Variables in Workflows

When running locally, variables are shared in-memory between targets. In CI/CD workflows, targets may run in separate jobs on different machines. Atom bridges this gap by mapping variables to the platform's native output/variable mechanism.

How It Works

1. A target declares .ProducesVariable("BuildVersion") and writes it with WriteVariable.
2. A downstream target declares .ConsumesVariable(nameof(SetupBuildInfo), "BuildVersion").
3. Locally: the variable is stored in-memory and read directly.
4. In CI: the WorkflowResolver detects the dependency, creates a job boundary if needed, and the platform module maps the variable to:
   • GitHub Actions: job outputs (${{ needs.job.outputs.var }})
   • Azure DevOps: pipeline variables

Platform Variable Providers

Each platform module registers an IVariableProvider:

Module	Provider	Mechanism
DecSm.Atom.Module.GithubWorkflows	GithubVariableProvider	Writes to $GITHUB_OUTPUT, reads from job outputs
DecSm.Atom.Module.DevopsWorkflows	DevopsVariableProvider	Uses Azure DevOps pipeline variables

The built-in AtomVariableProvider handles local execution.

Example

Target SetupVersion => t => t
    .ProducesVariable("BuildVersion")
    .Executes(async ct =>
    {
        await WriteVariable("BuildVersion", "1.2.3", ct);
    });

Target Pack => t => t
    .ConsumesVariable(nameof(SetupVersion), "BuildVersion")
    .DependsOn(SetupVersion)
    .Executes(async ct =>
    {
        var version = await ReadVariable(nameof(SetupVersion), "BuildVersion", ct);
        // use version
    });

In the generated YAML, SetupVersion and Pack will be in separate jobs with the variable passed as a job output.

Next Steps

→ Debugging Workflows