Package com.linkedin.venice.schema.writecompute

Interface WriteComputeHandler

All Known Implementing Classes:

WriteComputeHandlerV1, WriteComputeHandlerV2

public interface WriteComputeHandler

This interface provides methods to execute write-compute operation a value

Method Summary

Modifier and Type	Method	Description
org.apache.avro.generic.GenericRecord	mergeUpdateRecord(org.apache.avro.Schema valueSchema, org.apache.avro.generic.GenericRecord currUpdateRecord, org.apache.avro.generic.GenericRecord newUpdateRecord)	Merge two partial update record that has the same update schema.
org.apache.avro.generic.GenericRecord	updateValueRecord(org.apache.avro.Schema valueSchema, org.apache.avro.generic.GenericRecord valueRecord, org.apache.avro.generic.GenericRecord writeComputeRecord)	Execute write-compute operation on a value record.

Method Details

updateValueRecord

org.apache.avro.generic.GenericRecord updateValueRecord(org.apache.avro.Schema valueSchema,
 org.apache.avro.generic.GenericRecord valueRecord,
 org.apache.avro.generic.GenericRecord writeComputeRecord)

Execute write-compute operation on a value record.

Parameters:

valueSchema - Schema of the value record.

valueRecord - Value record. Note that this object may be mutated in the implementation of this method and the returned object may be reference equal to this input object. When it is Optional.empty(), it means that there is currently no value.

writeComputeRecord - Write compute schema

Returns:

updated value record.

mergeUpdateRecord

org.apache.avro.generic.GenericRecord mergeUpdateRecord(org.apache.avro.Schema valueSchema,
 org.apache.avro.generic.GenericRecord currUpdateRecord,
 org.apache.avro.generic.GenericRecord newUpdateRecord)

Merge two partial update record that has the same update schema. Note that the semantic of the merged update record is equivalent to two partial update applied one by one with different timestamp for each field: (1) If any one of the field operation is NoOp, result will be the same as the other field operation. (2) If both field operation are set field, the latter one will win. (2) For collection field: -- If 2nd update is a set field, it will override the previous one; -- If the first one is set field and the 2nd one is collection merge, they will merge into a set field. This is to make sure the first set field semantic get applied correctly and fully override previous value. -- If both are collection merge, they will also merge into a collection merge. Note that: (1) If an element is removed in the first update and added in the latter update, it will still present in the merged collection merge operation. This API for now does not perform de-dupe if an element present in both union and diff operation. Venice Server will be able to handle this correctly regardless. (2) The expected priority for the two collection merge operation is: (a) current update's union op < (b) current update's diff op < (c) next update's union op < (d) next update's diff op. Within the same update, diff always has higher priority than union op, which is also exercised in Venice Backend conflict resolution logic. Also, note that this API may modify the content of input update records.

Parameters:

currUpdateRecord - Current update record. If existing update record is null, return the next update record.

newUpdateRecord - Next update record.

Returns:

Merged update record.