<tool id="hadoop_galaxy_cat_paths" name="Cat paths" version="0.1.0">
  <description>Concatenate all components of a pathset into a single file.</description>
  <requirements>
    <requirement type="package" version="0.11">pydoop</requirement>
    <requirement type="package" version="0.1.1">hadoop-galaxy</requirement>
  </requirements>

  <command>
    #if $use_hadoop
      dist_cat_paths
    #else
      cat_paths
    #end if
    #if $delete_source
      --delete-source
    #end if
    $input_pathset $output_path
  </command>

  <inputs>
    <param name="input_pathset" type="data" format="pathset" label="Input pathset">
      <validator type="empty_field" />
    </param>
    <param name="delete_source" type="boolean" checked="false" label="Delete remote input data"
        help="This option makes the tool move the data rather than copy it" />
    <param name="use_hadoop" type="boolean" checked="false" label="Use Hadoop-based program"
        help="The Galaxy workspace must be accessible by the Hadoop cluster (see help for details)" />
  </inputs>

  <outputs>
      <!-- TODO: can we read the format from input pathset and transfer it to output? -->
      <data name="output_path" format="data" label="Concatenated dataset $input_pathset.name" />
  </outputs>

  <stdio>
    <exit_code range="1:" level="fatal" />
  </stdio>

  <help>
Datasets represented as pathsets can be split in a number of files.
This tool takes all of them and concatenates them into a single output file.

In your workflow, you'll need to explicitly set the appropriate data format on the
output dataset with an Action to "Change Datatype".

"Delete remote input data" option
====================================
With this option, after the data has been concated into the new Galaxy dataset,
the original files that were referenced by the pathset are deleted.  This effectively
tells the action to "move" the data instead of a "copying" it and helps
avoid amassing intermediate data in your Hadoop workspace.


"Use Hadoop-based program" option
====================================

With this option you will use your entire Hadoop cluster to simultaneously write
multiple parts of the final file.  For this to be possible, the Hadoop nodes
must be able to access the Galaxy file space directly.  In addition, to achieve
reasonable results the Galaxy workspace should on a parallel shared file system.
  </help>
</tool>