AvroIndexedParquetRDD is an RDD which reads an index file, and (for a given CombinedFilter, which contains a predicate on the entries of the index) has as many partitions as there are unique row groups within the indexed Parquet files whose index entries satisfy the predicate.
ParquetLister materializes the records within a Parquet row group as an Iterator[T].
ParquetPartition is a Partition implementation, used in our RDD implementations, which corresponds to a single row group in a single Parquet file.
This is a reimplementation of the Type / GroupType / PrimitiveType / MessageType hierarchy from parquet.
PartitionPageReadStore is a holder for mapping between the ParquetColumnDescriptor values and the PageReader values.