######################################################################
   Archive::Tar::Merge 0.01
######################################################################

NAME
   Archive::Tar::Merge - Merge two or more tarballs into one

SYNOPSIS
       use Archive::Tar::Merge;

       my $merger = Archive::Tar::Merge->new(
           source_tarballs => ["a.tgz", "b.tgz"],
           dest_tarball    => "c.tgz"
       );

       $merger->merge();

DESCRIPTION
   "Archive::Tar::Merge" takes two or more tarballs, merges their files and
   writes the resulting directory/file structure out to a destination
   tarball.

   In the easiest case, there's no overlap between files. Then the result
   is just a union of all files in all source tarballs.

   If there's overlap, but the overlapping files are identical in the
   source and destination tarballs, there's no conflict and
   "Archive::Tar::Merge" just puts the files into the destination tarball
   without further ado.

 Deciders
   If there is a conflict, e.g. two files under the same path in two source
   tarballs which have different content or permissions, a decision needs
   to be made which of the source files is going to be used in the
   destination tarball. The "decider" parameter takes a reference to a
   function, which receives both the logical path and the physical
   locations of the two (or more) conflicting files. Based on this, it
   decides what to do:

   *   Return the array index of the winning file. If the decider gets a
       reference to an array of 2 source tarballs, returning

           return { index => 0 }

       will pick the first file while

           return { index => 1 }

       will pick the second one.

   *   Return the content of the winning file. How it comes up with this
       content is up to the decider, typically, it applies a filter to one
       or more of the source files, derives the content of the destination
       file from them and returns its content as a string.

           return { content => "content of destination file" }

   *   Ignore the file. Leave it out of the destination tarball:

           return { action => "ignore" };

   As parameters, the decider receives the file's logical source path in
   the tarball and a list of paths to the unpacked source file candidates.
   Here's an example of a decider that resolves conflicts by prioritizing
   files from the second source tarball ("b.tgz"):

       my $merger = Archive::Tar::Merge->new(
           source_tarballs => ["a.tgz", "b.tgz"],
           dest_tarball    => "c.tgz"
           decider         => \&decider,
       );

         # If there's a conflict, let the source file of the
         # last candidate win.
       sub decider {
           my($logical_src_path, $candidate_physical_paths) = @_;

             # Always return the index of the last candidate
           return { index => "-1" };
       }

       $merger->merge();

   If a decider wants to make a decision based on the content of one or
   more of the source files, it has to open and read them. A somewhat
   contrived example would be a decider which appends the content of all
   conflicting source files and adds them under the given logical path to
   the destination tarball:

       use File::Slurp;

       sub decider {
           my($logical_src_path, $candidate_physical_paths) = @_;

           my $content;

           for my $path (@$candidate_physical_paths) {
               $content .= read_file($path);
           }

           return { content => $content };
       }

   A decider receives a reference to the outgoing "Archive::Tar::Wrapper"
   object as a third parameter:

       sub decider {
           my($logical_src_path, $candidate_physical_paths, $out_tar) = @_;

   This allows for adding extra files to the outgoing tarball and perform
   all kinds of scary manipulations.

 Hooks
   Sometimes not all of the source files are supposed to be merged to the
   destination. A hook, called with the logical source path and a list of
   source file candidate physical paths, can act as a filter to suppress
   files, modify their content, or store them under a different location.

       my $merger = Archive::Tar::Merge->new(
           source_tarballs => ["a.tgz", "b.tgz"],
           dest_tarball    => "c.tgz"
           hook            => \&hook,
       );

         # Keep only "bin/*" files
       sub hook {
           my($logical_src_path, $candidate_paths) = @_;

           if($logical_src_path =~ /^bin/) {
                 # keep
               return undef;
           }

           return { action => "ignore" };
       }

   Just like deciders, hooks receive the file's logical source path in the
   tarball and a reference to an array of paths to the source file
   candidates. Hooks *differ* from deciders in that they are called with
   *every* file that is merged and not just with conflicting files.

   If a merger defines both a decider and a hook, in case of a conflict, it
   will <only> call the decider. It doesn't make sense to call the hook in
   this case because the decider already had the opportunity to totally
   render the content of the file itself, making things like source paths
   useless.

   If a hook returns "undef", the file will be kept unmodified. On

           return { action => "ignore" };

   the file will be ignored. And a hook can, just like a decider, filter
   the original file or create its content from scratch by using the
   "content" parameter.

   A hook also receives a reference to the outgoing "Archive::Tar::Wrapper"
   object.

 Using a decider to add multiple files
   If you have two tarballs which both have a different version of
   "path/to/file", you might want to add these entries as "path/to/file.1"
   and "path/to/file.2" to the outgoing tarball.

   Since the decider gets a reference to the outgoing tarball's
   "Archive::Tar::Wrapper" object, it can easily do that:

       my $merger = Archive::Tar::Merge->new(
           source_tarballs => ["a.tgz", "b.tgz"],
           decider => sub {
             my($logical_src_path, $candidate_physical_paths, $out_tar) = @_;

             my $idx = 1;
             for my $ppath (@$candidate_physical_paths) {
               $out_tar->add($logical_src_path . ".$idx", $ppath);
               $idx++;
             }

             return { action => "ignore" };
           }
       );

   Note that the decider code not only adds the different versions under
   different paths to the outgoing tarball but also tells the main
   "Archive::Tar::Merge" code to ignore the file to prevent it from adding
   yet another version.

 Post-processing the outgoing tarball
   If the "Archive::Tar::Merge" object gets the name of the outgoing
   tarball in its constructor call, its "merge()" method will write the
   outgoing tarball into the file specified right after the merge process
   has been completed.

   If "dest_tarball" isn't specified, as in

       my $merger = Archive::Tar::Merge->new(
           source_tarballs => ["a.tgz", "b.tgz"],
       );

   then writing out the resulting tarball can be done manually by using the
   "Archive::Tar::Wrapper" object returned by "merge()":

        my $out_tar = $merger->merge();
        $out_tar->write("c.tgz", $compressed);

   For more info on what to do with the "Archive::Tar::Wrapper" object,
   read its documentation.

TODO
   * Copy directories? * What if an entry is a symlink in one tarball and a
   file in another? * Permissions of target files created by content * N
   different symlinks (hash dst file)

LEGALESE
   Copyright 2007 by Mike Schilli, all rights reserved. This program is
   free software, you can redistribute it and/or modify it under the same
   terms as Perl itself.

AUTHOR
   2007, Mike Schilli <[email protected]>