Demo

The following demonstration covers most of the basic functions.

Try the demo yourself after installing pakit.

Install Pakit

To use pakit:

1. Ensure you have a build environment for compiling the programs.
2. Fetch pakit via pip or github.
3. Modify your $PATH.

Build Environment

At this point pakit has two limitations to be aware of:

• Relies on user's build environment.
• Pakit recipes can only depend on things pakit can build, currently limited pool. User needs to install any dependencies pakit can't build.

To use pakit, I suggest you have...

• c++ build environment
• git
• anything a recipe depends on that pakit can't build

For Ubuntu install these packages:

    sudo apt-get install build-essential automake autoconf python-pip git liblzma-dev libevent-dev ncurses-dev
  

Github

Fetch the latest from the source. Works unless the build badge says failing.

    git clone https://github.com/starcraftman/pakit.git
    export PATH=$(pwd)/pakit/bin:$PATH
    python pakit/setup.py deps release
  

pip

Install the latest stable pip release. It might be old but working.

    sudo -H pip install pakit
  

$PATH

By default, pakit will install programs under pakit.paths.prefix and link everything to pakit.paths.link. To use the built programs, pakit.paths.link/bin must be on your $PATH. So for example, with the default value of pakit.paths.link, you would need to:

    export PATH=/tmp/pakit/links/bin:$PATH
  

The above exports will only last for the terminal session. To make them permanent for bash, edit $HOME/.bashrc or $HOME/.bash_aliases.

More Information

From inside the pakit source folder:

• Help: pakit --help
• Consult man: man pakit
• Read pydocs: pydoc pakit or pydoc pakit.shell and so on...
• Install all development packages: python setup.py deps
• Run the test suite: tox
• See Waffle for things I'm working on.
• Read DESIGN.md for details on design. A bit out of date.

Contributors

• @starcraftman/Jeremy Pallats

Overview

Basically I want to make a universal package manager on python. Runs everywhere, builds anything and handles dependencies. A bit like a meta build tool tying arbitrary recipes together. At the end of the day, will probably resemble Homebrew at least a little.

Importantly, the recipes should be configurable via a single YAML file that users can modify without changing the recipes. Say you want to pass particular flags to the vim or ag build, you'd just put them in an entry in the config.

Expected Feature Overview:

• Python only, with minimal dependencies.
• Package manager interface, install remove and update recipes.
• 100% tested, framework & supported recipes.
• Should work on any POSIX system, emphasis on Linux.
• Simple recipe specification.
• Configuration via a single YAML file.
• Available via pip.
• Traceability via logs for every command.
• Premade & tested recipes available for use.

See DESIGN.md for more details.