Extension Dtypes#

cuDF supports a number of extension dtypes that build on top of the types that pandas supports. These dtypes are not directly available in pandas, which instead relies on object dtype arrays that run at Python rather than native speeds. The following dtypes are supported:

cudf.CategoricalDtype#

CategoricalDtype([categories, ordered])

Type for categorical data with the categories and orderedness.

Properties and Methods#

`CategoricalDtype.categories`	An `Index` containing the unique categories allowed.
`CategoricalDtype.construct_from_string`()	Construct this type from a string.
`CategoricalDtype.deserialize`(header, frames)	Generate an object from a serialized representation.
`CategoricalDtype.device_deserialize`(header, ...)	Perform device-side deserialization tasks.
`CategoricalDtype.device_serialize`()	Serialize data and metadata associated with device memory.
`CategoricalDtype.from_pandas`(dtype)	Convert a `pandas.CategrocialDtype` to `cudf.CategoricalDtype`
`CategoricalDtype.host_deserialize`(header, frames)	Perform device-side deserialization tasks.
`CategoricalDtype.host_serialize`()	Serialize data and metadata associated with host memory.
`CategoricalDtype.is_dtype`(dtype)	Check if we match 'dtype'.
`CategoricalDtype.name`	A string identifying the data type.
`CategoricalDtype.ordered`	Whether the categories have an ordered relationship.
`CategoricalDtype.serialize`()	Generate an equivalent serializable representation of an object.
`CategoricalDtype.str`
`CategoricalDtype.to_pandas`()	Convert a `cudf.CategoricalDtype` to `pandas.CategoricalDtype`
`CategoricalDtype.type`	The scalar type for the array, e.g.

cudf.Decimal32Dtype#

Decimal32Dtype(precision[, scale])

Type to represent a decimal32 data.

Properties and Methods#

`Decimal32Dtype.ITEMSIZE`
`Decimal32Dtype.MAX_PRECISION`
`Decimal32Dtype.deserialize`(header, frames)	Generate an object from a serialized representation.
`Decimal32Dtype.device_deserialize`(header, frames)	Perform device-side deserialization tasks.
`Decimal32Dtype.device_serialize`()	Serialize data and metadata associated with device memory.
`Decimal32Dtype.from_arrow`(typ)	Construct a cudf decimal dtype from a `pyarrow` dtype
`Decimal32Dtype.host_deserialize`(header, frames)	Perform device-side deserialization tasks.
`Decimal32Dtype.host_serialize`()	Serialize data and metadata associated with host memory.
`Decimal32Dtype.is_dtype`(dtype)	Check if we match 'dtype'.
`Decimal32Dtype.itemsize`	Length of one column element in bytes.
`Decimal32Dtype.precision`	The decimal precision, in number of decimal digits (an integer).
`Decimal32Dtype.scale`	The decimal scale (an integer).
`Decimal32Dtype.serialize`()	Generate an equivalent serializable representation of an object.
`Decimal32Dtype.str`
`Decimal32Dtype.to_arrow`()	Return the equivalent `pyarrow` dtype.

cudf.Decimal64Dtype#

Decimal64Dtype(precision[, scale])

Type to represent a decimal64 data.

Properties and Methods#

`Decimal64Dtype.ITEMSIZE`
`Decimal64Dtype.MAX_PRECISION`
`Decimal64Dtype.deserialize`(header, frames)	Generate an object from a serialized representation.
`Decimal64Dtype.device_deserialize`(header, frames)	Perform device-side deserialization tasks.
`Decimal64Dtype.device_serialize`()	Serialize data and metadata associated with device memory.
`Decimal64Dtype.from_arrow`(typ)	Construct a cudf decimal dtype from a `pyarrow` dtype
`Decimal64Dtype.host_deserialize`(header, frames)	Perform device-side deserialization tasks.
`Decimal64Dtype.host_serialize`()	Serialize data and metadata associated with host memory.
`Decimal64Dtype.is_dtype`(dtype)	Check if we match 'dtype'.
`Decimal64Dtype.itemsize`	Length of one column element in bytes.
`Decimal64Dtype.precision`	The decimal precision, in number of decimal digits (an integer).
`Decimal64Dtype.scale`	The decimal scale (an integer).
`Decimal64Dtype.serialize`()	Generate an equivalent serializable representation of an object.
`Decimal64Dtype.str`
`Decimal64Dtype.to_arrow`()	Return the equivalent `pyarrow` dtype.

cudf.Decimal128Dtype#

Decimal128Dtype(precision[, scale])

Type to represent a decimal128 data.

Properties and Methods#

`Decimal128Dtype.ITEMSIZE`
`Decimal128Dtype.MAX_PRECISION`
`Decimal128Dtype.deserialize`(header, frames)	Generate an object from a serialized representation.
`Decimal128Dtype.device_deserialize`(header, ...)	Perform device-side deserialization tasks.
`Decimal128Dtype.device_serialize`()	Serialize data and metadata associated with device memory.
`Decimal128Dtype.from_arrow`(typ)	Construct a cudf decimal dtype from a `pyarrow` dtype
`Decimal128Dtype.host_deserialize`(header, frames)	Perform device-side deserialization tasks.
`Decimal128Dtype.host_serialize`()	Serialize data and metadata associated with host memory.
`Decimal128Dtype.is_dtype`(dtype)	Check if we match 'dtype'.
`Decimal128Dtype.itemsize`	Length of one column element in bytes.
`Decimal128Dtype.precision`	The decimal precision, in number of decimal digits (an integer).
`Decimal128Dtype.scale`	The decimal scale (an integer).
`Decimal128Dtype.serialize`()	Generate an equivalent serializable representation of an object.
`Decimal128Dtype.str`
`Decimal128Dtype.to_arrow`()	Return the equivalent `pyarrow` dtype.

cudf.ListDtype#

ListDtype(element_type)

Type to represent list data.

Properties and Methods#

`ListDtype.deserialize`(header, frames)	Generate an object from a serialized representation.
`ListDtype.device_deserialize`(header, frames)	Perform device-side deserialization tasks.
`ListDtype.device_serialize`()	Serialize data and metadata associated with device memory.
`ListDtype.element_type`	Returns the element type of the `ListDtype`.
`ListDtype.from_arrow`(typ)	Creates a `ListDtype` from `pyarrow.ListType`.
`ListDtype.host_deserialize`(header, frames)	Perform device-side deserialization tasks.
`ListDtype.host_serialize`()	Serialize data and metadata associated with host memory.
`ListDtype.is_dtype`(dtype)	Check if we match 'dtype'.
`ListDtype.leaf_type`	Returns the type of the leaf values.
`ListDtype.serialize`()	Generate an equivalent serializable representation of an object.
`ListDtype.to_arrow`()	Convert to a `pyarrow.ListType`
`ListDtype.type`	The scalar type for the array, e.g.

cudf.StructDtype#

StructDtype(fields)

Type to represent a struct data.

Properties and Methods#

`StructDtype.deserialize`(header, frames)	Generate an object from a serialized representation.
`StructDtype.device_deserialize`(header, frames)	Perform device-side deserialization tasks.
`StructDtype.device_serialize`()	Serialize data and metadata associated with device memory.
`StructDtype.fields`	Returns an ordered dict of column name and dtype key-value.
`StructDtype.from_arrow`(typ)	Convert a `pyarrow.StructType` to `StructDtype`.
`StructDtype.host_deserialize`(header, frames)	Perform device-side deserialization tasks.
`StructDtype.host_serialize`()	Serialize data and metadata associated with host memory.
`StructDtype.is_dtype`(dtype)	Check if we match 'dtype'.
`StructDtype.serialize`()	Generate an equivalent serializable representation of an object.
`StructDtype.to_arrow`()	Convert a `StructDtype` to a `pyarrow.StructType`.
`StructDtype.type`	The scalar type for the array, e.g.