… | |
… | |
42 | #define INIT_SIZE 32 // initial scalar size to be allocated |
42 | #define INIT_SIZE 32 // initial scalar size to be allocated |
43 | #define INDENT_STEP 3 // spaces per indentation level |
43 | #define INDENT_STEP 3 // spaces per indentation level |
44 | |
44 | |
45 | #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 | |
46 | |
|
|
47 | #define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8) |
|
|
48 | |
47 | #define SB do { |
49 | #define SB do { |
48 | #define SE } while (0) |
50 | #define SE } while (0) |
49 | |
51 | |
50 | #if __GNUC__ >= 3 |
52 | #if __GNUC__ >= 3 |
51 | # define expect(expr,value) __builtin_expect ((expr), (value)) |
53 | # define expect(expr,value) __builtin_expect ((expr), (value)) |
… | |
… | |
187 | return SvUTF8 (sv) |
189 | return SvUTF8 (sv) |
188 | ? utf8_distance (offset, SvPVX (sv)) |
190 | ? utf8_distance (offset, SvPVX (sv)) |
189 | : offset - SvPVX (sv); |
191 | : offset - SvPVX (sv); |
190 | } |
192 | } |
191 | |
193 | |
|
|
194 | ///////////////////////////////////////////////////////////////////////////// |
|
|
195 | // fp hell |
|
|
196 | |
|
|
197 | // scan a group of digits, and a trailing exponent |
|
|
198 | static void |
|
|
199 | json_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 | |
|
|
270 | static NV |
|
|
271 | json_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 | } |
192 | ///////////////////////////////////////////////////////////////////////////// |
288 | ///////////////////////////////////////////////////////////////////////////// |
193 | // encoder |
289 | // encoder |
194 | |
290 | |
195 | // structure used for encoding JSON |
291 | // structure used for encoding JSON |
196 | typedef struct |
292 | typedef struct |
… | |
… | |
1119 | } |
1215 | } |
1120 | |
1216 | |
1121 | len -= *start == '-' ? 1 : 0; |
1217 | len -= *start == '-' ? 1 : 0; |
1122 | |
1218 | |
1123 | // does not fit into IV or UV, try NV |
1219 | // does not fit into IV or UV, try NV |
1124 | if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) |
1220 | if (len <= NV_DIG) |
1125 | #if defined (LDBL_DIG) |
|
|
1126 | || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len) |
|
|
1127 | #endif |
|
|
1128 | ) |
|
|
1129 | // fits into NV without loss of precision |
1221 | // fits into NV without loss of precision |
1130 | return newSVnv (Atof (start)); |
1222 | return newSVnv (json_atof (start)); |
1131 | |
1223 | |
1132 | // everything else fails, convert it to a string |
1224 | // everything else fails, convert it to a string |
1133 | return newSVpvn (start, dec->cur - start); |
1225 | return newSVpvn (start, dec->cur - start); |
1134 | } |
1226 | } |
1135 | |
1227 | |
1136 | // loss of precision here |
1228 | // loss of precision here |
1137 | return newSVnv (Atof (start)); |
1229 | return newSVnv (json_atof (start)); |
1138 | |
1230 | |
1139 | fail: |
1231 | fail: |
1140 | return 0; |
1232 | return 0; |
1141 | } |
1233 | } |
1142 | |
1234 | |
… | |
… | |
1311 | dSP; |
1403 | dSP; |
1312 | int count; |
1404 | int count; |
1313 | |
1405 | |
1314 | ENTER; SAVETMPS; PUSHMARK (SP); |
1406 | ENTER; SAVETMPS; PUSHMARK (SP); |
1315 | XPUSHs (HeVAL (he)); |
1407 | XPUSHs (HeVAL (he)); |
|
|
1408 | sv_2mortal (sv); |
1316 | |
1409 | |
1317 | PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
1410 | PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
1318 | |
1411 | |
1319 | if (count == 1) |
1412 | if (count == 1) |
1320 | { |
1413 | { |
1321 | sv = newSVsv (POPs); |
1414 | sv = newSVsv (POPs); |
1322 | FREETMPS; LEAVE; |
1415 | FREETMPS; LEAVE; |
1323 | return sv; |
1416 | return sv; |
1324 | } |
1417 | } |
1325 | |
1418 | |
|
|
1419 | SvREFCNT_inc (sv); |
1326 | FREETMPS; LEAVE; |
1420 | FREETMPS; LEAVE; |
1327 | } |
1421 | } |
1328 | } |
1422 | } |
1329 | |
1423 | |
1330 | if (dec->json.cb_object) |
1424 | if (dec->json.cb_object) |
… | |
… | |
1456 | if (offset > json->max_size && json->max_size) |
1550 | if (offset > json->max_size && json->max_size) |
1457 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
1551 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
1458 | (unsigned long)SvCUR (string), (unsigned long)json->max_size); |
1552 | (unsigned long)SvCUR (string), (unsigned long)json->max_size); |
1459 | } |
1553 | } |
1460 | |
1554 | |
1461 | if (json->flags & F_UTF8) |
1555 | if (DECODE_WANTS_OCTETS (json)) |
1462 | sv_utf8_downgrade (string, 0); |
1556 | sv_utf8_downgrade (string, 0); |
1463 | else |
1557 | else |
1464 | sv_utf8_upgrade (string); |
1558 | sv_utf8_upgrade (string); |
1465 | |
1559 | |
1466 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1560 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
… | |
… | |
1837 | PPCODE: |
1931 | PPCODE: |
1838 | { |
1932 | { |
1839 | if (!self->incr_text) |
1933 | if (!self->incr_text) |
1840 | self->incr_text = newSVpvn ("", 0); |
1934 | self->incr_text = newSVpvn ("", 0); |
1841 | |
1935 | |
|
|
1936 | /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */ |
|
|
1937 | if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text)) |
|
|
1938 | if (DECODE_WANTS_OCTETS (self)) |
|
|
1939 | { |
|
|
1940 | if (self->incr_pos) |
|
|
1941 | self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text), |
|
|
1942 | (U8 *)SvPVX (self->incr_text) + self->incr_pos); |
|
|
1943 | |
|
|
1944 | sv_utf8_downgrade (self->incr_text, 0); |
|
|
1945 | } |
|
|
1946 | else |
|
|
1947 | { |
|
|
1948 | sv_utf8_upgrade (self->incr_text); |
|
|
1949 | |
|
|
1950 | if (self->incr_pos) |
|
|
1951 | self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos) |
|
|
1952 | - (U8 *)SvPVX (self->incr_text); |
|
|
1953 | } |
|
|
1954 | |
1842 | // append data, if any |
1955 | // append data, if any |
1843 | if (jsonstr) |
1956 | if (jsonstr) |
1844 | { |
1957 | { |
|
|
1958 | /* make sure both strings have same encoding */ |
|
|
1959 | if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text)) |
1845 | if (SvUTF8 (jsonstr)) |
1960 | if (SvUTF8 (jsonstr)) |
|
|
1961 | sv_utf8_downgrade (jsonstr, 0); |
1846 | { |
1962 | else |
1847 | if (!SvUTF8 (self->incr_text)) |
|
|
1848 | { |
|
|
1849 | /* utf-8-ness differs, need to upgrade */ |
|
|
1850 | sv_utf8_upgrade (self->incr_text); |
|
|
1851 | |
|
|
1852 | if (self->incr_pos) |
|
|
1853 | self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos) |
|
|
1854 | - (U8 *)SvPVX (self->incr_text); |
|
|
1855 | } |
|
|
1856 | } |
|
|
1857 | else if (SvUTF8 (self->incr_text)) |
|
|
1858 | sv_utf8_upgrade (jsonstr); |
1963 | sv_utf8_upgrade (jsonstr); |
1859 | |
1964 | |
|
|
1965 | /* and then just blindly append */ |
1860 | { |
1966 | { |
1861 | STRLEN len; |
1967 | STRLEN len; |
1862 | const char *str = SvPV (jsonstr, len); |
1968 | const char *str = SvPV (jsonstr, len); |
1863 | STRLEN cur = SvCUR (self->incr_text); |
1969 | STRLEN cur = SvCUR (self->incr_text); |
1864 | |
1970 | |
… | |
… | |
1963 | json_init (&json); |
2069 | json_init (&json); |
1964 | json.flags |= ix; |
2070 | json.flags |= ix; |
1965 | XPUSHs (decode_json (jsonstr, &json, 0)); |
2071 | XPUSHs (decode_json (jsonstr, &json, 0)); |
1966 | } |
2072 | } |
1967 | |
2073 | |
1968 | |
|
|