In 0.1 you could do something like table.variants_table().to_dataframe().write.mode(‘overwrite’).parquet() but I dont see anything similar in the new version.
Is the underlying data structure different between the versions? 0.1 specifics its stored as parquet under the hood but 0.2 seems to refere to Hail table format.
0.2 does have a new native format for MatrixTable and Table. It is massively faster for us to read, and doesn’t suffer from critical limitations in Parquet with respect to table width.
You can still write Parquet by converting to a Spark dataframe – mt.rows().to_spark()