[deniskuzZ]

What changes were proposed in this pull request?

• Request no column for a projection with no column, in both DataWritableReadSupport helpers (vectorized getRequestedSchema, init)
• The whole-file-schema fallback was added by HIVE-10975 as a workaround for Parquet 1.8.0 rejecting an empty schema ("ParquetReadSupport doesn't support empty field to construct a group type") and copied into the vectorized helper by HIVE-17696. PARQUET-363 (1.8.2, 1.9.0) allows an empty requested schema for row counting; Hive is on 1.18.1. RCFile, ORC and LLAP already read no column for an empty projection

Why are the changes needed?

• count(*) and partition-only reads read every column of the file and decode none of them. On LLAP the empty projection carried the table's nested columns, so count(*) declined the native path to the fallback reader (parquet_llap_counters.q: HDFS_BYTES_READ 18,616 -> 0, ROWS_EMITTED 2098)
• Bug-Fix: a vectorized read of only row lineage columns failed with IndexOutOfBoundsException: Index 2 out of bounds for length 2 in ParquetRowGroupDecoder.buildColumnReaders. The lineage columns were appended to the whole file schema but to an empty colsToInclude, and the two are paired by position

Does this PR introduce any user-facing change?

No

How was this patch tested?

• TestDataWritableReadSupport#testNoProjectedColumnRequestsNoColumn (both helpers)
• parquet_llap_counters.q: count(*) with no projected column on the LLAP native path
• row_lineage.q: a read of only row lineage columns; fails with the IndexOutOfBoundsException on master
• Row mode: the existing parquet_partitioned.q (SELECT part, COUNT(0) ... GROUP BY part, vectorization off) now reads with an empty schema

[Aggarwal-Raghav]

LGTM +1, pending tests

Before Patch, Tested the query on local setup and after its working ✅

Caused by: java.lang.IndexOutOfBoundsException: Index 2 out of bounds for length 2
	at java.base/jdk.internal.util.Preconditions.outOfBounds(Preconditions.java:100)
	at java.base/jdk.internal.util.Preconditions.outOfBoundsCheckIndex(Preconditions.java:106)
	at java.base/jdk.internal.util.Preconditions.checkIndex(Preconditions.java:302)
	at java.base/java.util.Objects.checkIndex(Objects.java:385)
	at java.base/java.util.ArrayList.get(ArrayList.java:427)
	at org.apache.hadoop.hive.ql.io.parquet.vector.VectorizedParquetRecordReader.checkEndOfRowGroup(VectorizedParquetRecordReader.java:462)
	at org.apache.hadoop.hive.ql.io.parquet.vector.VectorizedParquetRecordReader.nextBatch(VectorizedParquetRecordReader.java:421)
	at org.apache.hadoop.hive.ql.io.parquet.vector.VectorizedParquetRecordReader.next(VectorizedParquetRecordReader.java:359)
	at org.apache.hadoop.hive.ql.io.parquet.vector.VectorizedParquetRecordReader.next(VectorizedParquetRecordReader.java:99)
	at org.apache.iceberg.mr.hive.vector.HiveBatchIterator.advance(HiveBatchIterator.java:92)
	at org.apache.iceberg.mr.hive.vector.HiveBatchIterator.hasNext(HiveBatchIterator.java:200)
	at org.apache.iceberg.mr.mapreduce.IcebergRecordReader.nextKeyValue(IcebergRecordReader.java:98)
	at org.apache.iceberg.mr.hive.vector.HiveIcebergVectorizedRecordReader.next(HiveIcebergVectorizedRecordReader.java:48)
	at org.apache.iceberg.mr.hive.vector.HiveIcebergVectorizedRecordReader.next(HiveIcebergVectorizedRecordReader.java:34)

[sonarqubecloud]

Quality Gate passed

Issues
6 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud