parthchandra commented on code in PR #1103:
URL: https://github.com/apache/datafusion-comet/pull/1103#discussion_r1849161016
##########
spark/src/main/scala/org/apache/comet/serde/QueryPlanSerde.scala:
##########
@@ -2507,23 +2508,22 @@ object QueryPlanSerde extends Logging with
ShimQueryPlanSerde with CometExprShim
partitions.foreach(p => {
val inputPartitions =
p.asInstanceOf[DataSourceRDDPartition].inputPartitions
inputPartitions.foreach(partition => {
- partition2Proto(partition.asInstanceOf[FilePartition],
nativeScanBuilder)
+ partition2Proto(partition.asInstanceOf[FilePartition],
nativeScanBuilder, scan)
})
})
case rdd: FileScanRDD =>
rdd.filePartitions.foreach(partition => {
- partition2Proto(partition, nativeScanBuilder)
+ partition2Proto(partition, nativeScanBuilder, scan)
})
case _ =>
+ assert(false)
}
- val requiredSchemaParquet =
- new
SparkToParquetSchemaConverter(conf).convert(scan.requiredSchema)
- val dataSchemaParquet =
- new
SparkToParquetSchemaConverter(conf).convert(scan.relation.dataSchema)
+ val projection_vector: Array[java.lang.Long] =
scan.requiredSchema.fields.map(field => {
Review Comment:
This change essentially means that any schema 'adaptation' made in
`SparkToParquetSchemaConverter.convert` to support legacy timestamps and
decimals will not be supported. But we will probably fail tests with incorrect
results.
Also, Comet's Parquet file reader uses
`CometParquetReadSupport.clipParquetSchema` to do similar conversion and it
includes support for Parquet
[field_id](https://github.com/apache/parquet-format/blob/c70281359087dfaee8bd43bed9748675f4aabe11/src/main/thrift/parquet.thrift#L473)
which is desirable for delta sources like Iceberg.
Basically a field_id, if present, identifies a field more precisely (in the
event of field name changes) in a schema.
##########
spark/src/main/scala/org/apache/comet/serde/QueryPlanSerde.scala:
##########
@@ -3191,9 +3191,25 @@ object QueryPlanSerde extends Logging with
ShimQueryPlanSerde with CometExprShim
private def partition2Proto(
partition: FilePartition,
- nativeScanBuilder: OperatorOuterClass.NativeScan.Builder): Unit = {
+ nativeScanBuilder: OperatorOuterClass.NativeScan.Builder,
+ scan: CometScanExec): Unit = {
val partitionBuilder = OperatorOuterClass.SparkFilePartition.newBuilder()
+ val sparkContext = scan.session.sparkContext
+ var schema_saved: Boolean = false;
partition.files.foreach(file => {
+ if (!schema_saved) {
+ // TODO: This code shouldn't be here, but for POC it's fine.
+ // Extract the schema and stash it.
+ val hadoopConf =
+
scan.relation.sparkSession.sessionState.newHadoopConfWithOptions(scan.relation.options)
+ val broadcastedHadoopConf =
+ sparkContext.broadcast(new SerializableConfiguration(hadoopConf))
+ val sharedConf = broadcastedHadoopConf.value.value
+ val footer = FooterReader.readFooter(sharedConf, file)
Review Comment:
You're right. This can never be in production code. For one, this is
expensive.
--
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
