Use data stored as Avro without a column-level schema to construct an object container file. You have three possibilities for constructing an object container file in this scenario.
The Schemas are the Same, Without a Column-Level Schema
Simply retrieve a schema from an instance, and use this to construct the object container file, following the guidelines of Scenario 1.
The Schemas are Different, But Compatible
- Determine the desired schema used to describe all binary-encoded Avro values that compose the object container file.
- Use the AvroProject method of the DATASET type to construct Avro instances with the desired schema.
- Retrieve only the binary-encoded Avro values from the instances created in step 2 using the getRawData or getRawDataLob methods of the DATASET data type. If the length of these binary-encoded Avro values is also needed,use the getRawDataSize method.
Using the determined upon schema and the projected binary-encoded Avro values, you can construct an object container file using the Avro libraries.
The Schemas are Different, But Not Compatible
Each schema must be isolated, along with any binary-encoded Avro values described by that schema. For each isolated pairing, follow the steps in Scenario 1 to produce an object container file.