… | |
… | |
32 | #define F_DEFAULT (9UL << S_MAXDEPTH) |
32 | #define F_DEFAULT (9UL << S_MAXDEPTH) |
33 | |
33 | |
34 | #define INIT_SIZE 32 // initial scalar size to be allocated |
34 | #define INIT_SIZE 32 // initial scalar size to be allocated |
35 | #define INDENT_STEP 3 // spaces per indentation level |
35 | #define INDENT_STEP 3 // spaces per indentation level |
36 | |
36 | |
37 | #define SHORT_STRING_LEN 512 // special-case strings of up to this size |
37 | #define SHORT_STRING_LEN 16384 // special-case strings of up to this size |
38 | |
38 | |
39 | #define SB do { |
39 | #define SB do { |
40 | #define SE } while (0) |
40 | #define SE } while (0) |
41 | |
41 | |
42 | #if __GNUC__ >= 3 |
42 | #if __GNUC__ >= 3 |
… | |
… | |
178 | STRLEN clen; |
178 | STRLEN clen; |
179 | UV uch; |
179 | UV uch; |
180 | |
180 | |
181 | if (is_utf8) |
181 | if (is_utf8) |
182 | { |
182 | { |
183 | //uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY); |
|
|
184 | uch = decode_utf8 (str, end - str, &clen); |
183 | uch = decode_utf8 (str, end - str, &clen); |
185 | if (clen == (STRLEN)-1) |
184 | if (clen == (STRLEN)-1) |
186 | croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); |
185 | croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); |
187 | } |
186 | } |
188 | else |
187 | else |
… | |
… | |
497 | encode_str (enc, str, len, SvUTF8 (sv)); |
496 | encode_str (enc, str, len, SvUTF8 (sv)); |
498 | encode_ch (enc, '"'); |
497 | encode_ch (enc, '"'); |
499 | } |
498 | } |
500 | else if (SvNOKp (sv)) |
499 | else if (SvNOKp (sv)) |
501 | { |
500 | { |
|
|
501 | // trust that perl will do the right thing w.r.t. JSON syntax. |
502 | need (enc, NV_DIG + 32); |
502 | need (enc, NV_DIG + 32); |
503 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
503 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
504 | enc->cur += strlen (enc->cur); |
504 | enc->cur += strlen (enc->cur); |
505 | } |
505 | } |
506 | else if (SvIOKp (sv)) |
506 | else if (SvIOKp (sv)) |
507 | { |
507 | { |
508 | // we assume we can always read an IV as a UV |
508 | // we assume we can always read an IV as a UV |
509 | if (SvUV (sv) & ~(UV)0x7fff) |
509 | if (SvUV (sv) & ~(UV)0x7fff) |
510 | { |
510 | { |
|
|
511 | // large integer, use the (rather slow) snprintf way. |
511 | need (enc, 32); |
512 | need (enc, sizeof (UV) * 3); |
512 | enc->cur += |
513 | enc->cur += |
513 | SvIsUV(sv) |
514 | SvIsUV(sv) |
514 | ? snprintf (enc->cur, 32, "%"UVuf, (UV)SvUVX (sv)) |
515 | ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv)) |
515 | : snprintf (enc->cur, 32, "%"IVdf, (IV)SvIVX (sv)); |
516 | : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv)); |
516 | } |
517 | } |
517 | else |
518 | else |
518 | { |
519 | { |
519 | // optimise the "small number case" |
520 | // optimise the "small number case" |
520 | // code will likely be branchless and use only a single multiplication |
521 | // code will likely be branchless and use only a single multiplication |
521 | I32 i = SvIV (sv); |
522 | I32 i = SvIV (sv); |
522 | U32 u; |
523 | U32 u; |
|
|
524 | char digit, nz = 0; |
523 | |
525 | |
524 | need (enc, 6); |
526 | need (enc, 6); |
525 | |
527 | |
526 | *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0; |
528 | *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0; |
527 | u = i < 0 ? -i : i; |
529 | u = i < 0 ? -i : i; |
528 | |
530 | |
529 | // convert to 4.28 fixed-point representation |
531 | // convert to 4.28 fixed-point representation |
530 | u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits |
532 | u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits |
531 | |
533 | |
532 | char digit, nz = 0; |
534 | // now output digit by digit, each time masking out the integer part |
533 | |
535 | // and multiplying by 5 while moving the decimal point one to the right, |
|
|
536 | // resulting in a net multiplication by 10. |
|
|
537 | // we always write the digit to memory but conditionally increment |
|
|
538 | // the pointer, to ease the usage of conditional move instructions. |
534 | digit = u >> 28; *enc->cur = digit + '0'; nz |= digit; enc->cur += nz ? 1 : 0; u = (u & 0xfffffff) * 5; |
539 | digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5; |
535 | digit = u >> 27; *enc->cur = digit + '0'; nz |= digit; enc->cur += nz ? 1 : 0; u = (u & 0x7ffffff) * 5; |
540 | digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5; |
536 | digit = u >> 26; *enc->cur = digit + '0'; nz |= digit; enc->cur += nz ? 1 : 0; u = (u & 0x3ffffff) * 5; |
541 | digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5; |
537 | digit = u >> 25; *enc->cur = digit + '0'; nz |= digit; enc->cur += nz ? 1 : 0; u = (u & 0x1ffffff) * 5; |
542 | digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5; |
538 | digit = u >> 24; *enc->cur = digit + '0'; nz |= digit; enc->cur += 1; |
543 | digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
539 | } |
544 | } |
540 | } |
545 | } |
541 | else if (SvROK (sv)) |
546 | else if (SvROK (sv)) |
542 | encode_rv (enc, SvRV (sv)); |
547 | encode_rv (enc, SvRV (sv)); |
543 | else if (!SvOK (sv)) |
548 | else if (!SvOK (sv)) |
… | |
… | |
625 | decode_4hex (dec_t *dec) |
630 | decode_4hex (dec_t *dec) |
626 | { |
631 | { |
627 | signed char d1, d2, d3, d4; |
632 | signed char d1, d2, d3, d4; |
628 | unsigned char *cur = (unsigned char *)dec->cur; |
633 | unsigned char *cur = (unsigned char *)dec->cur; |
629 | |
634 | |
630 | d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("four hexadecimal digits expected"); |
635 | d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected"); |
631 | d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("four hexadecimal digits expected"); |
636 | d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected"); |
632 | d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("four hexadecimal digits expected"); |
637 | d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected"); |
633 | d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("four hexadecimal digits expected"); |
638 | d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected"); |
634 | |
639 | |
635 | dec->cur += 4; |
640 | dec->cur += 4; |
636 | |
641 | |
637 | return ((UV)d1) << 12 |
642 | return ((UV)d1) << 12 |
638 | | ((UV)d2) << 8 |
643 | | ((UV)d2) << 8 |
… | |
… | |
646 | static SV * |
651 | static SV * |
647 | decode_str (dec_t *dec) |
652 | decode_str (dec_t *dec) |
648 | { |
653 | { |
649 | SV *sv = 0; |
654 | SV *sv = 0; |
650 | int utf8 = 0; |
655 | int utf8 = 0; |
|
|
656 | char *dec_cur = dec->cur; |
651 | |
657 | |
652 | do |
658 | do |
653 | { |
659 | { |
654 | char buf [SHORT_STRING_LEN + UTF8_MAXBYTES]; |
660 | char buf [SHORT_STRING_LEN + UTF8_MAXBYTES]; |
655 | char *cur = buf; |
661 | char *cur = buf; |
656 | |
662 | |
657 | do |
663 | do |
658 | { |
664 | { |
659 | unsigned char ch = *(unsigned char *)dec->cur++; |
665 | unsigned char ch = *(unsigned char *)dec_cur++; |
660 | |
666 | |
661 | if (expect_false (ch == '"')) |
667 | if (expect_false (ch == '"')) |
662 | { |
668 | { |
663 | --dec->cur; |
669 | --dec_cur; |
664 | break; |
670 | break; |
665 | } |
671 | } |
666 | else if (expect_false (ch == '\\')) |
672 | else if (expect_false (ch == '\\')) |
667 | { |
673 | { |
668 | switch (*dec->cur) |
674 | switch (*dec_cur) |
669 | { |
675 | { |
670 | case '\\': |
676 | case '\\': |
671 | case '/': |
677 | case '/': |
672 | case '"': *cur++ = *dec->cur++; break; |
678 | case '"': *cur++ = *dec_cur++; break; |
673 | |
679 | |
674 | case 'b': ++dec->cur; *cur++ = '\010'; break; |
680 | case 'b': ++dec_cur; *cur++ = '\010'; break; |
675 | case 't': ++dec->cur; *cur++ = '\011'; break; |
681 | case 't': ++dec_cur; *cur++ = '\011'; break; |
676 | case 'n': ++dec->cur; *cur++ = '\012'; break; |
682 | case 'n': ++dec_cur; *cur++ = '\012'; break; |
677 | case 'f': ++dec->cur; *cur++ = '\014'; break; |
683 | case 'f': ++dec_cur; *cur++ = '\014'; break; |
678 | case 'r': ++dec->cur; *cur++ = '\015'; break; |
684 | case 'r': ++dec_cur; *cur++ = '\015'; break; |
679 | |
685 | |
680 | case 'u': |
686 | case 'u': |
681 | { |
687 | { |
682 | UV lo, hi; |
688 | UV lo, hi; |
683 | ++dec->cur; |
689 | ++dec_cur; |
684 | |
690 | |
|
|
691 | dec->cur = dec_cur; |
685 | hi = decode_4hex (dec); |
692 | hi = decode_4hex (dec); |
|
|
693 | dec_cur = dec->cur; |
686 | if (hi == (UV)-1) |
694 | if (hi == (UV)-1) |
687 | goto fail; |
695 | goto fail; |
688 | |
696 | |
689 | // possibly a surrogate pair |
697 | // possibly a surrogate pair |
690 | if (hi >= 0xd800) |
698 | if (hi >= 0xd800) |
691 | if (hi < 0xdc00) |
699 | if (hi < 0xdc00) |
692 | { |
700 | { |
693 | if (dec->cur [0] != '\\' || dec->cur [1] != 'u') |
701 | if (dec_cur [0] != '\\' || dec_cur [1] != 'u') |
694 | ERR ("missing low surrogate character in surrogate pair"); |
702 | ERR ("missing low surrogate character in surrogate pair"); |
695 | |
703 | |
696 | dec->cur += 2; |
704 | dec_cur += 2; |
697 | |
705 | |
|
|
706 | dec->cur = dec_cur; |
698 | lo = decode_4hex (dec); |
707 | lo = decode_4hex (dec); |
|
|
708 | dec_cur = dec->cur; |
699 | if (lo == (UV)-1) |
709 | if (lo == (UV)-1) |
700 | goto fail; |
710 | goto fail; |
701 | |
711 | |
702 | if (lo < 0xdc00 || lo >= 0xe000) |
712 | if (lo < 0xdc00 || lo >= 0xe000) |
703 | ERR ("surrogate pair expected"); |
713 | ERR ("surrogate pair expected"); |
… | |
… | |
717 | *cur++ = hi; |
727 | *cur++ = hi; |
718 | } |
728 | } |
719 | break; |
729 | break; |
720 | |
730 | |
721 | default: |
731 | default: |
722 | --dec->cur; |
732 | --dec_cur; |
723 | ERR ("illegal backslash escape sequence in string"); |
733 | ERR ("illegal backslash escape sequence in string"); |
724 | } |
734 | } |
725 | } |
735 | } |
726 | else if (expect_true (ch >= 0x20 && ch <= 0x7f)) |
736 | else if (expect_true (ch >= 0x20 && ch <= 0x7f)) |
727 | *cur++ = ch; |
737 | *cur++ = ch; |
728 | else if (ch >= 0x80) |
738 | else if (ch >= 0x80) |
729 | { |
739 | { |
730 | STRLEN clen; |
740 | STRLEN clen; |
731 | UV uch; |
741 | UV uch; |
732 | |
742 | |
733 | --dec->cur; |
743 | --dec_cur; |
734 | |
744 | |
735 | uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen); |
745 | uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
736 | if (clen == (STRLEN)-1) |
746 | if (clen == (STRLEN)-1) |
737 | ERR ("malformed UTF-8 character in JSON string"); |
747 | ERR ("malformed UTF-8 character in JSON string"); |
738 | |
748 | |
739 | do |
749 | do |
740 | *cur++ = *dec->cur++; |
750 | *cur++ = *dec_cur++; |
741 | while (--clen); |
751 | while (--clen); |
742 | |
752 | |
743 | utf8 = 1; |
753 | utf8 = 1; |
744 | } |
754 | } |
745 | else |
755 | else |
746 | { |
756 | { |
747 | --dec->cur; |
757 | --dec_cur; |
748 | |
758 | |
749 | if (!ch) |
759 | if (!ch) |
750 | ERR ("unexpected end of string while parsing JSON string"); |
760 | ERR ("unexpected end of string while parsing JSON string"); |
751 | else |
761 | else |
752 | ERR ("invalid character encountered while parsing JSON string"); |
762 | ERR ("invalid character encountered while parsing JSON string"); |
… | |
… | |
765 | } |
775 | } |
766 | else |
776 | else |
767 | sv = newSVpvn (buf, len); |
777 | sv = newSVpvn (buf, len); |
768 | } |
778 | } |
769 | } |
779 | } |
770 | while (*dec->cur != '"'); |
780 | while (*dec_cur != '"'); |
771 | |
781 | |
772 | ++dec->cur; |
782 | ++dec_cur; |
773 | |
783 | |
774 | if (sv) |
784 | if (sv) |
775 | { |
785 | { |
776 | SvPOK_only (sv); |
786 | SvPOK_only (sv); |
777 | *SvEND (sv) = 0; |
787 | *SvEND (sv) = 0; |
… | |
… | |
780 | SvUTF8_on (sv); |
790 | SvUTF8_on (sv); |
781 | } |
791 | } |
782 | else |
792 | else |
783 | sv = newSVpvn ("", 0); |
793 | sv = newSVpvn ("", 0); |
784 | |
794 | |
|
|
795 | dec->cur = dec_cur; |
785 | return sv; |
796 | return sv; |
786 | |
797 | |
787 | fail: |
798 | fail: |
|
|
799 | dec->cur = dec_cur; |
788 | return 0; |
800 | return 0; |
789 | } |
801 | } |
790 | |
802 | |
791 | static SV * |
803 | static SV * |
792 | decode_num (dec_t *dec) |
804 | decode_num (dec_t *dec) |
… | |
… | |
854 | { |
866 | { |
855 | // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so |
867 | // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so |
856 | if (*start == '-') |
868 | if (*start == '-') |
857 | switch (dec->cur - start) |
869 | switch (dec->cur - start) |
858 | { |
870 | { |
859 | case 2: return newSViv (-( start [1] - '0' )); |
871 | case 2: return newSViv (-( start [1] - '0' * 1)); |
860 | case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); |
872 | case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); |
861 | case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
873 | case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
862 | case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
874 | case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
863 | } |
875 | } |
864 | else |
876 | else |
865 | switch (dec->cur - start) |
877 | switch (dec->cur - start) |
866 | { |
878 | { |
867 | case 1: return newSViv ( start [0] - '0' ); |
879 | case 1: return newSViv ( start [0] - '0' * 1); |
868 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
880 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
869 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
881 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
870 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
882 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
871 | } |
883 | } |
872 | |
884 | |
… | |
… | |
879 | if (uv < (UV)IV_MIN) |
891 | if (uv < (UV)IV_MIN) |
880 | return newSViv (-(IV)uv); |
892 | return newSViv (-(IV)uv); |
881 | } |
893 | } |
882 | else |
894 | else |
883 | return newSVuv (uv); |
895 | return newSVuv (uv); |
|
|
896 | |
|
|
897 | // here would likely be the place for bigint support |
884 | } |
898 | } |
885 | } |
899 | } |
886 | |
900 | |
|
|
901 | // if we ever support bigint or bigfloat, this is the place for bigfloat |
887 | return newSVnv (Atof (start)); |
902 | return newSVnv (Atof (start)); |
888 | |
903 | |
889 | fail: |
904 | fail: |
890 | return 0; |
905 | return 0; |
891 | } |
906 | } |
… | |
… | |
992 | |
1007 | |
993 | static SV * |
1008 | static SV * |
994 | decode_sv (dec_t *dec) |
1009 | decode_sv (dec_t *dec) |
995 | { |
1010 | { |
996 | decode_ws (dec); |
1011 | decode_ws (dec); |
|
|
1012 | |
|
|
1013 | // the beauty of JSON: you need exactly one character lookahead |
|
|
1014 | // to parse anything. |
997 | switch (*dec->cur) |
1015 | switch (*dec->cur) |
998 | { |
1016 | { |
999 | case '"': ++dec->cur; return decode_str (dec); |
1017 | case '"': ++dec->cur; return decode_str (dec); |
1000 | case '[': ++dec->cur; return decode_av (dec); |
1018 | case '[': ++dec->cur; return decode_av (dec); |
1001 | case '{': ++dec->cur; return decode_hv (dec); |
1019 | case '{': ++dec->cur; return decode_hv (dec); |
… | |
… | |
1130 | MODULE = JSON::XS PACKAGE = JSON::XS |
1148 | MODULE = JSON::XS PACKAGE = JSON::XS |
1131 | |
1149 | |
1132 | BOOT: |
1150 | BOOT: |
1133 | { |
1151 | { |
1134 | int i; |
1152 | int i; |
1135 | |
|
|
1136 | memset (decode_hexdigit, 0xff, 256); |
|
|
1137 | |
1153 | |
1138 | for (i = 0; i < 256; ++i) |
1154 | for (i = 0; i < 256; ++i) |
1139 | decode_hexdigit [i] = |
1155 | decode_hexdigit [i] = |
1140 | i >= '0' && i <= '9' ? i - '0' |
1156 | i >= '0' && i <= '9' ? i - '0' |
1141 | : i >= 'a' && i <= 'f' ? i - 'a' + 10 |
1157 | : i >= 'a' && i <= 'f' ? i - 'a' + 10 |