Flink: Dynamic Iceberg Sink: Add HashKeyGenerator / RowDataEvolver / TableUpdateOperator #13277
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…TableUpdateOperator
This change adds the following components for the Flink Dynamic Iceberg Sink:
*** HashKeyGenerator
A hash key generator which will be used in DynamicIcebergSink class (next PR) to
implement one of Iceberg's DistributionModes (NONE, HASH, RANGE).
The HashKeyGenerator is responsible for creating the appropriate hash key for Flink's keyBy
operation. The hash key is generated depending on the user-provided DynamicRecord and the table
metadata. Under the hood, we maintain a set of Flink {@link KeySelector}s which implement the
appropriate Iceberg {@link DistributionMode}. For every table, we randomly select a consistent
subset of writer subtasks which receive data via their associated keys, depending on the chosen
DistributionMode.
Caching ensures that a new key selector is also created when the table metadata (e.g. schema,
spec) or the user-provided metadata changes (e.g. distribution mode, write
parallelism).
*** RowDataEvolver
RowDataEvolver is responsible to change the input RowData to make it compatible
with the target schema. This is done when:
1. The input schema has fewer fields than the target schema.
2. The table types are wider than the input type.
3. The field order differs for source and target schema.
The resolution is as follows:
In the first case, we would add a null values for the missing field (if the field is optional).
In the second case, we would convert the data for the input field to a wider type, e.g. int (input type) => long (table type).
In the third case, we would rearrange the input data to match the target table.
*** DynamicUpdateOperator
A dedicated operator to updating the schema / spec for the table associated with
a DynamicRecord.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This change adds the following components for the Flink Dynamic Iceberg Sink:
HashKeyGenerator
A hash key generator which will be used in DynamicIcebergSink class (next PR) to implement one of Iceberg's DistributionModes (NONE, HASH, RANGE).
The HashKeyGenerator is responsible for creating the appropriate hash key for Flink's keyBy operation. The hash key is generated depending on the user-provided DynamicRecord and the table metadata. Under the hood, we maintain a set of Flink KeySelectors which implement the appropriate Iceberg DistributionMode. For every table, we randomly select a consistent subset of writer subtasks which receive data via their associated keys, depending on the chosen DistributionMode.
Caching ensures that a new key selector is also created when the table metadata (e.g. schema, spec) or the user-provided metadata changes (e.g. distribution mode, write parallelism).
RowDataEvolver
RowDataEvolver is responsible to change the input RowData to make it compatible with the target schema. This is done when:
The resolution is as follows:
In the first case, we would add a null values for the missing field (if the field is optional). In the second case, we would convert the data for the input field to a wider type, e.g. int (input type) => long (table type). In the third case, we would rearrange the input data to match the target table.
DynamicUpdateOperator
A dedicated operator to updating the schema / spec for the table associated with a DynamicRecord.