voonhous commented on PR #13927:
URL: https://github.com/apache/hudi/pull/13927#issuecomment-3325155186
For reader properties, I have verified that my current fix allows the
following below to work:
```
spark.sql("SET hoodie.hfile*")
```
The mechanism in which wiring for reader configs are done is as such:
When a reading a table, the entrypoint for file index is
`org.apache.hudi.HoodieFileIndex`.
The session properties are available in:
```
options: Map[String, String]
```
And it is used to build `configProperties`:
```
configProperties = getConfigProperties(spark, options,
metaClient.getTableConfig),
configProperties and options will live in the attribute space of
HoodieFileIndex .
```
`HoodieFileIndex` extends `SparkHoodieTableFileIndex` extends
`BaseHoodieTableFileIndex`.
In `BaseHoodieTableFileIndex` , it initializes `metadataConfig` using the
code below in the constructor using `configProperties`:
```
this.metadataConfig = HoodieMetadataConfig.newBuilder()
.fromProperties(configProperties)
.enable(configProperties.getBoolean(ENABLE.key(),
DEFAULT_METADATA_ENABLE_FOR_READERS)
&& HoodieTableMetadataUtil.isFilesPartitionAvailable(metaClient))
.build();
```
So, session properties are **injected** into the `metadataConfig` within the
`HoodieFileIndex` scope.
`HoodieTableMetadata` is created as such in `BaseHoodieTableFileIndex`,
where `metadataConfig` is stored as an attribute in its scope.
```
HoodieTableMetadata newTableMetadata = metadataFactory.create(
engineContext, storage, metadataConfig, basePath.toString(), true);
```
From here on, it's quite straight forward, `HoodieAvroReaderContext` is
initialized like this:
```java
TypedProperties props = buildFileGroupReaderProperties(metadataConfig);
HoodieReaderContext<IndexedRecord> readerContext = new
HoodieAvroReaderContext(
storageConf,
metadataMetaClient.getTableConfig(),
instantRange,
Option.of(predicate),
baseFileReaders,
props);
HoodieFileGroupReader<IndexedRecord> fileGroupReader =
HoodieFileGroupReader.<IndexedRecord>newBuilder()
.withReaderContext(readerContext)
.withHoodieTableMetaClient(metadataMetaClient)
.withLatestCommitTime(latestMetadataInstantTime)
.withFileSlice(fileSlice)
.withDataSchema(SCHEMA)
.withRequestedSchema(SCHEMA)
.withProps(props)
.withRecordBufferLoader(recordBufferLoader)
.withEnableOptimizedLogBlockScan(metadataConfig.isOptimizedLogBlocksScanEnabled())
.build();
```
Also, proof that session properties are injected with some manual testing
when reading **files** and **partition**:
<img width="3804" height="1682" alt="image"
src="https://github.com/user-attachments/assets/a88ab478-4cb3-4721-aa33-cc777eb1a2cc";
/>
<img width="4024" height="1970" alt="image"
src="https://github.com/user-attachments/assets/00aee61a-e64f-4bbc-beea-ab1ae57de9f1";
/>
--
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]
