Parquet's 'float' type is actually a 32 bit float and its 'double' type is 64 bit.
If you write a parquet out from Panads, Athena will refuse to read the file if you set the datatype to flat, because it's a double within the parquet file.
Your query has the following error(s):
HIVE_BAD_DATA: Field X's type DOUBLE in parquet is incompatible with type float defined in table schema
Parquet's 'float' type is actually a 32 bit float and its 'double' type is 64 bit.
If you write a parquet out from Panads, Athena will refuse to read the file if you set the datatype to flat, because it's a double within the parquet file.
Parquet file formats are enumerated here.
https://drill.apache.org/docs/parquet-format/
and Sparks are here:
http://spark.apache.org/docs/2.2.1/api/python/_modules/pyspark/sql/types.html#FloatType