Nested type support was added to parquet writer in #7230 and #6075 but the data is always plain encoded and not dictionary encoded. This needs significant changes to the dictionary encoding code in parquet writer, which currently hardcodes the hash table size and assumes that the number of rows == number of values.