Define resources

Add resources to the hit counter construct

Now, let’s define the AWS Lambda function and the DynamoDB table in our HitCounter construct.

As usual, we first need to install the DynamoDB construct library (we already have the Lambda library installed):

pip install aws_cdk.aws_dynamodb

Now, go back to hello/ and add the following highlighted code:

from aws_cdk import (
    aws_lambda as _lambda,
    aws_dynamodb as ddb,

class HitCounter(core.Construct):

    def handler(self):
        return self._handler    

    def __init__(self, scope: core.Construct, id: str, downstream: _lambda.IFunction, **kwargs):
        super().__init__(scope, id, **kwargs)

        table = ddb.Table(
            self, 'Hits',
            partition_key={'name': 'path', 'type': ddb.AttributeType.STRING}

        self._handler = _lambda.Function(
            self, 'HitCountHandler',
                'DOWNSTREAM_FUNCTION_NAME': downstream.function_name,
                'HITS_TABLE_NAME': table.table_name,

What did we do here?

This code is hopefully quite easy to understand:

  • We defined a DynamoDB table with path as the partition key (every DynamoDB table must have a single partition key).
  • We defined a Lambda function which is bound to the lambda/hitcounter.handler code.
  • We wired the Lambda’s environment variables to the function_name and table_name of our resources.

Late-bound values

The function_name and table_name properties are values that only resolve when we deploy our stack (notice that we haven’t configured these physical names when we defined the table/function, only logical IDs). This means that if you print their values during synthesis, you will get a “TOKEN”, which is how the CDK represents these late-bound values. You should treat tokens as opaque strings. This means you can concatenate them together for example, but don’t be tempted to parse them in your code.