Add terraform output variables to environment

[emjacobs]

Child function for GetEnvVars that runs terraform output and adds each output to the list of environment variables, has another child function for formatting the json returned by the terraform output command

[rmvangun]

For command/exec specifically, the shell package is used when executing commands. This approach abstracts execution so that in the future, we could execute commands against other types of shells--docker exec, SSH, etc.

You can find examples of this elsewhere, but it will end up being something like shell.ExecSilent("terraform", "-chdir=...")

[rmvangun]

This would be shimmed for test coverage

[rmvangun]

Very close!

Other than a couple small things, the key piece is that, if you're in a terraform project folder, the TF_VAR_* that should be injected would be from the previous Terraform modules. In the blueprint.yaml, that would mean, for something like:

terraform:
  - path: projectA
  - path: projectB

If an output.tf existed in the first folder, those outputs would be available on the second. So, you might create an example project, like so:

terraform/
├─ projectA/
│  ├─ main.tf
│  ├─ variables.tf
│  └─ output.tf
└─ projectB/
   ├─ main.tf
   ├─ variables.tf
   └─ output.tf

if projectA/output.tf is:

output "some_output" {
  value = some_resource.this.some_value
}

Then when you are in the terraform/projectB folder, running windsor env would include an output like:

...
export TF_VAR_some_output=some-value
...

[rmvangun]

Implemented in #1478