Skip to main content

A rocker plugin to help installing apt and pip dependencies

Project description

deps_rocker

Continuous Integration Status

Ci Codecov GitHub issues GitHub pull-requests merged GitHub release PyPI - Downloads License Python

Installation

pip install deps-rocker

Usage

#recursively search for *.deps.yaml and install those packages on top of an existing image
rocker --deps ubuntu:22.04  

Motivation

Docker enables easy isolation of dependencies from the host system, but it is not easy to dynamically combine docker files from separate projects into a single unified environment.

Intro

This is a rocker extension for automating dependency installation. The aim is to allow a projects to define its development dependencies in a deps.yaml file which are added to the rocker container. If two projects define their dependencies in separate files, the extension will combine the common commands into the same docker layer to help reduce image size and duplication of work.

For example:

pkg a requires git, make and ffmpeg and pkg_b requires git-lfs and pip. Their deps.yaml files would look something like:

pkg_a.deps.yaml:

apt_sources:
  - git

apt_language-toolchain:
  - make
  - gcc

apt:
  - ffmpeg

pkg_b.deps.yaml

apt_sources:
  - git
  - git-lfs

apt_language-toolchain:
  - python3-pip

If you wanted a container that had the dependencies of both installed deps-rocker would combine the dependencies to produce a file like:

apt_sources:
  - git
  - git-lfs

apt_language-toolchain:
  - make
  - gcc
  - python3-pip

apt:
  - ffmpeg

Each heading in the yaml file produces a docker layer based on the command and the label. The format of the labels is {command_name}_{command-label}. The layer names are delimited by _ so layer names should use - eg: language-toolchain.

This makes it easy to define the dependencies for a single project, but enable reuse of common dependencies across multiple projects. However, deps rocker does not restrict what is defined in each layer and so relies on a common convention for multiple packages to play nicely with eachother. If one package adds "make" to apt_sources and other package adds "make" to apt_langage_toolchain, the deps-rocker will not complain and will not deduplicate that install step.

Methodology:

The algorithm works by splitting each entry in the yaml file into a command and a layer. The entries from all the deps.yaml files are grouped by the command and layer into a list of entries for that command. The order of the commands is defined by the order they appear in the deps.yaml file. As long as all the files follow the same order of commands then a dependency tree of commands can be created and executed in a deterministic order. However if two files define conflicting orders deps-rocker will not be able to produce a deterministic set of commands and fail. e.g:

pkg_a.deps.yaml:

apt_sources:
  - git

apt_language-toolchain:
  - make
  - gcc

pkg_b.deps.yaml

apt_language-toolchain:
  - python3-pip

apt_sources:
  - git
  - git-lfs

pkg_a says that apt_langage-toolchain comes before apt_sources, and pkg_b says that apt_sources comes before apt_language-toolchain, which is a conflict.

The pseudocode for the deps-rocker algorithm is as follows:

dependencies_dictionary
for file in glob(*.deps.yaml):
  for entry in file.entries:
    add 

If two packages have unqiue layers that depend on a common layer

pkg_a.deps.yaml:

apt_sources:
  - git

apt_pkg_a_custom:
  - custom1

pkg_b.deps.yaml

apt_sources:
  - git-lfs

apt_pkg_b_custom:
  - custom1

Here apt_pkg_b_custom and apt_pkg_a_custom both need to be run after apt_sources. They will be run run in alphabetical order (to ensure determinism)

Commands

Commands are defined in templates/commandname_snippet.Dockerfile.

They use the empy templating langage that is used by rocker. deps-rocker has some basic commands already implemented but adding a new command is as simple as adding a _snippet.Dockerfile.

Existing Commands:

  • apt: apt install packages
  • add-apt-repository: add repositories to apt
  • env: define environment variables
  • pip: install pip packages
  • run: RUN a docker command
  • script: run a script.
  • pyproject: look for any local pyproject.toml files and install dependencies listed there.

script:

If you have sudo inside your script deps-rocker will automatically remove them. This is so that you can run the script on the host machine where sudo is required.

Layer conventions

As mentioned above, deps-rocker does not enforce any particular layer order so the user can define them as they see fit, however to enhance interoperation of packages we define a suggested layer order. Examples of deps.yaml can be found in manifest_rocker

the template_pkg has common layers and dependencies that go in each layer as a guide to maximise reusability and caching. template_pkg

# Template package  Uncomment or modify these entries.

env_base:
  - DEPS_ROCKER=1

apt_base: #lowest level of dependency that changes very infrequently
  - build-essentials

apt_io: #graphics sound, input devices etc
  - libasound2

apt_sources: #apt dependencies for setting up software sources
  - ca-certificates #needed for wget
  - wget
  - curl
  - lsb-release
  - gnupg2
  - git
  - git-lfs

script_sources: #scripts for adding repositories or repo keys
  - sources.sh

apt_language-toolchain: #packges related to setting up languages e.g. c++,python,rust etc
  - python3-pip
  - make
  - gcc

pip_language-toolchain: #install basic development tools which almost never change
  - pip #this updates pip to latest version
  - flit
  - pytest
  - ruff

apt_tools: #any other development tools
  - colcon

apt: #the main dependencies of the package
  - fsearch

pyproject: #Scan for all pyproject.tomls and install
  - all

script_build: #any build steps
  - build.sh

script_lint: 
  - lint.sh

script_test:
  - test.sh


## limitations/TODO

This has only been tested on the ubuntu base image. It assumes you have access to apt-get.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

deps_rocker-0.2.6.0.tar.gz (10.0 kB view details)

Uploaded Source

Built Distribution

deps_rocker-0.2.6.0-py2.py3-none-any.whl (10.4 kB view details)

Uploaded Python 2 Python 3

File details

Details for the file deps_rocker-0.2.6.0.tar.gz.

File metadata

  • Download URL: deps_rocker-0.2.6.0.tar.gz
  • Upload date:
  • Size: 10.0 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/5.1.1 CPython/3.11.10

File hashes

Hashes for deps_rocker-0.2.6.0.tar.gz
Algorithm Hash digest
SHA256 e10d799669c1d604809af03ef7d97e2cb9c02707d6e61a988e9d32cc144915f7
MD5 877e23ae08a6fbee3504ebdbcebf263c
BLAKE2b-256 6e14a243c1225ae719caf5097b7f1f2f70b3549a86941c497b6307d29a8d6813

See more details on using hashes here.

File details

Details for the file deps_rocker-0.2.6.0-py2.py3-none-any.whl.

File metadata

File hashes

Hashes for deps_rocker-0.2.6.0-py2.py3-none-any.whl
Algorithm Hash digest
SHA256 b0f3cefdf56b921a65d8cc40b028cf7f9123d1ef6c72b8a16d857bf35c6b0dd0
MD5 9f2df1ebe91a86cf8c2c1c202021c213
BLAKE2b-256 f0b2597efeda5c28b0c37a6e2b78594c9e0af928b8a181abf0b3b44382ca355b

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page