Steven Haryanto


Unixish - Data transformation framework, inspired by Unix toolbox philosophy


version 1.0.4




This document specifies Unixish, Perl framework for data processing (transformation, conversion, whatever) using the tried-and-true Unix toolbox philosophy. For the implementation, see Data::Unixish.


Early draft. The 1.0 series does not guarantee full backward compatibility between revisions, so caveat implementor. However, major incompatibility will bump the version to 1.1.


The Unix philosophy says a program should do only one thing and do it well. Problem is solved by sewing or chaining together a sequence of small, specialized programs. From Douglas McIlroy, the original developer of Unix pipelines:

 This is the Unix philosophy: Write programs that do one thing and do it well.
 Write programs to work together. Write programs to handle text streams, because
 that is a universal interface.

In Unixish, programs translate to functions. Unixish is essentially a set of guidelines and tools on how to write such functions.

The goal of the framework is to let users easily create functions that can be used as normal Perl functions operating on arrays and streams (filehandles), as well as functions that can become Unix command-line utilities.


Data::Unixish is the Perl implementation.

dux is a short notation for Data::Unixish.


The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119.

  • Function should accept a hash argument %args

    This future-proofs the function when more and more arguments are added.

  • Arguments should be described in Rinci metadata

    See Rinci and Rinci::function for more details.

  • There are some standard arguments: in, out

    in and out are analogous to standard input and output streams, explained below.

  • Arguments should have good defaults

  • Input data is given in $args{in}

    It is a "stream", usually actually a reference to array or a tied array. Function can iterate it as follows:

     while (my ($index, $item) = each @{ $args{in} }) {

    Function SHOULD NOT slurp it in memory like this, unless necessary:

     # CAUTION!
     for (@{ $args{in} }) {

    Remember that in Perl 5 for() is not lazy, the stream might contain very large amount of data or is infinite.

  • Output should be written (appended) to $args{out}

    It is a "stream", usually actually a reference to array or a tied array. Function can append output as follows:

     while (my ($index, $item) = each @{ $args{in} }) {
         push @{ $args{out} }, $res;

    Note that assigning an array directly doesn't work the way you think:

     $args{out} = [1, 2, 3];
  • Error messages can be logged to Log::Any

    Standard format for error message will be specified in the future.

  • When processing, undef/invalid/non-applicable value should generally be skipped (passed unchanged)

    For example, the date dux function accepts either an integer (assumed as Unix timestamp) or a DateTime object. Other values like undef, an empty string, or other kinds of unsupported objects should not be processed and just passed to the output stream unprocessed. A warning can be logged if needed.

A well-written dux function can be readily transformed into a usual Unix command-line utility.


Unixish is the specification.

Data::Unixish is the implementation.

Each dux function should be written in all-lowercase name, put under Data::Unixish::FUNCTION_NAME package. The function itself is put in that package with the same name. For example the Data::Unixish::date package contains the Data::Unixish::date::date function.

A further subpackaging is allowed, for example: Data::Unixish::English::count_syllables.

App::dux is a utility to access dux functions via the command-line.


Rinci and Rinci::function, another specification to leverage functions.


Similar projects on CPAN

  • Text::Pipe (2007-now)

    Similarly inspired by Unix pipes, OO. I didn't find this project before I started Unixish.

    Some of the differences: Text::Pipe is, as the name suggests, more text-oriented. It was created to unify text/template processing. Unixish on the other hand focuses on functions that can accept streams/arrays.

  • Perl Power Tools (1999-now)

    (Search for ppt in CPAN).

    Actually not quite the same thing, but the end result is roughly the same for many text-oriented Unix utilities (like wc, sum, head, tail, etc).


Steven Haryanto <>


This software is copyright (c) 2013 by Steven Haryanto.

This is free software; you can redistribute it and/or modify it under the same terms as the Perl 5 programming language system itself.