Type Mapping

Scala	Avro	BigQuery	Bigtable7	Datastore	Parquet	Protobuf	TensorFlow
Unit	null	x	x	Null	x	x	x
Boolean	boolean	BOOL	Byte	Boolean	BOOLEAN	Boolean	INT643
Char	int3	INT643	Char	Integer3	INT323	Int3	INT643
Byte	int3	INT643	Byte	Integer3	INT329	Int3	INT643
Short	int3	INT643	Short	Integer3	INT329	Int3	INT643
Int	int	INT643	Int	Integer3	INT329	Int	INT643
Long	long	INT64	Long	Integer	INT649	Long	INT64
Float	float	FLOAT643	Float	Double3	FLOAT	Float	FLOAT
Double	double	FLOAT64	Double	Double	DOUBLE	Double	FLOAT3
CharSequence	string	x	x	x	x	x	x
String	string	STRING	String	String	BINARY	String	BYTES3
Array[Byte]	bytes	BYTES	ByteString	Blob	BINARY	ByteString	BYTES
ByteString	x	x	ByteString	Blob	x	ByteString	BYTES
ByteBuffer	bytes	x	x		x	x	x
Enum1	enum	STRING3	String	String3	BINARY/ENUM9	Enum	BYTES3
BigInt	x	x	BigInt	x	x	x	x
BigDecimal	bytes4	NUMERIC6	Int scale + unscaled BigInt	x	LOGICAL[DECIMAL]9,14	x	x
Option[T]	union[null, T]5	NULLABLE	Empty as None	Absent as None	OPTIONAL	optional10	Size <= 1
Iterable[T]2	array[T]	REPEATED	x	Array	REPEATED13	repeated	Size >= 0
Nested	record	STRUCT	Flat8	Entity	Group	Message	Flat8
Map[K, V]	map[V]15	x	x	x	x	map<K, V>	x
java.time.Instant	long11	TIMESTAMP	x	Timestamp	LOGICAL[TIMESTAMP]9	x	x
java.time.LocalDateTime	long11	DATETIME	x	x	LOGICAL[TIMESTAMP]9	x	x
java.time.OffsetTime	x	x	x	x	LOGICAL[TIME]9	x	x
java.time.LocalTime	long11	TIME	x	x	LOGICAL[TIME]9	x	x
java.time.LocalDate	int11	DATE	x	x	LOGICAL[DATE]9	x	x
org.joda.time.LocalDate	int11	x	x	x	x	x	x
org.joda.time.DateTime	int11	x	x	x	x	x	x
org.joda.time.LocalTime	int11	x	x	x	x	x	x
java.util.UUID	string4	x	ByteString (16 bytes)	x	FIXED[16]	x	x
(Long, Long, Long)12	fixed[12]	x	x	x	x	x	x

1. Those wrapped inUnsafeEnum are encoded as strings, see enums.md for more
2. Any subtype of Iterable[T]
3. Unsafe conversions, import magnolify.$MODULE.unsafe._
4. Avro logical types (doc)
5. UNION of [NULL, T] and defaults to NULL (doc)
6. Fixed precision of 38 and scale of 9 (doc)
7. All Scala types are encoded as big endian ByteString for Bigtable
8. Nested fields are encoded flat with field names joined with ., e.g. level1.level2.level3
9. More information on Parquet logical type schemas can be found here. Time types are available at multiple precisions; import magnolify.parquet.logical.micros._, magnolify.avro.logical.millis._, or magnolify.avro.logical.nanos._ accordingly.
10. See protobuf.md for more
11. Logical types available at micro- or milli-second precision; import magnolify.avro.logical.micros._ or magnolify.avro.logical.millis._ accordingly. BigQuery-compatible conversions are available in magnolify.avro.logical.bigquery._.
12. Special tuple used to represent Duration in the Avro spec. This has not been made implicit in Magnolify; import AvroType.afDuration implicitly to enable
13. If magnolify.parquet.ParquetArray.AvroCompat._ is imported, array fields use the nested, Avro-compatible schema format: required group $FIELDNAME (LIST) { repeated $FIELDTYPE array ($FIELDSCHEMA); }.
14. Parquet’s Decimal logical format supports multiple representations, and are not implicitly scoped by default. Import one of: magnolify.parquet.ParquetField.{decimal32, decimal64, decimalFixed, decimalBinary}.
15. Map key type in avro is fixed to string. Scala Map key type must be either String or CharSequence.