Dominik Dębowczyk created FLINK-39106:
-----------------------------------------
Summary: Support ROW and ARRAY<ROW> data types in DynamoDB Table
API / SQL sink
Key: FLINK-39106
URL: https://issues.apache.org/jira/browse/FLINK-39106
Project: Flink
Issue Type: Improvement
Components: Connectors / DynamoDB
Affects Versions: 2.1.1, 2.2.0, 1.20.3, 2.0.1, 2.1.0, 1.20.2, 1.20.1,
1.20.0, 2.0.0
Reporter: Dominik Dębowczyk
h3. Motivation
The DynamoDB Table API / SQL sink ({{{}RowDataToAttributeValueConverter{}}})
does not support the Flink {{ROW}} data type. When a table schema includes a
{{ROW}} column, the converter falls through to
{{{}EnhancedType.of(dataType.getConversionClass()){}}}, which resolves to the
Flink internal RowData class. The DynamoDB Enhanced Client has no built-in
{{AttributeConverter}} for {{{}RowData{}}}, causing a runtime failure.
This prevents users from writing nested/structured data to DynamoDB via Flink
SQL. DynamoDB's Map (M) / document attribute type is a natural fit for Flink's
{{ROW}} type.
h3. Changes
ROW columns are now serialized as DynamoDB Map (M) attributes. {{ARRAY<ROW>}}
columns are serialized as DynamoDB List (L) of Maps. Recursive nesting ({{ROW}}
within {{ROW}}, {{ARRAY<ROW>}} within {{ROW}}) is fully supported.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
