The London Perl and Raku Workshop takes place on 26th Oct 2024. If your company depends on Perl, please consider sponsoring and/or attending.


Zilla::Dist - Dist::Zilla Mixed Up



This document describes Zilla::Dist version 0.1.24.


    > zild Meta
    > # Edit the Meta file.
    > zild release


This module is a formalization of a Perl package directory layout and release process that I have been evolving for some time. I use the same layout for Bash, Ruby, Python and Node.js package publishing.

Zilla::Dist provides a Makefile and set of scripts that take a modern code layout and transform it into something that looks like a standard old Perl distribution. Under the hood zild generates everything that Dist::Zilla wants and lets dzil do the heavy lifting, but you never need to interact with Dist::Zilla stuff directly.

Directory Layout

A fully stacked top level CPAN package repository might look like this:

    Changes         # History in YAML
    Contributing    # A generated instruction file for contributing
    Meta            # Meta info for all metadata needs (including dzil)
    ReadMe.pod      # Generated from `doc/Module.swim`
    .travis.yml     # Travis file (generated)
    bin/            # Scripts
    doc/            # Swim docs
    ext/            # External repos (subrepos)
    eg/             # Examples
    lib/            # Perl `.pm` code
    pkg/            # Packaging related files
    note/           # Project notes, todo lists, ideas, specs, etc
    share/          # Shared files to distribute
    test/           # Test suite

Note a few things:

  • Sane/Readable names

  • Directories are lowercase. Never plural

  • Files are TitleCase

  • No file extensions (if possible)

  • No extra meta files like dist.ini, .travis.yml, bower.json etc

These are the best of from all the package systems I've used. They make me happy, and not tied to poor legacy standards.


Zilla::Dist uses a Makefile to do everything, but you never see it. You run commands like:

    zild make test

Run zild make help to get a list of all the targets.

The most common targets don't require you to type make. For instance, you can simply:

    zild test

Here are the most important targets:

zild release

Build the dist, then cpan-upload it.

zild test

Run the test suite.

zild install

Build and install the software. Same as install from CPAN.

zild update

Rebuild the ReadMe.pod and other generated files.

zild prereqs

Install the prereqs from CPAN that are listed in the Meta file's requires: field.

zild cpan

Turn repo into a Dist::Zilla ready subdirectory called ./cpan/. This directory has a dist.ini file.

zild dist

Basically the same as zild cpan; cd cpan; dzil build.


Start by running:

    cp `zild sharedir`/Meta .

and you'll get a Meta file template. You need to customize the Meta file with information specific to your project.

To do a release, just set a new version in the Meta file and add a Changes section using the same version. Then run:

    zild release

This will:

  • Make sure things are ready for release.

  • Update the modules with $VERSION to the new version.

  • Make a dzil ready directory of your stuff called ./cpan/.

  • Call dzil build.

  • Call cpan-upload to send the dist to CPAN.

  • Tag the git repo with the version string.

  • git push the repo and tag upstream.

Release in Depth

This section lists all the things that actually happen during the zildrelease step.

…to be completed…


I've published a lot of packages in a lot of programming languages. I like taking the best ideas and spreading them around. I like reusing ideas and code and tools as much as possible between these packages.

I trust dzil to DTRT with regard to the CPAN release process. I use almost the exact same dist.ini for some 50 CPAN packages that I've converted so far.

I don't like cluttered repos and adding new metadata files for each new tool that needs one. The dist.ini file is not bad, but I can generate it from metadata easily. So I do.

As much as these great new ideas differ from the norm, I want my CPAN publishings to be normal to normal mongers (if there's such a thing). The zild release process does just that. End users would have to look hard to know this wasn't a "normal" dzil release.

I'm packaging this packaging process as Zilla::Dist for others to use. It's also a decent example of a CPAN package packaged with itself.


Some of the tools in Zilla::Dist are Bash, some are Perl. I'm doing a lot in the area of Bash Package packaging. See

I use the term Package where CPAN people have used the term Distribution. Perl is the only language (in my packaging experience) to do so.

The name t/ is another outlier. The most common is test/ followed by tests/.

I don't like plural directory names. Try singular. I think you'll like it too.


Commiting Generated Code/Files

People think that committing generated code/files is a bad idea and in general I concur, but there are exceptions.

Sometimes tools like Travis-CI require you to commit a config file. Zilla::Dist generates these files from metadata, which is a whole lot easier than maintaining them yourself, but you end up commiting generated code.

The dist.ini file is only needed locally, however, during dist build time, so no need to commit that.

In general, when an external tool requires files, and it's easiest to generate those files, it's OK to commit generated code.

Modules Published to CPAN w/ zild


Ingy döt Net <>


Copyright 2014-2024. Ingy döt Net.

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