SAM: use 'environmentVariables' without adding them to the template

[cyrfer]

Describe the bug

Developers need to be able to specify environment variables for debugging Lambda functions.

I tried specifying those values in lambda.environmentVariables (seen below) with no luck. My workaround was to add them to the template.yaml, but that is not very flexible or natural when working with VS Code debug configurations.

All documentation I can find leaves lambda.environmentVariables an empty object. For example, here.

The toolkit generated this .vscode/launch.json for me:

{
  "configurations": [
    {
      "type": "aws-sam",
      "request": "direct-invoke",
      "name": "lambda-nodejs:Function",
      "invokeTarget": {
        "target": "template",
        "templatePath": "${workspaceFolder}/template.yaml",
        "logicalId": "Function"
      },
      "lambda": {
        "payload": {"path": "src/test/events/my.json"},
        "environmentVariables": {
          "key": "value"
        },
      }
    }
  ]
}

To Reproduce

Steps to reproduce the behavior:

1. use the toolkit to add a debug configuration for lambda functions
2. specify environment variables with custom keys
3. read the environment variable key in your code (nodejs for me)
4. notice the key gives no value.

Expected behavior

I expect the debugging context to allow code to read from process.env.key where key is any key specified in the debug configuration lambda.environmentVariables.

Desktop (please complete the following information):

• OS: Darwin x64 19.6.0
• Visual Studio Code Extension Host Version: 1.60.1
• AWS Toolkit Version: 1.29.0

[JadenSimon]

Seems like SAM CLI is not using the environment variables we pass in:
No environment variables found for function 'HelloWorldFunction'
Looking into it.

[JadenSimon]

Currently the values in lambda.environmentVariables are treated as overrides to environment variables specified in the template. The documentation is not clear on that, with the only reference being this snippet (found here):

When specifying environment variables for a serverless application, you must add configurations to both the AWS SAM template (template.yaml) and the launch.json file.

Which again does not elaborate as to why this is needed. What this means is that configuring lambda.environmentVariables does work as long as you place dummy values in the template.

My workaround was to add them to the template.yaml, but that is not very flexible or natural when working with VS Code debug configurations.

I agree. Intuitively, configuring environment variables in the launch config should just be added on top of the template, regardless of the existence of the variable in the template. This may be a quick fix on our end, but I'll need to do some more research to make sure we don't break any use-cases (if any) that relied on this behavior.