Regarding pushdown filter predicate

[ankit11519]

We are using apache spark to connect to Dynamodb using emr-dynamodb-connector.
Now, below are my code statements in pyspark:--

dynamoDf = spark.read.option('region', 'REGION')
.option("tableName", "TABLE_NAME")
.format("dynamodb")
.load()
dynamoDfFilter = dynamoDf.filter((F.col("colFilter").startswith('ABC')) | (F.col("colFilter").startswith('XYZ')))
print(dynamoDfFilter.count())

So, wanted to know if these filter conditions will also push down to the DynamoDB for server side filtering or will they be applied after full scan operation data being loaded into "dynamoDf"?

[kevnzhao]

Which connector are you using?
Predicate Pushdown is only available for Hive table in EMR DDB Hive Connector package. You can find the supported data types and operators at HERE.

@mimaomao feel free to comment if I miss anything.

[ankit11519]

We are using this connector:--
[Accessing data in Amazon DynamoDB with Apache Spark]

So, we are accessing amazon DynamoDB with apache spark for scan operation. My question is if query is like this :--

dynamoDfFilter = dynamoDf.filter((F.col("colFilter").startswith('ABC')) | (F.col("colFilter").startswith('XYZ')))

print(dynamoDfFilter.count())

When spark 'filter' is being applied in scan query, whether query which is being sent to dynamoDB has 'filterExpressions' associated with it for server side filtering Or it load entire data in scan operation in a dataFrame first and then apply the filter over that dataFrame.

[kevnzhao]

Then you are using Hadoop Connector. Predicate push-down is not available yet. So all data is loaded from DynamoDB.