In a terragrunt monorepo you will usually have multiple dependant modules where the order of execution is important. The dependencies can be represented as rings or layers where the innermost layer is dependend on by the layer above which is depended on by the layer above and so on. In the example above we can consider modules core/network and core/security to be in layer 0. Modules layer1/compute and layer1/storage and layer1/monitoring are in layer 1. Modules layer2/api and layer2/analytics and layer2/frontend are in layer2, the outermost layer. So usually what we want is if we have a change in core/network, we first want to plan and apply that change. Subsequently we want to plan and apply the changes of layer1/compute, layer1/storage and layer1/monitoring, since they depend on the outputs of core. Finally we want to plan and apply the modules in the layer2 group since they depend on the modules in layer1. In digger this could have been done individually by using the -p flag such as “digger apply -p core_network” and doing the appiles in the intended order, but this was cumbersome and tiring. With layering you can achieve this layer by layer in a single command. To enable layering in the terragrunt syntax you can do it as follows:
respect_layers: true

generate_projects:
  terragrunt_parsing:
    parallel: true
    createProjectName: true
    createWorkspace: true
    defaultWorkflow: default
    cascadeDependencies: false
    dependsOnOrdering: false
    executionOrderGroups: true
With layering enabled when you open a PR touching a module, digger will first display all impacted projects and the layers, it will then guide you on the next commands to start planning and applying the layers in the right order

Usage with other flags

Using layering with other include_patterns works well to coordinate changes across layers. For example say you have a VPC module which is used in to generate some subnets and you need to use those subnet layers in a few other modules. You can have digger.yml as follows:
respect_layers: true

projects:
  - name: vpc
    dir: vpc
    layer: 1
  - name: service_1
    dir: service_1
    layer: 2
    include_patterns: ["vpc/**"]
  - name: service_2
    dir: l3_1
    layer: 2
    include_patterns: ["vpc/**"]
and with that if you make a change to the vpc module you will be able to first plan and apply vpc, then plan and apply service_1 and service_2 in parallel.

Current Limitations

There are a few limitations currently which will be addressed in future releases:
  1. The aggregate status checks of digger/plan and digger/apply may not work as expected
  2. The automerge feature may not work as expected
this is because of how digger currently relies on single event to figure out if it includes all projects to perform the merge. A future refactor of the code will address these limitations.