****************
Planned Features
****************

This is a list of features that *may* be implemented in the future.


Reading commands from files
===========================

Commands can be read from files using :manpage:`xargs(1)`. However,
:command:`xargs` reads commands in chunks, so that :command:`para`
is called repeatedly for subsets of those commands, limiting
parallelization.

For example,

.. code:: bash

    $ for file in *.foo
    > do printf '%s\0' "$file" "${file%.foo}.bar"
    > done |
    > para -0f- -k'mv {} {}'

would be more efficient than

.. code:: bash

    $ for file in *.foo
    > do printf '%s\0' "$file" "${file%.foo}.bar"
    > done |
    > xargs -0 -n256 para -k'mv {} {}'


Dependencies
============

Para could run pipelines in parallel, rather than commands. This would allow
users to, say, to run :command:`foo` and :command:`bar` in parallel, but only
run :command:`baz` once, and if, :command:`bar` has succeeded; for example::

    $ para foo 'bar; baz'


Per-job environments and I/O streams
====================================

Jobs that lend themselves to parallelization tend to be similar.
Still, it may be useful to be able to specify different environments and
I/O redirections for different jobs, for example::

    $ para 'find /foo -name '*bar*' >find.out' 'PYTHONPATH=/qux qux >qux.out'

The syntax should mimic that of the :command:`sh`.


Calling commands via :command:`sh`
==================================

Another way to implement dependencies and per-job environments and I/O streams
would be to allow users to run commands via :command:`sh -c`.