Package Mgmt + Docker

Systems Dependency Management

• Tools that manage programs for a single host/machine/desktop
• In the beginning there was apt
• Not really, but this is probably as far back as we care about
• apt manages system dependencies on Debian-based linux distros (Ubuntu, Raspian, etc)
• $ apt search tmux
• $ sudo apt install tmux # need root privileges to make changes
• $ sudo apt remove tmux
• Problems that arise in software development
• different projects have different dependencies (Python 2 vs Python 3)
• different tools installed on different machines (“it works on my machine…”)
• developer environment setup can be cumbersome (long list of “apt install” commands)
• system packages can often be ancient and have quirks (node.js)

Software Project Dependency Management

• Manages language/runtime specific dependencies for your project
• In JavaScript-land we have npm (and yarn which is compatible with npm)
• npm was created for node.js but is now a standard tool for front-end development
• we create a package.json, and that manages our project level dependencies
• $ npm init # create a new package.json project in a directory
• $ npm install react # add a new project dependency
• $ npm install prettier –save-dev # add a new dev dependency
• Everything is managed inside the node_modules directory (which is generally .gitignored)

Software Project Dependency Management

• In Python we started with Pip
• $ pip search poetry # no longer works apparently
• Uses a requirements.txt file to manage dependencies for a project
• Typical installs are system-wide (all the problems with system dependencies)
• Then came virtual environments (venv)
• Create a separate python environment for each project
• Then came wrappers for both venv and pip (first pipenv then poetry)
• $ pip install poetry # install poetry system-wide
• $ poetry search request
• $ poetry init # create poetry project inside a directory (pyproject.toml)
• $ poetry add request # add a dependency on request
• $ poetry install # install all dependencies specified in pyproject.toml
• $ poetry run [whatever] # run a python program within the associated venv with dependencies

Then came Docker…

• …and all was right with the world
• You can think of Docker as a wrapper around “virtual systems”
• Technically, Docker is not implemented via virtualization (although when running on a mac you do run a virtual linux machine because Docker is linux-specific)
• Docker gives you a package management like front-end to /systems/ instead of dependencies
• These “systems” can be customized and shared directly as images, or indirectly through declarative Dockerfiles
• A “Dockerfile” becomes part of your project which allows you to track it with the rest of your version control
• ​

Lab (Part 1)

1. Install Docker Engine for your system
2. Create a simple “hello world” program in JavaScript
3. Run the following command:

docker run -v .:/src -w /src node:latest test.js

• Do the same with python
• Let’s talk about what we’ve done

​