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 1.95 by root, Sun Feb 22 06:55:28 2009 UTC vs.
Revision 1.103 by root, Wed Jan 6 08:01:39 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
76 77
77enum { 78enum {
78 INCR_M_WS = 0, // initial whitespace skipping, must be 0 79 INCR_M_WS = 0, // initial whitespace skipping, must be 0
79 INCR_M_STR, // inside string 80 INCR_M_STR, // inside string
80 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
81 INCR_M_JSON // outside anything, count nesting 84 INCR_M_JSON // outside anything, count nesting
82}; 85};
83 86
84#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON) 87#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
85 88
173 *s++ = 0x80 | ((ch >> 12) & 0x3f), 176 *s++ = 0x80 | ((ch >> 12) & 0x3f),
174 *s++ = 0x80 | ((ch >> 6) & 0x3f), 177 *s++ = 0x80 | ((ch >> 6) & 0x3f),
175 *s++ = 0x80 | ( ch & 0x3f); 178 *s++ = 0x80 | ( ch & 0x3f);
176 179
177 return s; 180 return s;
181}
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);
178} 190}
179 191
180///////////////////////////////////////////////////////////////////////////// 192/////////////////////////////////////////////////////////////////////////////
181// encoder 193// encoder
182 194
462 474
463 // for canonical output we have to sort by keys first 475 // for canonical output we have to sort by keys first
464 // actually, this is mostly due to the stupid so-called 476 // actually, this is mostly due to the stupid so-called
465 // security workaround added somewhere in 5.8.x 477 // security workaround added somewhere in 5.8.x
466 // that randomises hash orderings 478 // that randomises hash orderings
467 if (enc->json.flags & F_CANONICAL) 479 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
468 { 480 {
469 int count = hv_iterinit (hv); 481 int count = hv_iterinit (hv);
470 482
471 if (SvMAGICAL (hv)) 483 if (SvMAGICAL (hv))
472 { 484 {
750 : enc.json.flags & F_LATIN1 ? 0x000100UL 762 : enc.json.flags & F_LATIN1 ? 0x000100UL
751 : 0x110000UL; 763 : 0x110000UL;
752 764
753 SvPOK_only (enc.sv); 765 SvPOK_only (enc.sv);
754 encode_sv (&enc, scalar); 766 encode_sv (&enc, scalar);
767 encode_nl (&enc);
755 768
756 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 769 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
757 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 770 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
758 771
759 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) 772 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
939 else if (expect_true (ch >= 0x20 && ch < 0x80)) 952 else if (expect_true (ch >= 0x20 && ch < 0x80))
940 *cur++ = ch; 953 *cur++ = ch;
941 else if (ch >= 0x80) 954 else if (ch >= 0x80)
942 { 955 {
943 STRLEN clen; 956 STRLEN clen;
944 UV uch;
945 957
946 --dec_cur; 958 --dec_cur;
947 959
948 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 960 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
949 if (clen == (STRLEN)-1) 961 if (clen == (STRLEN)-1)
950 ERR ("malformed UTF-8 character in JSON string"); 962 ERR ("malformed UTF-8 character in JSON string");
951 963
952 do 964 do
953 *cur++ = *dec_cur++; 965 *cur++ = *dec_cur++;
1075 1087
1076 // special case the rather common 1..5-digit-int case 1088 // special case the rather common 1..5-digit-int case
1077 if (*start == '-') 1089 if (*start == '-')
1078 switch (len) 1090 switch (len)
1079 { 1091 {
1080 case 2: return newSViv (-( start [1] - '0' * 1)); 1092 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1081 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1093 case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1082 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); 1094 case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1083 case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); 1095 case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
1084 case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); 1096 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1085 } 1097 }
1086 else 1098 else
1087 switch (len) 1099 switch (len)
1088 { 1100 {
1089 case 1: return newSViv ( start [0] - '0' * 1); 1101 case 1: return newSViv ( start [0] - '0' * 1);
1090 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1102 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1091 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); 1103 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1092 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); 1104 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1093 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); 1105 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1094 } 1106 }
1095 1107
1096 { 1108 {
1097 UV uv; 1109 UV uv;
1098 int numtype = grok_number (start, len, &uv); 1110 int numtype = grok_number (start, len, &uv);
1408fail: 1420fail:
1409 return 0; 1421 return 0;
1410} 1422}
1411 1423
1412static SV * 1424static SV *
1413decode_json (SV *string, JSON *json, STRLEN *offset_return) 1425decode_json (SV *string, JSON *json, char **offset_return)
1414{ 1426{
1415 dec_t dec; 1427 dec_t dec;
1416 STRLEN offset;
1417 SV *sv; 1428 SV *sv;
1418 1429
1419 /* work around bugs in 5.10 where manipulating magic values 1430 /* work around bugs in 5.10 where manipulating magic values
1420 * will perl ignore the magic in subsequent accesses 1431 * will perl ignore the magic in subsequent accesses
1421 */ 1432 */
1433 * and hope for the best. 1444 * and hope for the best.
1434 * Damnit, SvPV_nolen still trips over yet another assertion. This 1445 * Damnit, SvPV_nolen still trips over yet another assertion. This
1435 * assertion business is seriously broken, try yet another workaround 1446 * assertion business is seriously broken, try yet another workaround
1436 * for the broken -DDEBUGGING. 1447 * for the broken -DDEBUGGING.
1437 */ 1448 */
1449 {
1438#ifdef DEBUGGING 1450#ifdef DEBUGGING
1439 offset = SvOK (string) ? sv_len (string) : 0; 1451 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1440#else 1452#else
1441 offset = SvCUR (string); 1453 STRLEN offset = SvCUR (string);
1442#endif 1454#endif
1443 1455
1444 if (offset > json->max_size && json->max_size) 1456 if (offset > json->max_size && json->max_size)
1445 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 1457 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1446 (unsigned long)SvCUR (string), (unsigned long)json->max_size); 1458 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1459 }
1447 1460
1448 if (json->flags & F_UTF8) 1461 if (json->flags & F_UTF8)
1449 sv_utf8_downgrade (string, 0); 1462 sv_utf8_downgrade (string, 0);
1450 else 1463 else
1451 sv_utf8_upgrade (string); 1464 sv_utf8_upgrade (string);
1463 1476
1464 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1477 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1465 1478
1466 decode_ws (&dec); 1479 decode_ws (&dec);
1467 sv = decode_sv (&dec); 1480 sv = decode_sv (&dec);
1481
1482 if (offset_return)
1483 *offset_return = dec.cur;
1468 1484
1469 if (!(offset_return || !sv)) 1485 if (!(offset_return || !sv))
1470 { 1486 {
1471 // check for trailing garbage 1487 // check for trailing garbage
1472 decode_ws (&dec); 1488 decode_ws (&dec);
1475 { 1491 {
1476 dec.err = "garbage after JSON object"; 1492 dec.err = "garbage after JSON object";
1477 SvREFCNT_dec (sv); 1493 SvREFCNT_dec (sv);
1478 sv = 0; 1494 sv = 0;
1479 } 1495 }
1480 }
1481
1482 if (offset_return || !sv)
1483 {
1484 offset = dec.json.flags & F_UTF8
1485 ? dec.cur - SvPVX (string)
1486 : utf8_distance (dec.cur, SvPVX (string));
1487
1488 if (offset_return)
1489 *offset_return = offset;
1490 } 1496 }
1491 1497
1492 if (!sv) 1498 if (!sv)
1493 { 1499 {
1494 SV *uni = sv_newmortal (); 1500 SV *uni = sv_newmortal ();
1500 SAVEVPTR (PL_curcop); 1506 SAVEVPTR (PL_curcop);
1501 PL_curcop = &cop; 1507 PL_curcop = &cop;
1502 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1508 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1503 LEAVE; 1509 LEAVE;
1504 1510
1505 croak ("%s, at character offset %d [\"%s\"]", 1511 croak ("%s, at character offset %d (before \"%s\")",
1506 dec.err, 1512 dec.err,
1507 (int)offset, 1513 ptr_to_index (string, dec.cur),
1508 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1514 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1509 } 1515 }
1510 1516
1511 sv = sv_2mortal (sv); 1517 sv = sv_2mortal (sv);
1512 1518
1521 1527
1522static void 1528static void
1523incr_parse (JSON *self) 1529incr_parse (JSON *self)
1524{ 1530{
1525 const char *p = SvPVX (self->incr_text) + self->incr_pos; 1531 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1532
1533 // the state machine here is a bit convoluted and could be simplified a lot
1534 // but this would make it slower, so...
1526 1535
1527 for (;;) 1536 for (;;)
1528 { 1537 {
1529 //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 1538 //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
1530 switch (self->incr_mode) 1539 switch (self->incr_mode)
1531 { 1540 {
1532 // only used for intiial whitespace skipping 1541 // only used for initial whitespace skipping
1533 case INCR_M_WS: 1542 case INCR_M_WS:
1534 for (;;) 1543 for (;;)
1535 { 1544 {
1536 if (*p > 0x20) 1545 if (*p > 0x20)
1537 { 1546 {
1547 if (*p == '#')
1548 {
1549 self->incr_mode = INCR_M_C0;
1550 goto incr_m_c;
1551 }
1552 else
1553 {
1538 self->incr_mode = INCR_M_JSON; 1554 self->incr_mode = INCR_M_JSON;
1539 goto incr_m_json; 1555 goto incr_m_json;
1556 }
1540 } 1557 }
1541 else if (!*p) 1558 else if (!*p)
1542 goto interrupt; 1559 goto interrupt;
1543 1560
1544 ++p; 1561 ++p;
1550 goto interrupt; 1567 goto interrupt;
1551 1568
1552 ++p; 1569 ++p;
1553 self->incr_mode = INCR_M_STR; 1570 self->incr_mode = INCR_M_STR;
1554 goto incr_m_str; 1571 goto incr_m_str;
1572
1573 // inside #-style comments
1574 case INCR_M_C0:
1575 case INCR_M_C1:
1576 incr_m_c:
1577 for (;;)
1578 {
1579 if (*p == '\n')
1580 {
1581 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1582 break;
1583 }
1584 else if (!*p)
1585 goto interrupt;
1586
1587 ++p;
1588 }
1589
1590 break;
1555 1591
1556 // inside a string 1592 // inside a string
1557 case INCR_M_STR: 1593 case INCR_M_STR:
1558 incr_m_str: 1594 incr_m_str:
1559 for (;;) 1595 for (;;)
1618 1654
1619 case ']': 1655 case ']':
1620 case '}': 1656 case '}':
1621 if (--self->incr_nest <= 0) 1657 if (--self->incr_nest <= 0)
1622 goto interrupt; 1658 goto interrupt;
1659 break;
1660
1661 case '#':
1662 self->incr_mode = INCR_M_C1;
1663 goto incr_m_c;
1623 } 1664 }
1624 } 1665 }
1625 } 1666 }
1626 1667
1627 modechange: 1668 modechange:
1628 ; 1669 ;
1629 } 1670 }
1630 1671
1631interrupt: 1672interrupt:
1632 self->incr_pos = p - SvPVX (self->incr_text); 1673 self->incr_pos = p - SvPVX (self->incr_text);
1674 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1633 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D 1675 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1634} 1676}
1635 1677
1636///////////////////////////////////////////////////////////////////////////// 1678/////////////////////////////////////////////////////////////////////////////
1637// XS interface functions 1679// XS interface functions
1652 json_stash = gv_stashpv ("JSON::XS" , 1); 1694 json_stash = gv_stashpv ("JSON::XS" , 1);
1653 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 1695 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1654 1696
1655 json_true = get_bool ("JSON::XS::true"); 1697 json_true = get_bool ("JSON::XS::true");
1656 json_false = get_bool ("JSON::XS::false"); 1698 json_false = get_bool ("JSON::XS::false");
1699
1700 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1657} 1701}
1658 1702
1659PROTOTYPES: DISABLE 1703PROTOTYPES: DISABLE
1660 1704
1661void CLONE (...) 1705void CLONE (...)
1781 XPUSHs (decode_json (jsonstr, self, 0)); 1825 XPUSHs (decode_json (jsonstr, self, 0));
1782 1826
1783void decode_prefix (JSON *self, SV *jsonstr) 1827void decode_prefix (JSON *self, SV *jsonstr)
1784 PPCODE: 1828 PPCODE:
1785{ 1829{
1786 STRLEN offset; 1830 char *offset;
1787 EXTEND (SP, 2); 1831 EXTEND (SP, 2);
1788 PUSHs (decode_json (jsonstr, self, &offset)); 1832 PUSHs (decode_json (jsonstr, self, &offset));
1789 PUSHs (sv_2mortal (newSVuv (offset))); 1833 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
1790} 1834}
1791 1835
1792void incr_parse (JSON *self, SV *jsonstr = 0) 1836void incr_parse (JSON *self, SV *jsonstr = 0)
1793 PPCODE: 1837 PPCODE:
1794{ 1838{
1796 self->incr_text = newSVpvn ("", 0); 1840 self->incr_text = newSVpvn ("", 0);
1797 1841
1798 // append data, if any 1842 // append data, if any
1799 if (jsonstr) 1843 if (jsonstr)
1800 { 1844 {
1801 if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text)) 1845 if (SvUTF8 (jsonstr))
1802 { 1846 {
1847 if (!SvUTF8 (self->incr_text))
1848 {
1803 /* utf-8-ness differs, need to upgrade */ 1849 /* utf-8-ness differs, need to upgrade */
1804 sv_utf8_upgrade (self->incr_text); 1850 sv_utf8_upgrade (self->incr_text);
1805 1851
1806 if (self->incr_pos) 1852 if (self->incr_pos)
1807 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos) 1853 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1808 - (U8 *)SvPVX (self->incr_text); 1854 - (U8 *)SvPVX (self->incr_text);
1855 }
1809 } 1856 }
1857 else if (SvUTF8 (self->incr_text))
1858 sv_utf8_upgrade (jsonstr);
1810 1859
1811 { 1860 {
1812 STRLEN len; 1861 STRLEN len;
1813 const char *str = SvPV (jsonstr, len); 1862 const char *str = SvPV (jsonstr, len);
1814 STRLEN cur = SvCUR (self->incr_text); 1863 STRLEN cur = SvCUR (self->incr_text);
1823 } 1872 }
1824 1873
1825 if (GIMME_V != G_VOID) 1874 if (GIMME_V != G_VOID)
1826 do 1875 do
1827 { 1876 {
1828 STRLEN offset; 1877 char *offset;
1829 1878
1830 if (!INCR_DONE (self)) 1879 if (!INCR_DONE (self))
1831 { 1880 {
1832 incr_parse (self); 1881 incr_parse (self);
1833 1882
1839 break; 1888 break;
1840 } 1889 }
1841 1890
1842 XPUSHs (decode_json (self->incr_text, self, &offset)); 1891 XPUSHs (decode_json (self->incr_text, self, &offset));
1843 1892
1844 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
1845 self->incr_pos -= offset; 1893 self->incr_pos -= offset - SvPVX (self->incr_text);
1846 self->incr_nest = 0; 1894 self->incr_nest = 0;
1847 self->incr_mode = 0; 1895 self->incr_mode = 0;
1896
1897 sv_chop (self->incr_text, offset);
1848 } 1898 }
1849 while (GIMME_V == G_ARRAY); 1899 while (GIMME_V == G_ARRAY);
1850} 1900}
1851 1901
1852SV *incr_text (JSON *self) 1902SV *incr_text (JSON *self)

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines