Sharing Data in Terramate

Maintaining DRY (Don't Repeat Yourself) principles in your code is crucial to keep your project neat and manageable. Terramate facilitates this practice by enabling you to define data once and distribute it throughout your project. This can be accomplished using two main constructs: Globals and Metadata.

Globals are user-defined entities, similar to locals in Terraform, whereas Metadata is information supplied by Terramate itself. These are integrated with Terraform through a code generation process. To delve deeper into code generation process, read here.

Lazy Evaluation in Terramate

Given that globals can reference other globals and Terramate metadata, it is important to be clear about how/when evaluation happens.

Globals are lazily evaluated, which means globals block can reference variables that will exist at a later point in time. For example, a non-stack directory can have globals referencing stack metadata.

Below is a high-level overview of how globals evaluation is implemented for a given stack:

• The globals blocks are loaded from all Terramate files in the stack directory.
• The multiple globals blocks are merged into a single globals definition.
• Recursively, the parent directories have their globals blocks loaded and merged into the same single globals definition but giving preference for the global defined lower in the filesystem tree.
• Then resulting single globals block have its attributes evaluated.
  • The terramate.* namespace is set with the stack Metadata values.
  • Globals that depend on other globals are postponed until it's dependencies are evaluated.

This means that globals can reference globals on other configurations independent of how specific or general the configuration is since it is all merged together into a single globals set before evaluation.