The following code writes three records to DynamoDB. Create another file called dynamo_modify_items.py with the following Python code:

from boto3 import resource


class DynamoRepository:
    def __init__(self, target_dynamo_table, region='eu-west-1'):
        self.dynamodb = resource(service_name='dynamodb', 
                        region_name=region)
        self.target_dynamo_table = target_dynamo_table
        self.table = self.dynamodb.Table(self.target_dynamo_table)

    def update_dynamo_event_counter(self, event_name, 
                event_datetime, event_count=1):
        return self.table.update_item(
            Key={
                'EventId': event_name,
                'EventDay': event_datetime
            },
            ExpressionAttributeValues={":eventCount": event_count},
            UpdateExpression="ADD EventCount :eventCount")


def main():
    table_name = 'user-visits'
    dynamo_repo = DynamoRepository(table_name)
    print(dynamo_repo.update_dynamo_event_counter('324', 20171001))
    print(dynamo_repo.update_dynamo_event_counter('324', 20171001, 2))
    print(dynamo_repo.update_dynamo_event_counter('324', 20171002, 5))


if __name__ == '__main__':
    main()

Here, we use Boto3's resource(), which is a higher-level service resource with the repository pattern. We abstract all the DynamoDB-related code in the DynamoRepository() class that instantiates as dynamo_repo with table_name. self.dynamodb.Table() creates a table resource based on table_name. That will be used later on when calling update_dynamo_event_counter() to update DynamoDB records.

In self.table.update_item(), I first declare a variable called eventCount using ExpressionAttributeValues. I'm using this in the DynamoDB advanced Update Expressions (https://docs.aws.amazon.com/amazondynamodb/latest/developerguide/Expressions.UpdateExpressions.html), which is one of my favorite features in DynamoDB. Why? Because not all NoSQL databases can do something similar without having something like a semaphore lock and having the clients do a retry. It performs the following three actions in one atomic statement, while circumventing possible concurrency violations at the cost of eventual consistency:

1. Reads records matching the given EventId=event_name and EventDay=event_datetime
2. Creates a new item if it doesn't exist, setting EventCount=1
3. If it does already exist, then it increments EventCount by event_count

The first function calls dynamo_repo.update_dynamo_event_counter('324', 20171001), sets EventCount to 1; the second function call, dynamo_repo.update_dynamo_event_counter('324', 20171001, 2), increments EventCount by 2, so that it's now 3. The third function call adds a new record, as the EventCount or primary key, is different.