Python

Python app in my Docker quickstarts.

With a Dockerfile

FROM python:3.9

WORKDIR /usr/src/app

COPY requirements.txt ./
RUN pip install -r requirements.txt --no-cache-dir 

COPY . .

CMD [ "python", "./app.py" ]

Notes:

Use of dot as ./ is needed.
Packages are done first and then the rest of the app - this helps for caching of layers in image building.
Packages are installed without using pip cache - this keeps the image size down.

Then run using Docker CLI or Docker Compose CLI.

Here as make commands to build the image with the name and myapp and then run it, deleting the container when it is done and not giving it a name.

container:
	docker build -t myapp .
	docker run --rm myapp

CLI only

Here we use a Python image as a drop-in replacement for running the python command.

We specify python3.9 as the image and mount the current directory (such as the repo root) inside the container. No need for a Dockerfile to copy files into the image.

This can be useful for trying out different versions of Python or testing your app in Linux on a Window or macOS host. You can edit files on the host and get the changes immediately in the container.

If you want to install any other dependencies in the container such as Node or Postgres, you are maybe better off setting up a Dockerfile and installing those in there. Or have a standalone container for database services.

Basic

This will do nothing and then exit.

It also has no access to your app or packages.

$ docker run --rm python:3.9

Add volumes

Add the contents of your app to /usr/src/myapp and set that as your working directory.

$ docker run -it --rm \
    --name my-python-app \
    -v "$PWD:/usr/src/myapp" \
    -w /usr/src/myapp \
    python:3.9

Load the contents of your virtual environment. This evaluates where venv is pointing to, so if it is a symlink then the symlink’s target will be mounted.

The order matters - app the virtual env before your app.

$ docker run -it --rm \
    --name my-python-app \
    -v "$(realpath venv):/usr/src/myapp" \
    -v "$PWD:/usr/src/myapp" \
    -w /usr/src/myapp \
    python:3.9

If you want venv to be recognized by Python, you need to add it your PYTHONPATH environment variable.

.env - I keep this file around anyway because it helps VS Code work.
```
  PYTHONPATH=myapp:venv
```

Then load that inside the container. Here simplified without the

docker run \
    # other flags here ...
    python:3.9 \
    bash -c 'source .env && python myapp/main.py'

You can test with source .env && echo $PYTHONPATH and see myapp:venv as output.

Testing

Here, we get the venv directory on the host to correspond to the default location for packages installed in a container.

So effectively these will the same and you can install on the host or the container.

venv/lib/python3.9/site-packages on host.
/usr/local/lib/python3.9/site-packages on container.

    -v "$(realpath venv)/lib:/usr/local/lib"

I could not get this to work properly though. I installed python-dev so I could get an .so file for pandas but then pip was breaking. So in this case Dockerfile would have been better.

Also in general maybe it is safer to not be root and then to install for a target user, maybe in their user directory and not shared site-packages so they don’t break.

Interactive Python console

$ docker run -it --rm \
    --name my-python-app \
    -v "$PWD:/usr/src/myapp" \
    -w /usr/src/myapp \
    python:3.9

Run Python script

$ docker run --rm \
    --name my-python-app \
    -v "$PWD:/usr/src/myapp" \
    -w /usr/src/myapp \
    python:3.9 \
    python greet.py

Run a shell command

$ docker run --rm \
    --name my-python-app \
    -v "$PWD:/usr/src/myapp" \
    -w /usr/src/myapp \
    python:3.9 \
    bash -c "cd scripts && python lib.py"

Related