stardict-tools/stardict-tools-3.0.1/src/ncce2stardict.pl

   1 #!/usr/bin/perl -w
   2 # by ChaosLawful@SMTH at 2006-3-23
   3 use strict;
   4 use Encode;
   5 use FileHandle;
   6 use Fcntl ':seek';      # for constants SEEK_*
   7
   8 my @ncce_dicts=("ec","ce");
   9
  10 for my $ncce_dict (@ncce_dicts) {
  11         my $buf;
  12         my $idxFH=new FileHandle("$ncce_dict.idx") or die;
  13         my $libFH=new FileHandle("$ncce_dict.lib") or die;
  14         my $outFH=new FileHandle(">ncce_$ncce_dict.tab") or die;
  15         binmode($idxFH);
  16         binmode($libFH);
  17         binmode($outFH);
  18
  19         sysread($idxFH,$buf,4);
  20         my ($totalRecord)=unpack("L",$buf);     # got total record number
  21
  22         for my $idxNo (1..$totalRecord) {
  23                 seek($idxFH,$idxNo*4,SEEK_SET); # find offset
  24                 sysread($idxFH,$buf,4);
  25
  26                 my ($off)=unpack("L",$buf);     # seek into lib file
  27                 seek($libFH,$off,SEEK_SET);
  28                 {
  29                         # and read corresponding entry record
  30                         local $/=chr(1);
  31                         chomp($buf=<$libFH>);
  32                 }
  33
  34                 $buf=pack "C*",map $_+0x1e,unpack "C*",$buf;    # decrypt record
  35
  36                 $buf=~s/\\/\\\\/gs;
  37                 my @fields=split(/\x1e/,$buf);  # split entry into word and explanation
  38                 toTextDict($outFH,$fields[0],$fields[1]);       # output
  39         }
  40 }
  41
  42 sub toTextDict
  43 {
  44         my ($fh,$word,$explain)=@_;
  45         # Kingsoft custom dictionary's export format:
  46         # every line contains one entry, whose format is:
  47         # <word>|<explanation>
  48         # where <explanation>:=<literal>[\r\n<explanation>]
  49
  50         # strip leading and trailing spaces, squeeze inner spaces
  51         $word=~s/^\s+//gs;
  52         $word=~y/ / /s;
  53         $word=~s/\s+$//gs;
  54
  55         $explain=~s/^\s+//gs;
  56         $explain=~y/ / /s;
  57         $explain=~s/\s+$//gs;
  58
  59         # split NCCE entry into multiple explanations, optional
  60         $explain=~s/;\s*/\\n/gs;
  61         # convert fullwidth comma between alphadigits into halfwidth comma, optional
  62         $word=~s/\xa3\xac(?=\w)/,/gs;
  63         $explain=~s/\xa3\xac(?=\w)/,/gs;
  64         # convert to utf-8
  65         $word=encode("utf-8",decode("cp936",$word));
  66         $explain=encode("utf-8",decode("cp936",$explain));
  67         print $fh "$word\t$explain\n";
  68 }
  69