modelw-docker 2022.10.0a4

Utility to simplify Dockerfiles

Model W Docker

A tool so that your Dockerfile looks like this:

FROM modelw/base:2023.01

COPY --chown=user package.json package-lock.json ./

RUN modelw-docker install

COPY --chown=user . .

RUN modelw-docker build

CMD ["modelw-docker", "serve"]

Organization

This repository contains different elements that work together, found in sub-directories here:

• src — Contains the source of the modelw-docker package, that is published on Pypi.
• image — Is the source for the Docker image that can be used as a base for all Model W projects.
• demo — A demo project used to test the image during development

modelw-docker

This is a helper that is pre-installed in the Docker image and helps you build and run a Model W project.

If called from the root of a project, it will automatically detect the project's type and run appropriate commands for each step of the build. If later on the way the Docker image is built or the requirements of Model W change, it is expected that those changes can be reflected in the modelw-docker package without requiring the developers to change their Dockerfiles.

Available actions

• modelw-docker install — Installs the project's dependencies (creates the virtualenv, runs npm install or whatever is required). It only requires the dependencies files to run (package.json/package-lock.json for front components, pyproject.toml/poetry.lock for api components).
• modelw-docker build — Builds the project. It requires the project to be installed first. It also requires all the source code to be present.
• modelw-docker serve — Runs the project. It requires the project to be installed and built first.
• modelw-docker run — Runs a command in the project's virtualenv. It requires the project to be installed first.

The reason why install and build are separate and why you need first to copy just the dependencies list and then the source code is to allow for caching of the dependencies. This way, the dependencies are only re-installed when the dependencies list changes, not when the source code changes. This makes very fast builds when only the source code changes.

Dry run

There is a --dry-run option for all the commands that will just print what would have been done but not do it. The dry run mode is automatically enabled if you run the package outside of Docker in order to avoid fucking up your desktop.

Config file

All the settings are automatically detected, however if something isn't to your taste you can always override it using a model-w.toml file, following this structure:

[project]
# For printing purposes
name = "demo_project"
# Either "front" or "api"
component = "api"

[project.required_files]
# All the files to be created before the install step and their content
"src/demo_project/__init__.py" = ""

[apt.repos]
# APT repositories to be added, you need to give both the source and the key
# for each one of them
pgdg.source = "deb http://apt.postgresql.org/pub/repos/apt/ bullseye-pgdg main"
pgdg.key = { url = "https://www.postgresql.org/media/keys/ACCC4CF8.asc" }

[apt.packages]
# APT packages to be installed. Put * to install the default version, or a
# version number to install a specific version.
gdal-bin = "*"

In addition, Python project also have the following settings:

[project]
# [...]
# Either "gunicorn" or "daphne"
server = "daphne"

# Modules that have the WSGI and ASGI entry points
wsgi = "demo_project.django.wsgi:application"
asgi = "demo_project.django.asgi:application"