Question: How do references to other components' output work?

[jedfoster]

This is a more general question about working with Serverless Components, not necessarily specific to this repo. If there's a better place to post this, let me know.

The API component in this repo references output from the permissions and database components. How does that work?

See api/serverless.yml lines 9, 17, and 19:

component: express
app: fullstack
name: api

inputs:
  # Express application source code.
  src: ./
  # Permissions required for the AWS Lambda function to interact with other resources
  roleName: ${output:permissions.name}  # <<--- _HOW_ does this work??

AFAICT, the only linkages between the API component and the permissions component are the app values, and directory structure (api/ and permissions/ being siblings).

I'm trying to add a layer to a lambda using those conventions—sibling directories, same app values—but I keep getting invalid reference ${output:commoncode.arnVersion}. What am I missing? How do these output references work?

[tommedema]

Same question here.

Similarly, if you run serverless deploy in the root folder, how does it know the dependency graph to define the order of operations in which to deploy? I.e. the database must be deployed prior to the api etc.

[eahefnawy]

The output variable format is like this:

${output:<instanceName>.<outputName>}

This would find the instance named instanceName in the same app & stage you're deploying to. So in your case, the instance name is permissions. You can confirm that by looking at the name property of the yaml file inside the permissions folder.

@tommedema the CLI collects all the yaml files in child directories and does some basic static analysis, to figure out the output references.

[tommedema]

@eahefnawy does it then do a graph analysis to ensure that dependencies are deployed prior to dependents etc.?