The Perl Toolchain Summit needs more sponsors. If your company depends on Perl, please support this very important event.

=head1 NAME

IO::Compress::FAQ -- Frequently Asked Questions about IO::Compress

=head1 DESCRIPTION

Common questions answered.

=head2 Compatibility with Unix compress/uncompress.

Although C<Compress::Zlib> has a pair of functions called C<compress> and
C<uncompress>, they are I<not> related to the Unix programs of the same
name. The C<Compress::Zlib> module is not compatible with Unix
C<compress>.

If you have the C<uncompress> program available, you can use this to read
compressed files

    open F, "uncompress -c $filename |";
    while (<F>)
    {
        ...

Alternatively, if you have the C<gunzip> program available, you can use
this to read compressed files

    open F, "gunzip -c $filename |";
    while (<F>)
    {
        ...

and this to write compress files, if you have the C<compress> program
available

    open F, "| compress -c $filename ";
    print F "data";
    ...
    close F ;

=head2 Accessing .tar.Z files

The C<Archive::Tar> module can optionally use C<Compress::Zlib> (via the
C<IO::Zlib> module) to access tar files that have been compressed with
C<gzip>. Unfortunately tar files compressed with the Unix C<compress>
utility cannot be read by C<Compress::Zlib> and so cannot be directly
accessed by C<Archive::Tar>.

If the C<uncompress> or C<gunzip> programs are available, you can use one
of these workarounds to read C<.tar.Z> files from C<Archive::Tar>

Firstly with C<uncompress>

    use strict;
    use warnings;
    use Archive::Tar;

    open F, "uncompress -c $filename |";
    my $tar = Archive::Tar->new(*F);
    ...

and this with C<gunzip>

    use strict;
    use warnings;
    use Archive::Tar;

    open F, "gunzip -c $filename |";
    my $tar = Archive::Tar->new(*F);
    ...

Similarly, if the C<compress> program is available, you can use this to
write a C<.tar.Z> file

    use strict;
    use warnings;
    use Archive::Tar;
    use IO::File;

    my $fh = new IO::File "| compress -c >$filename";
    my $tar = Archive::Tar->new();
    ...
    $tar->write($fh);
    $fh->close ;

=head2 Accessing Zip Files

This module provides support for reading/writing zip files using the
C<IO::Compress::Zip> and C<IO::Uncompress::Unzip> modules.

The primary focus of the C<IO::Compress::Zip> and C<IO::Uncompress::Unzip>
modules is to provide an C<IO::File> compatible streaming read/write
interface to zip files/buffers. They are not fully flegged archivers. If
you are looking for an archiver check out the C<Archive::Zip> module. You
can find it on CPAN at 

    http://www.cpan.org/modules/by-module/Archive/Archive-Zip-*.tar.gz    

=head2 Compressed files and Net::FTP

The C<Net::FTP> module provides two low-level methods called C<stor> and
C<retr> that both return filehandles. These filehandles can used with the
C<IO::Compress/Uncompress> modules to compress or uncompress files read
from or written to an FTP Server on the fly, without having to create a
temporary file.

Firstly, here is code that uses C<retr> to uncompressed a file as it is
read from the FTP Server.

    use Net::FTP;
    use IO::Uncompress::Gunzip qw(:all);

    my $ftp = new Net::FTP ...

    my $retr_fh = $ftp->retr($compressed_filename);
    gunzip $retr_fh => $outFilename, AutoClose => 1
        or die "Cannot uncompress '$compressed_file': $GunzipError\n";

and this to compress a file as it is written to the FTP Server 

    use Net::FTP;
    use IO::Compress::Gzip qw(:all);

    my $stor_fh = $ftp->stor($filename);
    gzip "filename" => $stor_fh, AutoClose => 1
        or die "Cannot compress '$filename': $GzipError\n";

=head2 How do I recompress using a different compression?

This is easier that you might expect if you realise that all the
C<IO::Compress::*> objects are derived from C<IO::File> and that all the
C<IO::Uncompress::*> modules can read from an C<IO::File> filehandle.

So, for example, say you have a file compressed with gzip that you want to
recompress with bzip2. Here is all that is needed to carry out the
recompression.

    use IO::Uncompress::Gunzip ':all';
    use IO::Compress::Bzip2 ':all';

    my $gzipFile = "somefile.gz";
    my $bzipFile = "somefile.bz2";

    my $gunzip = new IO::Uncompress::Gunzip $gzipFile
        or die "Cannot gunzip $gzipFile: $GunzipError\n" ;

    bzip2 $gunzip => $bzipFile 
        or die "Cannot bzip2 to $bzipFile: $Bzip2Error\n" ;

Note, there is a limitation of this technique. Some compression file
formats store extra information along with the compressed data payload. For
example, gzip can optionally store the original filename and Zip stores a
lot of information about the original file. If the original compressed file
contains any of this extra information, it will not be transferred to the
new compressed file usign the technique above.

=head2 Apache::GZip Revisited

Below is a mod_perl Apache compression module, called C<Apache::GZip>,
taken from
F<http://perl.apache.org/docs/tutorials/tips/mod_perl_tricks/mod_perl_tricks.html#On_the_Fly_Compression>

  package Apache::GZip;
  #File: Apache::GZip.pm
  
  use strict vars;
  use Apache::Constants ':common';
  use Compress::Zlib;
  use IO::File;
  use constant GZIP_MAGIC => 0x1f8b;
  use constant OS_MAGIC => 0x03;
  
  sub handler {
      my $r = shift;
      my ($fh,$gz);
      my $file = $r->filename;
      return DECLINED unless $fh=IO::File->new($file);
      $r->header_out('Content-Encoding'=>'gzip');
      $r->send_http_header;
      return OK if $r->header_only;
  
      tie *STDOUT,'Apache::GZip',$r;
      print($_) while <$fh>;
      untie *STDOUT;
      return OK;
  }
  
  sub TIEHANDLE {
      my($class,$r) = @_;
      # initialize a deflation stream
      my $d = deflateInit(-WindowBits=>-MAX_WBITS()) || return undef;
  
      # gzip header -- don't ask how I found out
      $r->print(pack("nccVcc",GZIP_MAGIC,Z_DEFLATED,0,time(),0,OS_MAGIC));
  
      return bless { r   => $r,
                     crc =>  crc32(undef),
                     d   => $d,
                     l   =>  0 
                   },$class;
  }
  
  sub PRINT {
      my $self = shift;
      foreach (@_) {
        # deflate the data
        my $data = $self->{d}->deflate($_);
        $self->{r}->print($data);
        # keep track of its length and crc
        $self->{l} += length($_);
        $self->{crc} = crc32($_,$self->{crc});
      }
  }
  
  sub DESTROY {
     my $self = shift;
     
     # flush the output buffers
     my $data = $self->{d}->flush;
     $self->{r}->print($data);
     
     # print the CRC and the total length (uncompressed)
     $self->{r}->print(pack("LL",@{$self}{qw/crc l/}));
  }
   
  1;

Here's the Apache configuration entry you'll need to make use of it.  Once
set it will result in everything in the /compressed directory will be
compressed automagically.

  <Location /compressed>
     SetHandler  perl-script
     PerlHandler Apache::GZip
  </Location>

Although at first sight there seems to be quite a lot going on in
C<Apache::GZip>, you could sum up what the code was doing as follows --
read the contents of the file in C<< $r->filename >>, compress it and write
the compressed data to standard output. That's all.

This code has to jump through a few hoops to achieve this because

=over

=item 1.

The gzip support in C<Compress::Zlib> version 1.x can only work with a real
filesystem filehandle. The filehandles used by Apache modules are not
associated with the filesystem.

=item 2.

That means all the gzip support has to be done by hand - in this case by
creating a tied filehandle to deal with creating the gzip header and
trailer.

=back

C<IO::Compress::Gzip> doesn't have that filehandle limitation (this was one
of the reasons for writing it in the first place). So if
C<IO::Compress::Gzip> is used instead of C<Compress::Zlib> the whole tied
filehandle code can be removed. Here is the rewritten code.

  package Apache::GZip;
  
  use strict vars;
  use Apache::Constants ':common';
  use IO::Compress::Gzip;
  use IO::File;
  
  sub handler {
      my $r = shift;
      my ($fh,$gz);
      my $file = $r->filename;
      return DECLINED unless $fh=IO::File->new($file);
      $r->header_out('Content-Encoding'=>'gzip');
      $r->send_http_header;
      return OK if $r->header_only;

      my $gz = new IO::Compress::Gzip '-', Minimal => 1
          or return DECLINED ;

      print $gz $_ while <$fh>;
  
      return OK;
  }
  
or even more succinctly, like this, using a one-shot gzip

  package Apache::GZip;
  
  use strict vars;
  use Apache::Constants ':common';
  use IO::Compress::Gzip qw(gzip);
  
  sub handler {
      my $r = shift;
      $r->header_out('Content-Encoding'=>'gzip');
      $r->send_http_header;
      return OK if $r->header_only;

      gzip $r->filename => '-', Minimal => 1
        or return DECLINED ;

      return OK;
  }
   
  1;

The use of one-shot C<gzip> above just reads from C<< $r->filename >> and
writes the compressed data to standard output.

Note the use of the C<Minimal> option in the code above. When using gzip
for Content-Encoding you should I<always> use this option. In the example
above it will prevent the filename being included in the gzip header and
make the size of the gzip data stream a slight bit smaller.

=head2 Using C<InputLength> to uncompress data embedded in a larger file/buffer.

A fairly common use-case is where compressed data is embedded in a larger
file/buffer and you want to read both.

As an example consider the structure of a zip file. This is a well-defined
file format that mixes both compressed and uncompressed sections of data in
a single file. 

For the purposes of this discussion you can think of a zip file as sequence
of compressed data streams, each of which is prefixed by an uncompressed
local header. The local header contains information about the compressed
data stream, including the name of the compressed file and, in particular,
the length of the compressed data stream. 

To illustrate how to use C<InputLength> here is a script that walks a zip
file and prints out how many lines are in each compressed file (if you
intend write code to walking through a zip file for real see
L<IO::Uncompress::Unzip/"Walking through a zip file"> )

    use strict;
    use warnings;

    use IO::File;
    use IO::Uncompress::RawInflate qw(:all);

    use constant ZIP_LOCAL_HDR_SIG  => 0x04034b50;
    use constant ZIP_LOCAL_HDR_LENGTH => 30;

    my $file = $ARGV[0] ;

    my $fh = new IO::File "<$file"
                or die "Cannot open '$file': $!\n";

    while (1)
    {
        my $sig;
        my $buffer;

        my $x ;
        ($x = $fh->read($buffer, ZIP_LOCAL_HDR_LENGTH)) == ZIP_LOCAL_HDR_LENGTH 
            or die "Truncated file: $!\n";

        my $signature = unpack ("V", substr($buffer, 0, 4));

        last unless $signature == ZIP_LOCAL_HDR_SIG;

        # Read Local Header
        my $gpFlag             = unpack ("v", substr($buffer, 6, 2));
        my $compressedMethod   = unpack ("v", substr($buffer, 8, 2));
        my $compressedLength   = unpack ("V", substr($buffer, 18, 4));
        my $uncompressedLength = unpack ("V", substr($buffer, 22, 4));
        my $filename_length    = unpack ("v", substr($buffer, 26, 2)); 
        my $extra_length       = unpack ("v", substr($buffer, 28, 2));

        my $filename ;
        $fh->read($filename, $filename_length) == $filename_length 
            or die "Truncated file\n";

        $fh->read($buffer, $extra_length) == $extra_length
            or die "Truncated file\n";

        if ($compressedMethod != 8 && $compressedMethod != 0)
        {
            warn "Skipping file '$filename' - not deflated $compressedMethod\n";
            $fh->read($buffer, $compressedLength) == $compressedLength
                or die "Truncated file\n";
            next;
        }

        if ($compressedMethod == 0 && $gpFlag & 8 == 8)
        {
            die "Streamed Stored not supported for '$filename'\n";
        }

        next if $compressedLength == 0;

        # Done reading the Local Header

        my $inf = new IO::Uncompress::RawInflate $fh,
                            Transparent => 1,
                            InputLength => $compressedLength
          or die "Cannot uncompress $file [$filename]: $RawInflateError\n"  ;

        my $line_count = 0;

        while (<$inf>)
        {
            ++ $line_count;
        }

        print "$filename: $line_count\n";
    }

The majority of the code above is concerned with reading the zip local
header data. The code that I want to focus on is at the bottom. 

    while (1) {
    
        # read local zip header data
        # get $filename
        # get $compressedLength

        my $inf = new IO::Uncompress::RawInflate $fh,
                            Transparent => 1,
                            InputLength => $compressedLength
          or die "Cannot uncompress $file [$filename]: $RawInflateError\n"  ;

        my $line_count = 0;

        while (<$inf>)
        {
            ++ $line_count;
        }

        print "$filename: $line_count\n";
    }

The call to C<IO::Uncompress::RawInflate> creates a new filehandle C<$inf>
that can be used to read from the parent filehandle C<$fh>, uncompressing
it as it goes. The use of the C<InputLength> option will guarantee that
I<at most> C<$compressedLength> bytes of compressed data will be read from
the C<$fh> filehandle (The only exception is for an error case like a
truncated file or a corrupt data stream).

This means that once RawInflate is finished C<$fh> will be left at the
byte directly after the compressed data stream. 

Now consider what the code looks like without C<InputLength> 

    while (1) {
    
        # read local zip header data
        # get $filename
        # get $compressedLength

        # read all the compressed data into $data
        read($fh, $data, $compressedLength);

        my $inf = new IO::Uncompress::RawInflate \$data,
                            Transparent => 1,
          or die "Cannot uncompress $file [$filename]: $RawInflateError\n"  ;

        my $line_count = 0;

        while (<$inf>)
        {
            ++ $line_count;
        }

        print "$filename: $line_count\n";
    }

The difference here is the addition of the temporary variable C<$data>.
This is used to store a copy of the compressed data while it is being
uncompressed.

If you know that C<$compressedLength> isn't that big then using temporary
storage won't be a problem. But if C<$compressedLength> is very large or
you are writing an application that other people will use, and so have no
idea how big C<$compressedLength> will be, it could be an issue.

Using C<InputLength> avoids the use of temporary storage and means the
application can cope with large compressed data streams.

One final point -- obviously C<InputLength> can only be used whenever you
know the length of the compressed data beforehand, like here with a zip
file. 

=head1 SEE ALSO

L<Compress::Zlib>, L<IO::Compress::Gzip>, L<IO::Uncompress::Gunzip>, L<IO::Compress::Deflate>, L<IO::Uncompress::Inflate>, L<IO::Compress::RawDeflate>, L<IO::Uncompress::RawInflate>, L<IO::Compress::Bzip2>, L<IO::Uncompress::Bunzip2>, L<IO::Compress::Lzop>, L<IO::Uncompress::UnLzop>, L<IO::Compress::Lzf>, L<IO::Uncompress::UnLzf>, L<IO::Uncompress::AnyInflate>, L<IO::Uncompress::AnyUncompress>

L<Compress::Zlib::FAQ|Compress::Zlib::FAQ>

L<File::GlobMapper|File::GlobMapper>, L<Archive::Zip|Archive::Zip>,
L<Archive::Tar|Archive::Tar>,
L<IO::Zlib|IO::Zlib>

=head1 AUTHOR

This module was written by Paul Marquess, F<pmqs@cpan.org>. 

=head1 MODIFICATION HISTORY

See the Changes file.

=head1 COPYRIGHT AND LICENSE

Copyright (c) 2005-2009 Paul Marquess. All rights reserved.

This program is free software; you can redistribute it and/or
modify it under the same terms as Perl itself.