Unless you're really, really good with information theory, I doubt you could come up with something better than a general-purpose compression algorithm. Take a look at the Compress:: modules, particularly Compress::Zlib and Compress::Bzip2. These will probably reduce the space by a lot more than most customized solutions, and will probably be easier to code.

----
I wanted to explore how Perl's closures can be manipulated, and ended up creating an object system by accident.
-- Schemer

Note: All code is untested, unless otherwise stated

Before showing you the code, some notes.

• You are mixing text (the boundary delimiters) and binary data. This is bad
• You are writing out hexadecimal representations of numbers, but you want to save space. This makes little sense
• You should really use pack

Now, I changed the encoding. Now it is:

• one long, for the index length
• the index. Each entry in the index is made of:
  1. one long: the index number
  2. one long: the string length
  3. some bytes: the string itself
• the content, as a series of longs

And now, the code:

Encoder:

#!/usr/bin/perl

@lines = <>;

binmode STDOUT;
$x = 1;
foreach $ln (@lines) {
  @words = split /\s+/, $ln;
  foreach $word (@words) {
    $index{$word} = $x++;
    $count{$word}++;
  }
}

$idx='';
foreach $key (keys %index) {
  $idx.=pack("N(N/a*)",$index{$key},$key);
}

print STDOUT pack("N",length($idx)),$idx;

foreach $ln (@lines) {
  @words = split /\s+/, $ln;
  foreach $word (@words) {
    print STDOUT pack("N",$index{$word});
  }
}
[download]

Decoder:

#!/usr/bin/perl
binmode STDIN;
undef $/;
$i=<>;
$is=unpack "N",$i;
$ind=substr($i,4,$is);$con=substr($i,4+$is);
%index=unpack "(N(N/a))*",$ind;
@con=unpack "N*",$con;
print join(' ',@index{@con});
[download]

Note the use of binmode and the unsetting of $/ (aka input record separator).

-- 
        dakkar - Mobilis in mobile

Point well taken on mixing text and binmode. Doing the decode was getting pretty hairy as I had it.

I must say that I do not understand the pack("N(N/a*)" syntax. I know what N and a are but cannot find any docs on the parens; I did find docs on the slash and understand it now. (Though I have to admit that I have really TOTALLY avoided pack and unpack as they seem VERY confusing to me.)

Also why 2 N's? Seems it could be written as N/a* per the perlfunc man page.

Also, I get an error on the decoder as follows:


/ must follow a numeric type at ./dcode2.pl line 7, <> chunk 1.
[download]


Extrapolating this out it would seem that there is a syntax error in the decoder as it is written unpack "(N(N/a))*",$ind; and is seems it should probably be unpack "N(N/a*)",$ind; but that does not produce any output, just blank lines.

Ed

pack("N(N/a*)",$index{$key},$key)! meaning that the $index{$key} is written as a Number, and the $key as a length-prefixed string.

The parenthesis are for grouping. Their meaning is half-hidden in the documentation...

Regarding the error: I don't get it. I tested the code.

The unpack is right as it is: it means "a series of pairs of Numbers and length-prefixed strings (in unpack you must not put the *, see the docs)

-- 
        dakkar - Mobilis in mobile

printf "%c", $i;
[download]

binmode FILE;
print FILE pack("N",$i);
[download]

pack takes a list of values, and returns a string obtained by packing them as indicated by the first argument. Look at perldoc.

-- 
        dakkar - Mobilis in mobile