On the build command, allow use the variable to render the images

[apenella]

At this moment, the image definitions are rendered in starting stages of Stevedore, then it is not possible to use the variables that are passed through the CLI when you render the image definition. Those variables are only available in the Dockerfile.

Imagine your applications are managed in a standardized way, and you can build all of the using only one Dockerfile and one image definition.

You have the following structure:

├── apps
│   ├── app1
│   │   └── app.sh
│   ├── app2
│   │   └── app.sh
│   └── app3
│       └── app.sh
├── build
│   └── Dockerfile

The apps folder contains your application's code. Since the images for those applications can be built in the same way, you get the most out of the feature that allows you to merge build context, therefore, you define a Dockerfile in the build folder which can be used to build the image for any of those applications.

Owing that your applications are being managed in a standardized way, you should be able to have only one image definition, a definition like the following one:

images:
  app:
    "*":
      version: "{{ .Version }}-{{ .Parent.Name }}{{ .Parent.Version }}"
      registry: registry.stevedore.test
      vars:
        app_name: app1
      builder:
        driver: docker
        options:
          context:
            - path: "apps/{{ .Vars.app_name }}"
            - path: build
      parents:
        busybox:
          - "1.36"

As you can see in the build context you would need to use something like - path: "apps/{{ .Vars.app_name }}" which will allow you to choose the source of your application's code at build time, and choose the app_name when you build the image.

$ stevedore build app -v v1 -i app1 -x app_name=app1
$ stevedore build app -v v1 -i app2 -x app_name=app2
$ stevedore build app -v v1 -i app3 -x app_name=app2