transform

pylibcudf.transform.bools_to_mask(

Column input,

stream=None,

DeviceMemoryResource mr=None,

) → tuple

Create a bitmask from a column of boolean elements

Parameters:

inputColumn

Column to produce new mask from.

streamStream | None

CUDA stream on which to perform the operation.

mrDeviceMemoryResource | None

Device memory resource used to allocate the returned mask’s device memory.

Returns:

tuple[gpumemoryview, int]

Two-tuple of a gpumemoryview wrapping the bitmask and the null count.

pylibcudf.transform.column_nans_to_nulls(

Column input,

stream=None,

DeviceMemoryResource mr=None,

) → Column

Create a column with nans converted to nulls.

For details, see column_nans_to_nulls().

Parameters:

inputColumn

Column to convert nans to nulls.

streamStream | None

CUDA stream on which to perform the operation.

mrDeviceMemoryResource | None

Device memory resource used to allocate the returned column’s device memory.

Returns:

Column

New column with nans converted to nulls.

pylibcudf.transform.compute_column(

Table input,

Expression expr,

stream=None,

DeviceMemoryResource mr=None,

) → Column

Create a column by evaluating an expression on a table.

For details see compute_column().

Parameters:

inputTable

Table used for expression evaluation

exprExpression

Expression to evaluate

streamStream | None

CUDA stream on which to perform the operation.

mrDeviceMemoryResource | None

Device memory resource used to allocate the returned column’s device memory.

Returns:

Column of the evaluated expression

pylibcudf.transform.compute_column_jit(

Table input,

Expression expr,

stream=None,

DeviceMemoryResource mr=None,

) → Column

Create a column by evaluating an expression on a table using a JIT-compiled kernel.

For details see compute_column_jit().

Parameters:

inputTable

Table used for expression evaluation

exprExpression

Expression to evaluate

streamStream | None

CUDA stream on which to perform the operation.

mrDeviceMemoryResource | None

Device memory resource used to allocate the returned column’s device memory.

Returns:

Column of the evaluated expression

pylibcudf.transform.encode(

Table input,

stream=None,

DeviceMemoryResource mr=None,

) → tuple

Encode the rows of the given table as integers.

Parameters:

inputTable

Table containing values to be encoded

streamStream | None

CUDA stream on which to perform the operation.

mrDeviceMemoryResource | None

Device memory resource used to allocate the returned objects’ device memory.

Returns:

tuple[Table, Column]

The distinct row of the input table in sorted order, and a column of integer indices representing the encoded rows.

pylibcudf.transform.mask_to_bools(

Py_ssize_t bitmask,

int begin_bit,

int end_bit,

stream=None,

DeviceMemoryResource mr=None,

) → Column

Creates a boolean column from given bitmask.

Parameters:

bitmaskint

Pointer to the bitmask which needs to be converted

begin_bitint

Position of the bit from which the conversion should start

end_bitint

Position of the bit before which the conversion should stop

streamStream | None

CUDA stream on which to perform the operation.

mrDeviceMemoryResource | None

Device memory resource used to allocate the returned column’s device memory.

Returns:

Column

Boolean column of the bitmask from [begin_bit, end_bit]

pylibcudf.transform.nans_to_nulls(

Column input,

stream=None,

DeviceMemoryResource mr=None,

) → tuple

Create a null mask preserving existing nulls and converting nans to null.

For details, see nans_to_nulls().

Parameters:

inputColumn

Column to produce new mask from.

streamStream | None

CUDA stream on which to perform the operation.

mrDeviceMemoryResource | None

Device memory resource used to allocate the returned mask’s device memory.

Returns:

Two-tuple of a gpumemoryview wrapping the null mask and the new null count.

pylibcudf.transform.one_hot_encode(

Column input,

Column categories,

stream=None,

DeviceMemoryResource mr=None,

) → Table

Encodes input by generating a new column for each value in categories indicating the presence of that value in input.

Parameters:

inputColumn

Column containing values to be encoded.

categoriesColumn

Column containing categories

streamStream | None

CUDA stream on which to perform the operation.

mrDeviceMemoryResource | None

Device memory resource used to allocate the returned table’s device memory.

Returns:

Column

A table of the encoded values.

pylibcudf.transform.transform(

inputs,

str transform_udf,

DataType output_type,

bool is_ptx,

null_aware is_null_aware,

output_nullability null_policy,

stream=None,

DeviceMemoryResource mr=None,

) → Column

Create a new column by applying a transform function against

multiple input columns.

Parameters:

inputsSequence[Column]

Columns to transform.

transform_udfstr

The PTX/CUDA string of the transform function to apply.

output_typeDataType

The output type that is compatible with the output type in the unary_udf.

is_ptxbool

If True, the UDF is treated as PTX code. If False, the UDF is treated as CUDA code.

is_null_aware: NullAware

If NO, the UDF gets non-nullable parameters If YES, the UDF gets nullable parameters

null_policy: OutputNullability

If PRESERVE, null-masks are produced if necessary. If ALL_VALID, null-masks are not produced. ALL_VALID has undefined behavior if the UDF can produce nulls.

streamStream | None

CUDA stream on which to perform the operation.

mrDeviceMemoryResource | None

Device memory resource used to allocate the returned column’s device memory.

Returns:

Column

The transformed column having the UDF applied to each element.