr_pos -= offset; 2001 self->incr_pos -= offset - SvPVX (self->incr_text);
1832 self->incr_nest = 0; 2002 self->incr_nest = 0;
1833 self->incr_mode = 0; 2003 self->incr_mode = 0;
2004
2005 sv_chop (self->incr_text, offset);
1834 } 2006 }
1835 while (GIMME_V == G_ARRAY); 2007 while (GIMME_V == G_ARRAY);
1836} 2008}
1837 2009
1838SV *incr_text (JSON *self) 2010SV *incr_text (JSON *self)
1857 self->incr_nest = 0; 2029 self->incr_nest = 0;
1858 self->incr_mode = 0; 2030 self->incr_mode = 0;
1859 } 2031 }
1860} 2032}
1861 2033
2034void incr_reset (JSON *self)
2035 CODE:
2036{
2037 SvREFCNT_dec (self->incr_text);
2038 self->incr_text = 0;
2039 self->incr_pos = 0;
2040 self->incr_nest = 0;
2041 self->incr_mode = 0;
2042}
2043
1862void DESTROY (JSON *self) 2044void DESTROY (JSON *self)
1863 CODE: 2045 CODE:
1864 SvREFCNT_dec (self->cb_sk_object); 2046 SvREFCNT_dec (self->cb_sk_object);
1865 SvREFCNT_dec (self->cb_object); 2047 SvREFCNT_dec (self->cb_object);
1866 SvREFCNT_dec (self->incr_text); 2048 SvREFCNT_dec (self->incr_text);
1871 ALIAS: 2053 ALIAS:
1872 to_json_ = 0 2054 to_json_ = 0
1873 encode_json = F_UTF8 2055 encode_json = F_UTF8
1874 PPCODE: 2056 PPCODE:
1875{ 2057{
1876 JSON json = { F_DEFAULT | ix }; 2058 JSON json;
2059 json_init (&json);
2060 json.flags |= ix;
1877 XPUSHs (encode_json (scalar, &json)); 2061 XPUSHs (encode_json (scalar, &json));
1878} 2062}
1879 2063
1880void decode_json (SV *jsonstr) 2064void decode_json (SV *jsonstr)
1881 ALIAS: 2065 ALIAS:
1882 from_json_ = 0 2066 from_json_ = 0
1883 decode_json = F_UTF8 2067 decode_json = F_UTF8
1884 PPCODE: 2068 PPCODE:
1885{ 2069{
1886 JSON json = { F_DEFAULT | ix }; 2070 JSON json;
2071 json_init (&json);
2072 json.flags |= ix;
1887 XPUSHs (decode_json (jsonstr, &json, 0)); 2073 XPUSHs (decode_json (jsonstr, &json, 0));
1888} 2074}
1889 2075
1890

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines