Adding `env_file:` to `process-compose.yml`

[dgmora]

Feature Request

Similarly to docker-compose env_file, this would allow each process to specify its own env file. I have searched the documentation but I don't find this option available

Use Case:

You want to share some environment variables between processes, but not all processes. You want to use an existing env file, but not for all processes

Proposed Change:

Add an env_file attribute that lets you load an env file for one process.

Who Benefits From The Change(s)?

I think this adds flexibility in how you organize and share env variables and files.

Alternative Approaches

You can probably achieve similar behaviour with a combination of passing -e and hardcoding the environment into environment

[F1bonacc1]

Hi @dgmora,

That's an interesting suggestion. As I'm not aware of your particular use case, I am wondering if those alternatives were considered and ruled out?

1. There is an option to configure process level env variables (see below).
2. This is an option to source your env vars as part of your process command (see below).

Process Environment

version: "0.5"
processes:
  hello:
    command: "echo hello $$WORLD"
    environment:
      - 'WORLD=world'

Env Sourcing in the Command

version: "0.5"
processes:
  hello:
    command: "source hello_env && echo hello $$WORLD"

[dgmora]

My use case is that I have different processes that need to share some environment variables and I wanted to avoid copy paste. What I've done finally is using vars.

I think there are use cases for specifying env_files in the same way that docker-compose does anyway:

You can share env files between processes rather than polluting the compose file. You can add multiple env files to one process, something like .env + .env.development.

[F1bonacc1]

Hi @dgmora,

Thanks for the detailed explanation.

While I understand the desire for parity with docker-compose, it's important to note the architectural difference here. Docker requires specific env_file directives because the process runs in an isolated filesystem, you can't easily source a file unless it is mounted into the container first.

Since process-compose runs processes directly on the host (or within the current shell context), we have the advantage of using standard shell features. Using command: source .env && ./my-app is actually more powerful than a native env_file implementation because it allows for dynamic variable expansion and shell scripting within the env file itself. Also the env file can be shared between different processes.

Adding a specific attribute for this adds logic to the codebase to solve a problem that the shell already solves natively and efficiently. Given that, I'm inclined to stick with the source method to keep the configuration schema simple.