Fix aws_serverless config when api folder missing

[bennettrogers]

Fixes #1550

Note: The ternary in the template string isn't necessarily the cleanest looking solution, but it seems to be the simplest.

[github-actions]

📦 Packages for this PR can be downloaded from
https://rw-pr-redwoodjs-com.s3.amazonaws.com/1553/create-redwood-app-0.21.0-ec9b5cf.tgz
https://rw-pr-redwoodjs-com.s3.amazonaws.com/1553/redwoodjs-api-0.21.0-ec9b5cf.tgz
https://rw-pr-redwoodjs-com.s3.amazonaws.com/1553/redwoodjs-api-server-0.21.0-ec9b5cf.tgz
https://rw-pr-redwoodjs-com.s3.amazonaws.com/1553/redwoodjs-auth-0.21.0-ec9b5cf.tgz
https://rw-pr-redwoodjs-com.s3.amazonaws.com/1553/redwoodjs-cli-0.21.0-ec9b5cf.tgz
https://rw-pr-redwoodjs-com.s3.amazonaws.com/1553/redwoodjs-core-0.21.0-ec9b5cf.tgz
https://rw-pr-redwoodjs-com.s3.amazonaws.com/1553/redwoodjs-dev-server-0.21.0-ec9b5cf.tgz
https://rw-pr-redwoodjs-com.s3.amazonaws.com/1553/redwoodjs-eslint-config-0.21.0-ec9b5cf.tgz
https://rw-pr-redwoodjs-com.s3.amazonaws.com/1553/redwoodjs-eslint-plugin-redwood-0.21.0-ec9b5cf.tgz
https://rw-pr-redwoodjs-com.s3.amazonaws.com/1553/redwoodjs-forms-0.21.0-ec9b5cf.tgz
https://rw-pr-redwoodjs-com.s3.amazonaws.com/1553/redwoodjs-internal-0.21.0-ec9b5cf.tgz
https://rw-pr-redwoodjs-com.s3.amazonaws.com/1553/redwoodjs-router-0.21.0-ec9b5cf.tgz
https://rw-pr-redwoodjs-com.s3.amazonaws.com/1553/redwoodjs-structure-0.21.0-ec9b5cf.tgz
https://rw-pr-redwoodjs-com.s3.amazonaws.com/1553/redwoodjs-testing-0.21.0-ec9b5cf.tgz
https://rw-pr-redwoodjs-com.s3.amazonaws.com/1553/redwoodjs-web-0.21.0-ec9b5cf.tgz

Install this PR by running yarn rw upgrade --pr 1553:0.21.0-ec9b5cf

[Tobbe]

Thanks for taking the time to submit this PR. Have you tried it, so it still works to deploy to AWS?

[bennettrogers]

I did try to run the Redwood AWS deploy generator, but that fails due to the api workspace being missing in the first place.

I did not try manually writing an AWS deploy config and triggering that, nor did I try generating an AWS deploy config before deleting the api workspace and then triggering the deploy. It seems like neither of those would work or make sense without the existence of an api side.

[Tobbe]

I didn't realize we only supported deploying the api side to AWS. I thought we deployed the web side to an s3 bucket as well... If it's just the API side, then, yeah, not much to test if there is no api side 😀

[Tobbe]

I generated aws deploy config (with the api stuff in place) and renamed the file to serverless.yml_. Then I installed the PR packages for this PR (need to be on latest canary release for that to work) and ran the generator again.

Want to tweak that, and then I think this is ready to merge!

Just to be clear, it's indented now, and it shouldn't be. This is what it looks with this PR installed (note line 32)

[bennettrogers]

I did notice that it was indenting that line, but forgot that it would be a problem since the result is a YAML file...

What's the best way to fix this? If I just try to move the functions: bit to its own line, the configured Prettier setup just puts it right back where it is. I guess // prettier-ignore would work? Is there a better way?

[Tobbe]

What's the best way to fix this? If I just try to move the functions: bit to its own line, the configured Prettier setup just puts it right back where it is. I guess // prettier-ignore would work? Is there a better way?

This seems to work for me, no prettier-ignore needed

package:
  individually: true

${
  fs.existsSync(path.resolve(getPaths().api.functions))
    ? `functions:
  ${fs
    .readdirSync(path.resolve(getPaths().api.functions))
    .map((file) => {

[Tobbe]

Thanks for the fix Bennett!