danny0405 commented on code in PR #13381:
URL: https://github.com/apache/hudi/pull/13381#discussion_r2119844459


##########
rfc/rfc-95/rfc-95.md:
##########
@@ -0,0 +1,263 @@
+<!--
+  Licensed to the Apache Software Foundation (ASF) under one or more
+  contributor license agreements.  See the NOTICE file distributed with
+  this work for additional information regarding copyright ownership.
+  The ASF licenses this file to You under the Apache License, Version 2.0
+  (the "License"); you may not use this file except in compliance with
+  the License.  You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+  Unless required by applicable law or agreed to in writing, software
+  distributed under the License is distributed on an "AS IS" BASIS,
+  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+  See the License for the specific language governing permissions and
+  limitations under the License.
+-->
+# RFC-95: Hudi Flink Source Implementation
+
+## Proposers
+
+- HuangZhenQiu
+
+## Approvers
+ - Danny Chan
+
+## Status
+
+JIRA: [HUDI-9483](https://issues.apache.org/jira/browse/HUDI-9483)
+
+## Abstract
+This RFC proposes support for new Flink source API for Flink hudi source. Hudi 
currently supports reading data via Flink Source Function APIs. However, it 
lacks a native Flink
+Source API implementation for consuming Hudi tables as a first-class Flink 
source. The proposal aims to fill that gap by implementing a Flink Source that 
adheres to Flink's Source API
+(introduced in Flink 1.11 and enhanced in 1.13+) to enable efficient, scalable 
and consistent reading of Hudi dataset in both streaming and batch modes.
+
+## Background
+Flink FLIP-27 solve several problems/shortcomings in the streaming source 
interface (SourceFunction) and simultaneously to unify the source interfaces 
between the batch and streaming APIs.
+By adopting the new source interface, Flink can read hudi data in batch mode 
for better processing efficiency of bounded data. It also unblocks Flink Hudi 
user to use Flink hybrid source to seamless
+switch reading data from hudi to kafka for back-fill use cases.
+
+## Implementation
+In the FLIP-27, the new source api split the reading logic into two major 
parts, SplitEnumerator, Reader. In the proposal, the implementation of native 
FlinkHudiSource using Flink's unified Source API 
+will be discussed. It supports both snapshot and incremental modes, integrate 
with Hudi's metadata and file indexing, provide fault-tolerant Flink sources. 
For the Flink SQL user, the split order and 
+watermark emitter for each split is very important for the correctness of 
result of window aggregation and window join. Thus, watermark emit and 
event-time processing will also be discussed below.
+
+![Hudi Flink Source](./hudi_flink_source.png)
+
+### Hudi Source
+
+```java
+public class HudiSource implements Source<RowData, HudiSourceSplit, 
HudiEnumeratorState> {
+  
+  //...
+  @Override
+  public Boundedness getBoundedness() {
+    return scanContext.isStreaming() ? Boundedness.CONTINUOUS_UNBOUNDED : 
Boundedness.BOUNDED;
+  }
+
+  @Override
+  public SourceReader<T, HudiSourceSplit> createReader(SourceReaderContext 
readerContext) {
+    HudiSourceReaderMetrics metrics =
+        new HudiSourceReaderMetrics(readerContext.metricGroup(), tableName);
+    return new HudiSourceReader<>(
+        emitter, metrics, readerFunction, splitComparator, readerContext);
+  }
+
+  @Override
+  public SplitEnumerator<HudiSourceSplit, HudiEnumeratorState> 
createEnumerator(
+      SplitEnumeratorContext<HudiSourceSplit> enumContext) {
+    return createEnumerator(enumContext, null);
+  }
+
+  @Override
+  public SplitEnumerator<HudiSourceSplit, HudiEnumeratorState> 
restoreEnumerator(
+      SplitEnumeratorContext<HudiSourceSplit> enumContext, HudiEnumeratorState 
enumState) {
+    return createEnumerator(enumContext, enumState);
+  }
+
+  @Override
+  public SimpleVersionedSerializer<HudiSourceSplit> getSplitSerializer() {
+    return new HudiSourceSplitSerializer(scanContext.caseSensitive());
+  }
+
+  @Override
+  public SimpleVersionedSerializer<HudiEnumeratorState> 
getEnumeratorCheckpointSerializer() {
+    return new HudiEnumeratorStateSerializer(scanContext.caseSensitive());
+  }
+  
+  // ...
+}
+```
+
+### Hudi Split
+Similar to existing MergeOnReadInputSplit that implements InputSplit API, A 
new MergeOnReadSourceSplit and CdcSourceSplit will be implemented with the new 
SourceSplit interface.
+FlinkRowDataReaderContext will be extended to be able to apply HudiSourceSplit 
and return a CloseableIterator that iterates RecordsWithSplitIds. Base on the 
change, the HudiSourceSplitReader

Review Comment:
   Can you elaborate this `RecordsWithSplitIds`?



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to