Amazon Lambda Functions with the event-filtering for DynamoDB Streams enables ease of integration and application development.
There are a couple of good blog posts with talks about how to setup triggers with filter patterns to your Lambda fns.
To know about what is DynamoDB Streams and it's use-cases -
{%post aws-builders/using-dynamodb-streams-and-lambda-in-your-serverless-applications-3epl%}
To know about what is the announcement of Lambda functions Event-filtering -
{%post aws-builders/trigger-lambda-functions-with-event-filtering-2pnb%}
To know about DynamoDB Streams which are helpful with event-filtering -
{%post aws-builders/new-dynamodb-streams-filtering-in-serverless-framework-3lc5%}
{%post nosqlknowhow/aws-lambda-event-filter-of-amazon-dynamodb-streams-16pc%}
{%post aws-builders/filtering-dynamodb-streams-before-lambda-p5p%}
Key takeaways
In this blog post, we will deep dive into the feature itself and we will learn about -
How the filtering works
The filter expression or the filter pattern JSON uses a strict JSON format which is used to match the filtering criteria. For a DynamoDB Stream based event, the JSON is validated against the key dynamodb which is part of the event JSON.
{
"Records": [
{
...
"dynamodb": { }
...
}
]
}
Whenever an action on DynamoDB putItem or updateItem or deleteItem occurs, the INSERT or MODIFY or REMOVE events are triggered respectively if DynamoDB Streams are enabled. This helps handling the needed business logic for specific actions. Now with event filtering with patterns are available, this performs additional validation before the Lambda function is invoked.
The flow diagram explains, whenever there is a DynamoDB Stream invoked, it checks if the target Lambda fn has any event filtering pattern defined or not. If the pattern is available, event JSON is validated against it and if the event JSON matches the criteria then the Lambda fn gets invoked.
If nothing is matched, the event is discarded. If the pattern itself is not defined, the Lambda function gets invoked as event-filter patterns are optional.
Event which is triggered whenever a DynamoDB INSERT operation happens is -
{
"Records": [
{
"eventID": "42a3c355ea58b7c68a96368be0f8fa68",
"eventName": "INSERT",
"eventVersion": "1.1",
"eventSource": "aws:dynamodb",
"awsRegion": "us-east-1",
"dynamodb": {
"ApproximateCreationDateTime": 1638968744,
"Keys": {
"sk": {
"S": "CAR"
},
"pk": {
"S": "USA3271395584644fordgalaxie5001972-01-01"
}
},
"NewImage": {
"miles_per_gallon": {
"N": "14"
},
"acceleration": {
"N": "13"
},
"car_name": {
"S": "Ford Galaxie 500"
},
"horsepower": {
"N": "153"
},
"year": {
"S": "1972-02-01"
},
"origin": {
"S": "USA"
},
"sk": {
"S": "CAR"
},
"displacement": {
"N": "351"
},
"pk": {
"S": "USA3271395584644fordgalaxie5001972-01-01"
},
"cylinders": {
"N": "8"
},
"weight_in_lbs": {
"N": "4129"
}
},
"SequenceNumber": "530563900000000048422878623",
"SizeBytes": 228,
"StreamViewType": "NEW_AND_OLD_IMAGES"
},
"eventSourceARN": "arn:aws:dynamodb:us-east-1:xxxxxxx:table/cars-demo/stream/2021-11-28T18:05:10.127"
}
]
}
Filter patterns supported
The event filtering uses JSON based patterns which are capable with the following criteria checks -
| Criteria | Example | Syntax |
| Null | origin is null | "origin": [ null ] |
| Empty | origin is empty | "origin": [""] |
| Equals | origin is "USA" | "origin": ["USA"] |
| And | origin is "USA" and cylinders is 8 | "origin": ["USA"], "cylinders": ["8"] |
| Or | origin is "USA" or "Germany" | "origin": ["USA","Germany"] |
| Not | origin is not Japan | "origin": { "anything-but": [ "Japan" ] } |
| Exists | displacement exists | "displacement ": [ { "exists": true } ] |
| Does not exists | displacement does not exists | "displacement ": [ { "exists": false} ] |
| Begins with | pk begins with "USA" | "pk": [ {"prefix": "USA" } ] |
Limitations with event-filtering
Use-cases and it's associated filter-patterns
Whenever you would want the Lambda function to be invoked only if -
- Have atleast one cylinder.
{
"filters": [
{
"pattern": "{\"dynamodb\" : \"{ \"NewImage\" : \"cylinder\" : \"N\" : { \"anything-but\": [ \"0\" ] } }\" }"
}
]
}
- A new car is added to the DynamoDB table.
{
"filters": [
{
"pattern": "{\"eventName\" : [\"INSERT\"], \"dynamodb\" : \"{ \"Key\" : \"sk\" : \"S\" : [\"CAR\"] }\" }"
}
]
}
- A car from USA origin is modified.
{
"filters": [
{
"pattern": "{\"eventName\" : [\"MODIFY\"], \"dynamodb\" : \"{ \"NewImage\" : \"origin\" : \"S\" : [\"USA\"] }\" }"
}
]
}
- A car name beginning with "Ford" is deleted.
{
"filters": [
{
"pattern": "{\"eventName\" : [\"REMOVE\"], \"dynamodb\" : \"{ \"OldImage\" : \"car_name\" : \"S\" : [{\"prefix\": \"Ford\" }] }\" }"
}
]
}
- Either if a new car is added or removed.
{
"filters": [
{
"pattern": "{\"eventName\" : [\"INSERT\",\"REMOVE\"], \"dynamodb\" : \"{ \"Key\" : \"sk\" : \"S\" : [\"CAR\"] }\" }"
}
]
}
- Either if a new car is added or if the any car with year 1971 is added / deleted / updated.
{
"filters": [
{
"pattern": "{\"eventName\" : [\"INSERT\"]}"
},
{
"pattern": "{\"dynamodb\" : \"{ \"OldImage\" : \"year\" : \"S\" : [{\"prefix\": \"1971\" }] }\" }"
}
]
}
Conclusion
The event-filtering with DynamoDB Streams helps in filtering out and focusing only on the pattern of data which needs the Lambda function to be invoked. This helps in efficient and optimized Lambda function code as the additional code to validate various conditions. This also adheres to the new Well Architected Model's new pillar Sustainability pillar which ensures there is less carbon foot-print because of limited Lambda function executions.
