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 1.130 by root, Sat Nov 26 06:09:29 2016 UTC vs.
Revision 1.142 by root, Mon Jun 21 00:12:57 2021 UTC

50#define F_ALLOW_TAGS 0x00004000UL 50#define F_ALLOW_TAGS 0x00004000UL
51#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing 51#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
52 52
53#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 53#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
54 54
55#define INIT_SIZE 32 // initial scalar size to be allocated 55#define INIT_SIZE 64 // initial scalar size to be allocated
56#define INDENT_STEP 3 // spaces per indentation level 56#define INDENT_STEP 3 // spaces per indentation level
57 57
58#define SHORT_STRING_LEN 16384 // special-case strings of up to this size 58#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
59 59
60#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8) 60#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
78 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) 78 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
79 79
80#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)" 80#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
81 81
82#ifdef USE_ITHREADS 82#ifdef USE_ITHREADS
83# define JSON_SLOW 1
84# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) 83# define JSON_STASH (expect_true (json_stash) ? json_stash : gv_stashpv ("JSON::XS", 1))
85# define BOOL_STASH (bool_stash ? bool_stash : gv_stashpv ("Types::Serialiser::Boolean", 1)) 84# define BOOL_STASH (expect_true (bool_stash) ? bool_stash : gv_stashpv ("Types::Serialiser::Boolean", 1))
85# define GET_BOOL(value) (expect_true (bool_ ## value) ? bool_ ## value : get_bool ("Types::Serialiser::" # value))
86#else 86#else
87# define JSON_SLOW 0
88# define JSON_STASH json_stash 87# define JSON_STASH json_stash
89# define BOOL_STASH bool_stash 88# define BOOL_STASH bool_stash
89# define GET_BOOL(value) bool_ ## value
90#endif 90#endif
91 91
92// the amount of HEs to allocate on the stack, when sorting keys 92// the amount of HEs to allocate on the stack, when sorting keys
93#define STACK_HES 64 93#define STACK_HES 64
94 94
95static HV *json_stash, *bool_stash; // JSON::XS::, Types::Serialiser::Boolean:: 95static HV *json_stash, *bool_stash; // JSON::XS::, Types::Serialiser::Boolean::
96static SV *bool_true, *bool_false, *sv_json; 96static SV *bool_false, *bool_true;
97static SV *sv_json;
97 98
98enum { 99enum {
99 INCR_M_WS = 0, // initial whitespace skipping, must be 0 100 INCR_M_WS = 0, // initial whitespace skipping, must be 0
101 INCR_M_TFN, // inside true/false/null
102 INCR_M_NUM, // inside number
100 INCR_M_STR, // inside string 103 INCR_M_STR, // inside string
101 INCR_M_BS, // inside backslash 104 INCR_M_BS, // inside backslash
102 INCR_M_C0, // inside comment in initial whitespace sequence 105 INCR_M_C0, // inside comment in initial whitespace sequence
103 INCR_M_C1, // inside comment in other places 106 INCR_M_C1, // inside comment in other places
104 INCR_M_JSON // outside anything, count nesting 107 INCR_M_JSON // outside anything, count nesting
117 // for the incremental parser 120 // for the incremental parser
118 SV *incr_text; // the source text so far 121 SV *incr_text; // the source text so far
119 STRLEN incr_pos; // the current offset into the text 122 STRLEN incr_pos; // the current offset into the text
120 int incr_nest; // {[]}-nesting level 123 int incr_nest; // {[]}-nesting level
121 unsigned char incr_mode; 124 unsigned char incr_mode;
125
126 SV *v_false, *v_true;
122} JSON; 127} JSON;
123 128
124INLINE void 129INLINE void
125json_init (JSON *json) 130json_init (JSON *json)
126{ 131{
127 Zero (json, 1, JSON); 132 static const JSON init = { F_ALLOW_NONREF, 512 };
128 json->max_depth = 512; 133
134 *json = init;
129} 135}
130 136
131///////////////////////////////////////////////////////////////////////////// 137/////////////////////////////////////////////////////////////////////////////
132// utility functions 138// utility functions
133 139
155 SvPV_renew (sv, SvCUR (sv) + 1); 161 SvPV_renew (sv, SvCUR (sv) + 1);
156#endif 162#endif
157 } 163 }
158} 164}
159 165
166/* adds two STRLENs together, slow, and with paranoia */
167static STRLEN
168strlen_sum (STRLEN l1, STRLEN l2)
169{
170 size_t sum = l1 + l2;
171
172 if (sum < (size_t)l2 || sum != (size_t)(STRLEN)sum)
173 croak ("JSON::XS: string size overflow");
174
175 return sum;
176}
177
178/* similar to SvGROW, but somewhat safer and guarantees exponential realloc strategy */
179static char *
180json_sv_grow (SV *sv, size_t len1, size_t len2)
181{
182 len1 = strlen_sum (len1, len2);
183 len1 = strlen_sum (len1, len1 >> 1);
184
185 if (len1 > 4096 - 24)
186 len1 = (len1 | 4095) - 24;
187
188 return SvGROW (sv, len1);
189}
190
160// decode an utf-8 character and return it, or (UV)-1 in 191// decode a utf-8 character and return it, or (UV)-1 in
161// case of an error. 192// case of an error.
162// we special-case "safe" characters from U+80 .. U+7FF, 193// we special-case "safe" characters from U+80 .. U+7FF,
163// but use the very good perl function to parse anything else. 194// but use the very good perl function to parse anything else.
164// note that we never call this function for a ascii codepoints 195// note that we never call this function for a ascii codepoints
165INLINE UV 196INLINE UV
365need (enc_t *enc, STRLEN len) 396need (enc_t *enc, STRLEN len)
366{ 397{
367 if (expect_false ((uintptr_t)(enc->end - enc->cur) < len)) 398 if (expect_false ((uintptr_t)(enc->end - enc->cur) < len))
368 { 399 {
369 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); 400 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
370 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); 401 char *buf = json_sv_grow (enc->sv, cur, len);
371 enc->cur = SvPVX (enc->sv) + cur; 402 enc->cur = buf + cur;
372 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 403 enc->end = buf + SvLEN (enc->sv) - 1;
373 } 404 }
374} 405}
375 406
376INLINE void 407INLINE void
377encode_ch (enc_t *enc, char ch) 408encode_ch (enc_t *enc, char ch)
393 424
394 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case 425 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
395 { 426 {
396 if (expect_false (ch == '"')) // but with slow exceptions 427 if (expect_false (ch == '"')) // but with slow exceptions
397 { 428 {
398 need (enc, len += 1); 429 need (enc, len + 1);
399 *enc->cur++ = '\\'; 430 *enc->cur++ = '\\';
400 *enc->cur++ = '"'; 431 *enc->cur++ = '"';
401 } 432 }
402 else if (expect_false (ch == '\\')) 433 else if (expect_false (ch == '\\'))
403 { 434 {
404 need (enc, len += 1); 435 need (enc, len + 1);
405 *enc->cur++ = '\\'; 436 *enc->cur++ = '\\';
406 *enc->cur++ = '\\'; 437 *enc->cur++ = '\\';
407 } 438 }
408 else 439 else
409 *enc->cur++ = ch; 440 *enc->cur++ = ch;
412 } 443 }
413 else 444 else
414 { 445 {
415 switch (ch) 446 switch (ch)
416 { 447 {
417 case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break; 448 case '\010': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
418 case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break; 449 case '\011': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
419 case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break; 450 case '\012': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
420 case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break; 451 case '\014': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
421 case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break; 452 case '\015': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
422 453
423 default: 454 default:
424 { 455 {
425 STRLEN clen; 456 STRLEN clen;
426 UV uch; 457 UV uch;
442 if (uch >= 0x10000UL) 473 if (uch >= 0x10000UL)
443 { 474 {
444 if (uch >= 0x110000UL) 475 if (uch >= 0x110000UL)
445 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); 476 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
446 477
447 need (enc, len += 11); 478 need (enc, len + 11);
448 sprintf (enc->cur, "\\u%04x\\u%04x", 479 sprintf (enc->cur, "\\u%04x\\u%04x",
449 (int)((uch - 0x10000) / 0x400 + 0xD800), 480 (int)((uch - 0x10000) / 0x400 + 0xD800),
450 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 481 (int)((uch - 0x10000) % 0x400 + 0xDC00));
451 enc->cur += 12; 482 enc->cur += 12;
452 } 483 }
453 else 484 else
454 { 485 {
455 need (enc, len += 5); 486 need (enc, len + 5);
456 *enc->cur++ = '\\'; 487 *enc->cur++ = '\\';
457 *enc->cur++ = 'u'; 488 *enc->cur++ = 'u';
458 *enc->cur++ = PL_hexdigit [ uch >> 12 ]; 489 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
459 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15]; 490 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
460 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15]; 491 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
468 *enc->cur++ = uch; 499 *enc->cur++ = uch;
469 str += clen; 500 str += clen;
470 } 501 }
471 else if (is_utf8) 502 else if (is_utf8)
472 { 503 {
473 need (enc, len += clen); 504 need (enc, len + clen);
474 do 505 do
475 { 506 {
476 *enc->cur++ = *str++; 507 *enc->cur++ = *str++;
477 } 508 }
478 while (--clen); 509 while (--clen);
479 } 510 }
480 else 511 else
481 { 512 {
482 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed 513 need (enc, len + UTF8_MAXBYTES - 1); // never more than 11 bytes needed
483 enc->cur = encode_utf8 (enc->cur, uch); 514 enc->cur = encode_utf8 (enc->cur, uch);
484 ++str; 515 ++str;
485 } 516 }
486 } 517 }
487 } 518 }
751 { 782 {
752 HV *stash = SvSTASH (sv); 783 HV *stash = SvSTASH (sv);
753 784
754 if (stash == bool_stash) 785 if (stash == bool_stash)
755 { 786 {
756 if (SvIV (sv))
757 encode_str (enc, "true", 4, 0); 787 if (SvIV (sv)) encode_str (enc, "true" , 4, 0);
758 else
759 encode_str (enc, "false", 5, 0); 788 else encode_str (enc, "false", 5, 0);
760 } 789 }
761 else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0))) 790 else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
762 { 791 {
763 int count; 792 int count;
764 dSP; 793 dSP;
765 794
766 ENTER; SAVETMPS; 795 ENTER; SAVETMPS;
767 SAVESTACK_POS ();
768 PUSHMARK (SP); 796 PUSHMARK (SP);
769 EXTEND (SP, 2); 797 EXTEND (SP, 2);
770 // we re-bless the reference to get overload and other niceties right 798 // we re-bless the reference to get overload and other niceties right
771 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash)); 799 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
772 PUSHs (sv_json); 800 PUSHs (sv_json);
784 encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash)); 812 encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
785 encode_ch (enc, '"'); 813 encode_ch (enc, '"');
786 encode_ch (enc, ')'); 814 encode_ch (enc, ')');
787 encode_ch (enc, '['); 815 encode_ch (enc, '[');
788 816
789 while (count) 817 if (count)
790 { 818 {
819 int i;
820
821 for (i = 0; i < count - 1; ++i)
822 {
791 encode_sv (enc, SP[1 - count--]); 823 encode_sv (enc, SP[i + 1 - count]);
792
793 if (count)
794 encode_ch (enc, ','); 824 encode_ch (enc, ',');
825 }
826
827 encode_sv (enc, TOPs);
828 SP -= count;
795 } 829 }
830
831 PUTBACK;
796 832
797 encode_ch (enc, ']'); 833 encode_ch (enc, ']');
798 834
799 FREETMPS; LEAVE; 835 FREETMPS; LEAVE;
800 } 836 }
882 if (SvIsUV (sv) ? SvUVX (sv) <= 59000 918 if (SvIsUV (sv) ? SvUVX (sv) <= 59000
883 : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000) 919 : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
884 { 920 {
885 // optimise the "small number case" 921 // optimise the "small number case"
886 // code will likely be branchless and use only a single multiplication 922 // code will likely be branchless and use only a single multiplication
887 // works for numbers up to 59074 923 // 4.28 works for numbers up to 59074, 4.29 works up to 59076
888 I32 i = SvIVX (sv); 924 I32 i = SvIVX (sv);
889 U32 u; 925 U32 u;
890 char digit, nz = 0; 926 char digit, nz = 0;
891 927
892 need (enc, 6); 928 need (enc, 6);
1002 else 1038 else
1003 break; 1039 break;
1004 } 1040 }
1005 else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09) 1041 else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
1006 break; // parse error, but let higher level handle it, gives better error messages 1042 break; // parse error, but let higher level handle it, gives better error messages
1007 1043 else
1008 ++dec->cur; 1044 ++dec->cur;
1009 } 1045 }
1010} 1046}
1011 1047
1012#define ERR(reason) SB dec->err = reason; goto fail; SE 1048#define ERR(reason) SB dec->err = reason; goto fail; SE
1013 1049
1168 1204
1169 if (sv) 1205 if (sv)
1170 { 1206 {
1171 STRLEN cur = SvCUR (sv); 1207 STRLEN cur = SvCUR (sv);
1172 1208
1173 if (SvLEN (sv) <= cur + len) 1209 if (SvLEN (sv) - cur <= len)
1174 SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); 1210 json_sv_grow (sv, cur, len);
1175 1211
1176 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1212 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
1177 SvCUR_set (sv, SvCUR (sv) + len); 1213 SvCUR_set (sv, SvCUR (sv) + len);
1178 } 1214 }
1179 else 1215 else
1470 1506
1471 DEC_DEC_DEPTH; 1507 DEC_DEC_DEPTH;
1472 sv = newRV_noinc ((SV *)hv); 1508 sv = newRV_noinc ((SV *)hv);
1473 1509
1474 // check filter callbacks 1510 // check filter callbacks
1475 if (dec->json.flags & F_HOOK) 1511 if (expect_false (dec->json.flags & F_HOOK))
1476 { 1512 {
1477 if (dec->json.cb_sk_object && HvKEYS (hv) == 1) 1513 if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
1478 { 1514 {
1479 HE *cb, *he; 1515 HE *cb, *he;
1480 1516
1490 { 1526 {
1491 dSP; 1527 dSP;
1492 int count; 1528 int count;
1493 1529
1494 ENTER; SAVETMPS; 1530 ENTER; SAVETMPS;
1495 SAVESTACK_POS ();
1496 PUSHMARK (SP); 1531 PUSHMARK (SP);
1497 XPUSHs (HeVAL (he)); 1532 XPUSHs (HeVAL (he));
1498 sv_2mortal (sv); 1533 sv_2mortal (sv);
1499 1534
1500 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1535 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1501 1536
1502 if (count == 1) 1537 if (count == 1)
1503 { 1538 {
1504 sv = newSVsv (POPs); 1539 sv = newSVsv (POPs);
1540 PUTBACK;
1505 FREETMPS; LEAVE; 1541 FREETMPS; LEAVE;
1506 return sv; 1542 return sv;
1507 } 1543 }
1544 else if (count)
1545 croak ("filter_json_single_key_object callbacks must not return more than one scalar");
1546
1547 PUTBACK;
1508 1548
1509 SvREFCNT_inc (sv); 1549 SvREFCNT_inc (sv);
1550
1510 FREETMPS; LEAVE; 1551 FREETMPS; LEAVE;
1511 } 1552 }
1512 } 1553 }
1513 1554
1514 if (dec->json.cb_object) 1555 if (dec->json.cb_object)
1515 { 1556 {
1516 dSP; 1557 dSP;
1517 int count; 1558 int count;
1518 1559
1519 ENTER; SAVETMPS; 1560 ENTER; SAVETMPS;
1520 SAVESTACK_POS ();
1521 PUSHMARK (SP); 1561 PUSHMARK (SP);
1522 XPUSHs (sv_2mortal (sv)); 1562 XPUSHs (sv_2mortal (sv));
1523 1563
1524 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; 1564 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1525 1565
1526 if (count == 1) 1566 if (count == 1)
1527 {
1528 sv = newSVsv (POPs); 1567 sv = newSVsv (POPs);
1529 FREETMPS; LEAVE; 1568 else if (count == 0)
1530 return sv;
1531 }
1532
1533 SvREFCNT_inc (sv); 1569 SvREFCNT_inc (sv);
1570 else
1571 croak ("filter_json_object callbacks must not return more than one scalar");
1572
1573 PUTBACK;
1574
1534 FREETMPS; LEAVE; 1575 FREETMPS; LEAVE;
1535 } 1576 }
1536 } 1577 }
1537 1578
1538 return sv; 1579 return sv;
1641 case '-': 1682 case '-':
1642 case '0': case '1': case '2': case '3': case '4': 1683 case '0': case '1': case '2': case '3': case '4':
1643 case '5': case '6': case '7': case '8': case '9': 1684 case '5': case '6': case '7': case '8': case '9':
1644 return decode_num (dec); 1685 return decode_num (dec);
1645 1686
1687 case 'f':
1688 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1689 {
1690 dec->cur += 5;
1691
1692 if (expect_false (!dec->json.v_false))
1693 dec->json.v_false = GET_BOOL (false);
1694
1695 return newSVsv (dec->json.v_false);
1696 }
1697 else
1698 ERR ("'false' expected");
1699
1700 break;
1701
1646 case 't': 1702 case 't':
1647 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1703 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1648 { 1704 {
1649 dec->cur += 4; 1705 dec->cur += 4;
1650#if JSON_SLOW 1706
1651 bool_true = get_bool ("Types::Serialiser::true"); 1707 if (expect_false (!dec->json.v_true))
1652#endif 1708 dec->json.v_true = GET_BOOL (true);
1709
1653 return newSVsv (bool_true); 1710 return newSVsv (dec->json.v_true);
1654 } 1711 }
1655 else 1712 else
1656 ERR ("'true' expected"); 1713 ERR ("'true' expected");
1657
1658 break;
1659
1660 case 'f':
1661 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1662 {
1663 dec->cur += 5;
1664#if JSON_SLOW
1665 bool_false = get_bool ("Types::Serialiser::false");
1666#endif
1667 return newSVsv (bool_false);
1668 }
1669 else
1670 ERR ("'false' expected");
1671 1714
1672 break; 1715 break;
1673 1716
1674 case 'n': 1717 case 'n':
1675 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) 1718 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
1759 else if (sv) 1802 else if (sv)
1760 { 1803 {
1761 // check for trailing garbage 1804 // check for trailing garbage
1762 decode_ws (&dec); 1805 decode_ws (&dec);
1763 1806
1764 if (*dec.cur) 1807 if (dec.cur != dec.end)
1765 { 1808 {
1766 dec.err = "garbage after JSON object"; 1809 dec.err = "garbage after JSON object";
1767 SvREFCNT_dec (sv); 1810 SvREFCNT_dec (sv);
1768 sv = 0; 1811 sv = 0;
1769 } 1812 }
1807 // the state machine here is a bit convoluted and could be simplified a lot 1850 // the state machine here is a bit convoluted and could be simplified a lot
1808 // but this would make it slower, so... 1851 // but this would make it slower, so...
1809 1852
1810 for (;;) 1853 for (;;)
1811 { 1854 {
1812 //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
1813 switch (self->incr_mode) 1855 switch (self->incr_mode)
1814 { 1856 {
1857 // reached end of a scalar, see if we are inside a nested structure or not
1858 end_of_scalar:
1859 self->incr_mode = INCR_M_JSON;
1860
1861 if (self->incr_nest) // end of a scalar inside array, object or tag
1862 goto incr_m_json;
1863 else // end of scalar outside structure, json text ends here
1864 goto interrupt;
1865
1815 // only used for initial whitespace skipping 1866 // only used for initial whitespace skipping
1816 case INCR_M_WS: 1867 case INCR_M_WS:
1817 for (;;) 1868 for (;;)
1818 { 1869 {
1819 if (*p > 0x20) 1870 if (*p > 0x20)
1861 ++p; 1912 ++p;
1862 } 1913 }
1863 1914
1864 break; 1915 break;
1865 1916
1917 // inside true/false/null
1918 case INCR_M_TFN:
1919 incr_m_tfn:
1920 for (;;)
1921 switch (*p++)
1922 {
1923 case 'r': case 'u': case 'e': // tRUE, falsE, nUll
1924 case 'a': case 'l': case 's': // fALSe, nuLL
1925 // allowed
1926 break;
1927
1928 default:
1929 --p;
1930 goto end_of_scalar;
1931 }
1932
1933 // inside a number
1934 case INCR_M_NUM:
1935 incr_m_num:
1936 for (;;)
1937 switch (*p++)
1938 {
1939 case 'e': case 'E': case '.': case '+':
1940 case '-':
1941 case '0': case '1': case '2': case '3': case '4':
1942 case '5': case '6': case '7': case '8': case '9':
1943 // allowed
1944 break;
1945
1946 default:
1947 --p;
1948 goto end_of_scalar;
1949 }
1950
1866 // inside a string 1951 // inside a string
1867 case INCR_M_STR: 1952 case INCR_M_STR:
1868 incr_m_str: 1953 incr_m_str:
1869 for (;;) 1954 for (;;)
1870 { 1955 {
1871 if (*p == '"') 1956 if (*p == '"')
1872 { 1957 {
1873 ++p; 1958 ++p;
1874 self->incr_mode = INCR_M_JSON;
1875
1876 if (!self->incr_nest)
1877 goto interrupt;
1878
1879 goto incr_m_json; 1959 goto end_of_scalar;
1880 } 1960 }
1881 else if (*p == '\\') 1961 else if (*p == '\\')
1882 { 1962 {
1883 ++p; // "virtually" consumes character after \ 1963 ++p; // "virtually" consumes character after \
1884 1964
1914 --p; // do not eat the whitespace, let the next round do it 1994 --p; // do not eat the whitespace, let the next round do it
1915 goto interrupt; 1995 goto interrupt;
1916 } 1996 }
1917 break; 1997 break;
1918 1998
1999 // the following three blocks handle scalars. this makes the parser
2000 // more strict than required inside arrays or objects, and could
2001 // be moved to a special case on the toplevel (except strings)
2002 case 't':
2003 case 'f':
2004 case 'n':
2005 self->incr_mode = INCR_M_TFN;
2006 goto incr_m_tfn;
2007
2008 case '-':
2009 case '0': case '1': case '2': case '3': case '4':
2010 case '5': case '6': case '7': case '8': case '9':
2011 self->incr_mode = INCR_M_NUM;
2012 goto incr_m_num;
2013
1919 case '"': 2014 case '"':
1920 self->incr_mode = INCR_M_STR; 2015 self->incr_mode = INCR_M_STR;
1921 goto incr_m_str; 2016 goto incr_m_str;
1922 2017
1923 case '[': 2018 case '[':
1970 : i >= 'A' && i <= 'F' ? i - 'A' + 10 2065 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1971 : -1; 2066 : -1;
1972 2067
1973 json_stash = gv_stashpv ("JSON::XS" , 1); 2068 json_stash = gv_stashpv ("JSON::XS" , 1);
1974 bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1); 2069 bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
2070 bool_false = get_bool ("Types::Serialiser::false");
1975 bool_true = get_bool ("Types::Serialiser::true"); 2071 bool_true = get_bool ("Types::Serialiser::true");
1976 bool_false = get_bool ("Types::Serialiser::false");
1977 2072
1978 sv_json = newSVpv ("JSON", 0); 2073 sv_json = newSVpv ("JSON", 0);
1979 SvREADONLY_on (sv_json); 2074 SvREADONLY_on (sv_json);
1980 2075
1981 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */ 2076 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1983 2078
1984PROTOTYPES: DISABLE 2079PROTOTYPES: DISABLE
1985 2080
1986void CLONE (...) 2081void CLONE (...)
1987 CODE: 2082 CODE:
2083 // as long as these writes are atomic, the race should not matter
2084 // as existing threads either already use 0, or use the old value,
2085 // which is sitll correct for the initial thread.
1988 json_stash = 0; 2086 json_stash = 0;
1989 bool_stash = 0; 2087 bool_stash = 0;
2088 bool_false = 0;
2089 bool_true = 0;
1990 2090
1991void new (char *klass) 2091void new (char *klass)
1992 PPCODE: 2092 PPCODE:
1993{ 2093{
1994 SV *pv = NEWSV (0, sizeof (JSON)); 2094 SV *pv = NEWSV (0, sizeof (JSON));
1997 XPUSHs (sv_2mortal (sv_bless ( 2097 XPUSHs (sv_2mortal (sv_bless (
1998 newRV_noinc (pv), 2098 newRV_noinc (pv),
1999 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 2099 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
2000 ))); 2100 )));
2001} 2101}
2102
2103void boolean_values (JSON *self, SV *v_false = 0, SV *v_true = 0)
2104 PPCODE:
2105 self->v_false = newSVsv (v_false);
2106 self->v_true = newSVsv (v_true);
2107 XPUSHs (ST (0));
2108
2109void get_boolean_values (JSON *self)
2110 PPCODE:
2111 if (self->v_false && self->v_true)
2112 {
2113 EXTEND (SP, 2);
2114 PUSHs (self->v_false);
2115 PUSHs (self->v_true);
2116 }
2002 2117
2003void ascii (JSON *self, int enable = 1) 2118void ascii (JSON *self, int enable = 1)
2004 ALIAS: 2119 ALIAS:
2005 ascii = F_ASCII 2120 ascii = F_ASCII
2006 latin1 = F_LATIN1 2121 latin1 = F_LATIN1
2159 { 2274 {
2160 STRLEN len; 2275 STRLEN len;
2161 const char *str = SvPV (jsonstr, len); 2276 const char *str = SvPV (jsonstr, len);
2162 STRLEN cur = SvCUR (self->incr_text); 2277 STRLEN cur = SvCUR (self->incr_text);
2163 2278
2164 if (SvLEN (self->incr_text) <= cur + len) 2279 if (SvLEN (self->incr_text) - cur <= len)
2165 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); 2280 json_sv_grow (self->incr_text, cur, len);
2166 2281
2167 Move (str, SvEND (self->incr_text), len, char); 2282 Move (str, SvEND (self->incr_text), len, char);
2168 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); 2283 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
2169 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there 2284 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
2170 } 2285 }
2243 self->incr_mode = 0; 2358 self->incr_mode = 0;
2244} 2359}
2245 2360
2246void DESTROY (JSON *self) 2361void DESTROY (JSON *self)
2247 CODE: 2362 CODE:
2363 SvREFCNT_dec (self->v_false);
2364 SvREFCNT_dec (self->v_true);
2248 SvREFCNT_dec (self->cb_sk_object); 2365 SvREFCNT_dec (self->cb_sk_object);
2249 SvREFCNT_dec (self->cb_object); 2366 SvREFCNT_dec (self->cb_object);
2250 SvREFCNT_dec (self->incr_text); 2367 SvREFCNT_dec (self->incr_text);
2251 2368
2252PROTOTYPES: ENABLE 2369PROTOTYPES: ENABLE

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines