jonvex commented on code in PR #10957:
URL: https://github.com/apache/hudi/pull/10957#discussion_r1571157403
##########
hudi-spark-datasource/hudi-spark-common/src/main/scala/org/apache/spark/sql/execution/datasources/parquet/HoodieFileGroupReaderBasedParquetFileFormat.scala:
##########
@@ -129,20 +138,15 @@ class
HoodieFileGroupReaderBasedParquetFileFormat(tableState: HoodieTableState,
file.partitionValues match {
// Snapshot or incremental queries.
case fileSliceMapping: HoodiePartitionFileSliceMapping =>
- val filePath =
sparkAdapter.getSparkPartitionedFileUtils.getPathFromPartitionedFile(file)
- val filegroupName = if (FSUtils.isLogFile(filePath)) {
- FSUtils.getFileId(filePath.getName).substring(1)
- } else {
- FSUtils.getFileId(filePath.getName)
- }
+ val filegroupName = FSUtils.getFileIdFromFilePath(sparkAdapter
+ .getSparkPartitionedFileUtils.getPathFromPartitionedFile(file))
fileSliceMapping.getSlice(filegroupName) match {
case Some(fileSlice) if !isCount =>
if (requiredSchema.isEmpty &&
!fileSlice.getLogFiles.findAny().isPresent) {
val hoodieBaseFile = fileSlice.getBaseFile.get()
baseFileReader(createPartitionedFile(fileSliceMapping.getPartitionValues,
hoodieBaseFile.getHadoopPath, 0, hoodieBaseFile.getFileLen))
} else {
- val readerContext: HoodieReaderContext[InternalRow] = new
SparkFileFormatInternalRowReaderContext(
- readerMaps)
+ val readerContext = new
SparkFileFormatInternalRowReaderContext(parquetFileReader.value,
tableState.recordKeyField, filters)
Review Comment:
The intention is to not have duplicate logic for creating the readers and
then passing in a map of already created readers. Now, we create the readers on
the executor.
Before we called buildReaderWithPartitionValues() a few times and had a map
from schema hash to PartitionedFile => Iterator[InternalRow]
Now we have a reader that we call read on the executor and can pass in the
schema and filters that we want with the file.
We have removed the limitation that schema and filters be known in the driver
--
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]
