… | |
… | |
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)) |
… | |
… | |
192 | ///////////////////////////////////////////////////////////////////////////// |
194 | ///////////////////////////////////////////////////////////////////////////// |
193 | // fp hell |
195 | // fp hell |
194 | |
196 | |
195 | // scan a group of digits, and a trailing exponent |
197 | // scan a group of digits, and a trailing exponent |
196 | static void |
198 | static void |
197 | json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp) |
199 | json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth) |
198 | { |
200 | { |
199 | UV uaccum = 0; |
201 | UV uaccum = 0; |
200 | int eaccum = 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; |
201 | |
209 | |
202 | for (;;) |
210 | for (;;) |
203 | { |
211 | { |
204 | U8 dig = (U8)*s - '0'; |
212 | U8 dig = (U8)*s - '0'; |
205 | |
213 | |
206 | if (expect_false (dig >= 10)) |
214 | if (expect_false (dig >= 10)) |
207 | { |
215 | { |
208 | if (dig == (U8)((U8)'.' - (U8)'0')) |
216 | if (dig == (U8)((U8)'.' - (U8)'0')) |
209 | { |
217 | { |
210 | ++s; |
218 | ++s; |
211 | json_atof_scan1 (s, accum, expo, 1); |
219 | json_atof_scan1 (s, accum, expo, 1, maxdepth); |
212 | } |
220 | } |
213 | else if ((dig | ' ') == 'e' - '0') |
221 | else if ((dig | ' ') == 'e' - '0') |
214 | { |
222 | { |
215 | int exp2 = 0; |
223 | int exp2 = 0; |
216 | int neg = 0; |
224 | int neg = 0; |
… | |
… | |
242 | // if we have too many digits, then recurse for more |
250 | // if we have too many digits, then recurse for more |
243 | // we actually do this for rather few digits |
251 | // we actually do this for rather few digits |
244 | if (uaccum >= (UV_MAX - 9) / 10) |
252 | if (uaccum >= (UV_MAX - 9) / 10) |
245 | { |
253 | { |
246 | if (postdp) *expo -= eaccum; |
254 | if (postdp) *expo -= eaccum; |
247 | json_atof_scan1 (s, accum, expo, postdp); |
255 | json_atof_scan1 (s, accum, expo, postdp, maxdepth); |
248 | if (postdp) *expo += eaccum; |
256 | if (postdp) *expo += eaccum; |
249 | |
257 | |
250 | break; |
258 | break; |
251 | } |
259 | } |
252 | } |
260 | } |
253 | |
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 | // (IEEE 754 conformant ones are required to be exact) |
254 | if (postdp) *expo -= eaccum; |
266 | if (postdp) *expo -= eaccum; |
255 | *accum += uaccum * pow (10., *expo); |
267 | *accum += uaccum * Perl_pow (10., *expo); |
256 | *expo += eaccum; |
268 | *expo += eaccum; |
257 | } |
269 | } |
258 | |
270 | |
259 | static NV |
271 | static NV |
260 | json_atof (const char *s) |
272 | json_atof (const char *s) |
… | |
… | |
267 | { |
279 | { |
268 | ++s; |
280 | ++s; |
269 | neg = 1; |
281 | neg = 1; |
270 | } |
282 | } |
271 | |
283 | |
|
|
284 | // a recursion depth of ten gives us >>500 bits |
272 | json_atof_scan1 (s, &accum, &expo, 0); |
285 | json_atof_scan1 (s, &accum, &expo, 0, 10); |
273 | |
286 | |
274 | return neg ? -accum : accum; |
287 | return neg ? -accum : accum; |
275 | } |
288 | } |
276 | ///////////////////////////////////////////////////////////////////////////// |
289 | ///////////////////////////////////////////////////////////////////////////// |
277 | // encoder |
290 | // encoder |
… | |
… | |
824 | encode_rv (enc, SvRV (sv)); |
837 | encode_rv (enc, SvRV (sv)); |
825 | else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) |
838 | else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) |
826 | encode_str (enc, "null", 4, 0); |
839 | encode_str (enc, "null", 4, 0); |
827 | else |
840 | else |
828 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
841 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
829 | SvPV_nolen (sv), SvFLAGS (sv)); |
842 | SvPV_nolen (sv), (unsigned int)SvFLAGS (sv)); |
830 | } |
843 | } |
831 | |
844 | |
832 | static SV * |
845 | static SV * |
833 | encode_json (SV *scalar, JSON *json) |
846 | encode_json (SV *scalar, JSON *json) |
834 | { |
847 | { |
… | |
… | |
1391 | dSP; |
1404 | dSP; |
1392 | int count; |
1405 | int count; |
1393 | |
1406 | |
1394 | ENTER; SAVETMPS; PUSHMARK (SP); |
1407 | ENTER; SAVETMPS; PUSHMARK (SP); |
1395 | XPUSHs (HeVAL (he)); |
1408 | XPUSHs (HeVAL (he)); |
|
|
1409 | sv_2mortal (sv); |
1396 | |
1410 | |
1397 | PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
1411 | PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
1398 | |
1412 | |
1399 | if (count == 1) |
1413 | if (count == 1) |
1400 | { |
1414 | { |
1401 | sv = newSVsv (POPs); |
1415 | sv = newSVsv (POPs); |
1402 | FREETMPS; LEAVE; |
1416 | FREETMPS; LEAVE; |
1403 | return sv; |
1417 | return sv; |
1404 | } |
1418 | } |
1405 | |
1419 | |
|
|
1420 | SvREFCNT_inc (sv); |
1406 | FREETMPS; LEAVE; |
1421 | FREETMPS; LEAVE; |
1407 | } |
1422 | } |
1408 | } |
1423 | } |
1409 | |
1424 | |
1410 | if (dec->json.cb_object) |
1425 | if (dec->json.cb_object) |
… | |
… | |
1506 | { |
1521 | { |
1507 | dec_t dec; |
1522 | dec_t dec; |
1508 | SV *sv; |
1523 | SV *sv; |
1509 | |
1524 | |
1510 | /* work around bugs in 5.10 where manipulating magic values |
1525 | /* work around bugs in 5.10 where manipulating magic values |
1511 | * will perl ignore the magic in subsequent accesses |
1526 | * will perl ignore the magic in subsequent accesses. |
|
|
1527 | * also make a copy of non-PV values, to get them into a clean |
|
|
1528 | * state (SvPV should do that, but it's buggy, see below). |
1512 | */ |
1529 | */ |
1513 | /*SvGETMAGIC (string);*/ |
1530 | /*SvGETMAGIC (string);*/ |
1514 | if (SvMAGICAL (string)) |
1531 | if (SvMAGICAL (string) || !SvPOK (string)) |
1515 | string = sv_2mortal (newSVsv (string)); |
1532 | string = sv_2mortal (newSVsv (string)); |
1516 | |
1533 | |
1517 | SvUPGRADE (string, SVt_PV); |
1534 | SvUPGRADE (string, SVt_PV); |
1518 | |
1535 | |
1519 | /* work around a bug in perl 5.10, which causes SvCUR to fail an |
1536 | /* work around a bug in perl 5.10, which causes SvCUR to fail an |
… | |
… | |
1536 | if (offset > json->max_size && json->max_size) |
1553 | if (offset > json->max_size && json->max_size) |
1537 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
1554 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
1538 | (unsigned long)SvCUR (string), (unsigned long)json->max_size); |
1555 | (unsigned long)SvCUR (string), (unsigned long)json->max_size); |
1539 | } |
1556 | } |
1540 | |
1557 | |
1541 | if (json->flags & F_UTF8) |
1558 | if (DECODE_WANTS_OCTETS (json)) |
1542 | sv_utf8_downgrade (string, 0); |
1559 | sv_utf8_downgrade (string, 0); |
1543 | else |
1560 | else |
1544 | sv_utf8_upgrade (string); |
1561 | sv_utf8_upgrade (string); |
1545 | |
1562 | |
1546 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1563 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
… | |
… | |
1588 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1605 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1589 | LEAVE; |
1606 | LEAVE; |
1590 | |
1607 | |
1591 | croak ("%s, at character offset %d (before \"%s\")", |
1608 | croak ("%s, at character offset %d (before \"%s\")", |
1592 | dec.err, |
1609 | dec.err, |
1593 | ptr_to_index (string, dec.cur), |
1610 | (int)ptr_to_index (string, dec.cur), |
1594 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1611 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1595 | } |
1612 | } |
1596 | |
1613 | |
1597 | sv = sv_2mortal (sv); |
1614 | sv = sv_2mortal (sv); |
1598 | |
1615 | |
… | |
… | |
1917 | PPCODE: |
1934 | PPCODE: |
1918 | { |
1935 | { |
1919 | if (!self->incr_text) |
1936 | if (!self->incr_text) |
1920 | self->incr_text = newSVpvn ("", 0); |
1937 | self->incr_text = newSVpvn ("", 0); |
1921 | |
1938 | |
|
|
1939 | /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */ |
|
|
1940 | if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text)) |
|
|
1941 | if (DECODE_WANTS_OCTETS (self)) |
|
|
1942 | { |
|
|
1943 | if (self->incr_pos) |
|
|
1944 | self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text), |
|
|
1945 | (U8 *)SvPVX (self->incr_text) + self->incr_pos); |
|
|
1946 | |
|
|
1947 | sv_utf8_downgrade (self->incr_text, 0); |
|
|
1948 | } |
|
|
1949 | else |
|
|
1950 | { |
|
|
1951 | sv_utf8_upgrade (self->incr_text); |
|
|
1952 | |
|
|
1953 | if (self->incr_pos) |
|
|
1954 | self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos) |
|
|
1955 | - (U8 *)SvPVX (self->incr_text); |
|
|
1956 | } |
|
|
1957 | |
1922 | // append data, if any |
1958 | // append data, if any |
1923 | if (jsonstr) |
1959 | if (jsonstr) |
1924 | { |
1960 | { |
|
|
1961 | /* make sure both strings have same encoding */ |
|
|
1962 | if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text)) |
1925 | if (SvUTF8 (jsonstr)) |
1963 | if (SvUTF8 (jsonstr)) |
|
|
1964 | sv_utf8_downgrade (jsonstr, 0); |
1926 | { |
1965 | else |
1927 | if (!SvUTF8 (self->incr_text)) |
|
|
1928 | { |
|
|
1929 | /* utf-8-ness differs, need to upgrade */ |
|
|
1930 | sv_utf8_upgrade (self->incr_text); |
|
|
1931 | |
|
|
1932 | if (self->incr_pos) |
|
|
1933 | self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos) |
|
|
1934 | - (U8 *)SvPVX (self->incr_text); |
|
|
1935 | } |
|
|
1936 | } |
|
|
1937 | else if (SvUTF8 (self->incr_text)) |
|
|
1938 | sv_utf8_upgrade (jsonstr); |
1966 | sv_utf8_upgrade (jsonstr); |
1939 | |
1967 | |
|
|
1968 | /* and then just blindly append */ |
1940 | { |
1969 | { |
1941 | STRLEN len; |
1970 | STRLEN len; |
1942 | const char *str = SvPV (jsonstr, len); |
1971 | const char *str = SvPV (jsonstr, len); |
1943 | STRLEN cur = SvCUR (self->incr_text); |
1972 | STRLEN cur = SvCUR (self->incr_text); |
1944 | |
1973 | |
… | |
… | |
1963 | if (self->incr_pos > self->max_size && self->max_size) |
1992 | if (self->incr_pos > self->max_size && self->max_size) |
1964 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
1993 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
1965 | (unsigned long)self->incr_pos, (unsigned long)self->max_size); |
1994 | (unsigned long)self->incr_pos, (unsigned long)self->max_size); |
1966 | |
1995 | |
1967 | if (!INCR_DONE (self)) |
1996 | if (!INCR_DONE (self)) |
|
|
1997 | { |
|
|
1998 | // as an optimisation, do not accumulate white space in the incr buffer |
|
|
1999 | if (self->incr_mode == INCR_M_WS && self->incr_pos) |
|
|
2000 | { |
|
|
2001 | self->incr_pos = 0; |
|
|
2002 | SvCUR_set (self->incr_text, 0); |
|
|
2003 | } |
|
|
2004 | |
1968 | break; |
2005 | break; |
|
|
2006 | } |
1969 | } |
2007 | } |
1970 | |
2008 | |
1971 | XPUSHs (decode_json (self->incr_text, self, &offset)); |
2009 | XPUSHs (decode_json (self->incr_text, self, &offset)); |
1972 | |
2010 | |
1973 | self->incr_pos -= offset - SvPVX (self->incr_text); |
2011 | self->incr_pos -= offset - SvPVX (self->incr_text); |