Fix reporting of row group size by parquet writer

Output of parquet-tools Before Row group 25: count: 2240892 26.32 B records start: 1479039171 total(compressed): 56.243 MB total(uncompressed):56.243 MB After Row group 25: count: 2244256 26.34 B records start: 1479178837 total(compressed): 56.370 MB total(uncompressed):167.418 MB
trinodb · May 11, 2023 · decd761 · decd761
1 parent a212bf1
commit decd761
Showing 1 changed file with 3 additions and 2 deletions.
diff --git a/lib/trino-parquet/src/main/java/io/trino/parquet/writer/ParquetWriter.java b/lib/trino-parquet/src/main/java/io/trino/parquet/writer/ParquetWriter.java
@@ -374,9 +374,10 @@ Slice getFooter(List<RowGroup> rowGroups, MessageType messageType)
 
     private void updateRowGroups(List<ColumnMetaData> columnMetaData)
     {
-        long totalBytes = columnMetaData.stream().mapToLong(ColumnMetaData::getTotal_compressed_size).sum();
+        long totalCompressedBytes = columnMetaData.stream().mapToLong(ColumnMetaData::getTotal_compressed_size).sum();
+        long totalBytes = columnMetaData.stream().mapToLong(ColumnMetaData::getTotal_uncompressed_size).sum();
         ImmutableList<org.apache.parquet.format.ColumnChunk> columnChunks = columnMetaData.stream().map(ParquetWriter::toColumnChunk).collect(toImmutableList());
-        rowGroupBuilder.add(new RowGroup(columnChunks, totalBytes, rows));
+        rowGroupBuilder.add(new RowGroup(columnChunks, totalBytes, rows).setTotal_compressed_size(totalCompressedBytes));
     }
 
     private static org.apache.parquet.format.ColumnChunk toColumnChunk(ColumnMetaData metaData)