Replacing

Files
file	cudf/replace.hpp

Enumerations
enum class	cudf::replace_policy : bool { PRECEDING , FOLLOWING }
	Policy to specify the position of replacement values relative to null rows. More...

Functions
std::unique_ptr< column >	cudf::replace_nulls (column_view const &input, column_view const &replacement, rmm::cuda_stream_view stream=cudf::get_default_stream(), rmm::device_async_resource_ref mr=rmm::mr::get_current_device_resource())
	Replaces all null values in a column with corresponding values of another column. More...
std::unique_ptr< column >	cudf::replace_nulls (column_view const &input, scalar const &replacement, rmm::cuda_stream_view stream=cudf::get_default_stream(), rmm::device_async_resource_ref mr=rmm::mr::get_current_device_resource())
	Replaces all null values in a column with a scalar. More...
std::unique_ptr< column >	cudf::replace_nulls (column_view const &input, replace_policy const &replace_policy, rmm::cuda_stream_view stream=cudf::get_default_stream(), rmm::device_async_resource_ref mr=rmm::mr::get_current_device_resource())
	Replaces all null values in a column with the first non-null value that precedes/follows. More...
std::unique_ptr< column >	cudf::replace_nans (column_view const &input, column_view const &replacement, rmm::cuda_stream_view stream=cudf::get_default_stream(), rmm::device_async_resource_ref mr=rmm::mr::get_current_device_resource())
	Replaces all NaN values in a column with corresponding values from another column. More...
std::unique_ptr< column >	cudf::replace_nans (column_view const &input, scalar const &replacement, rmm::cuda_stream_view stream=cudf::get_default_stream(), rmm::device_async_resource_ref mr=rmm::mr::get_current_device_resource())
	Replaces all NaN values in a column with a scalar. More...
std::unique_ptr< column >	cudf::find_and_replace_all (column_view const &input_col, column_view const &values_to_replace, column_view const &replacement_values, rmm::cuda_stream_view stream=cudf::get_default_stream(), rmm::device_async_resource_ref mr=rmm::mr::get_current_device_resource())
	Return a copy of input_col replacing any values_to_replace[i] found with replacement_values[i]. More...
std::unique_ptr< column >	cudf::clamp (column_view const &input, scalar const &lo, scalar const &lo_replace, scalar const &hi, scalar const &hi_replace, rmm::cuda_stream_view stream=cudf::get_default_stream(), rmm::device_async_resource_ref mr=rmm::mr::get_current_device_resource())
	Replaces values less than lo in input with lo_replace, and values greater than hi with hi_replace. More...
std::unique_ptr< column >	cudf::clamp (column_view const &input, scalar const &lo, scalar const &hi, rmm::cuda_stream_view stream=cudf::get_default_stream(), rmm::device_async_resource_ref mr=rmm::mr::get_current_device_resource())
	Replaces values less than lo in input with lo, and values greater than hi with hi. More...
std::unique_ptr< column >	cudf::normalize_nans_and_zeros (column_view const &input, rmm::cuda_stream_view stream=cudf::get_default_stream(), rmm::device_async_resource_ref mr=rmm::mr::get_current_device_resource())
	Copies from a column of floating-point elements and replaces -NaN and -0.0 with +NaN and +0.0, respectively. More...
void	cudf::normalize_nans_and_zeros (mutable_column_view &in_out, rmm::cuda_stream_view stream=cudf::get_default_stream())
	Modifies a column of floating-point elements to replace all -NaN and -0.0 with +NaN and +0.0, respectively. More...

Detailed Description

Enumeration Type Documentation

replace_policy

enum cudf::replace_policy : bool

strong

Policy to specify the position of replacement values relative to null rows.

PRECEDING means the replacement value is the first non-null value preceding the null row. FOLLOWING means the replacement value is the first non-null value following the null row.

Definition at line 41 of file cudf/replace.hpp.

Function Documentation

clamp() [1/2]

std::unique_ptr<column> cudf::clamp	(	column_view const &	input,
		scalar const &	lo,
		scalar const &	hi,
		rmm::cuda_stream_view	stream = cudf::get_default_stream(),
		rmm::device_async_resource_ref	mr = rmm::mr::get_current_device_resource()
	)

Replaces values less than lo in input with lo, and values greater than hi with hi.

if lo is invalid, then lo will not be considered while evaluating the input (Essentially considered minimum value of that type). if hi is invalid, then hi will not be considered while evaluating the input (Essentially considered maximum value of that type).

Example:
   input: {1, 2, 3, NULL, 5, 6, 7}
 
   valid lo and hi
   lo: 3, hi: 5
   output:{3, 3, 3, NULL, 5, 5, 5}
 
   invalid lo
   lo: NULL, hi:5
   output:{1, 2, 3, NULL, 5, 5, 5}
 
   invalid hi
   lo: 3, hi:NULL
   output:{3, 3, 3, NULL, 5, 6, 7}

Exceptions

cudf::logic_error	if lo.type() != hi.type()
cudf::logic_error	if lo.type() != input.type()

Parameters

[in]	input	Column whose elements will be clamped
[in]	lo	Minimum clamp value. All elements less than lo will be replaced by lo Ignored if null.
[in]	hi	Maximum clamp value. All elements greater than hi will be replaced by hi Ignored if null.
	stream	CUDA stream used for device memory operations and kernel launches
[in]	mr	Device memory resource used to allocate device memory of the returned column

Returns

Returns a clamped column as per lo and hi boundaries

clamp() [2/2]

std::unique_ptr<column> cudf::clamp	(	column_view const &	input,
		scalar const &	lo,
		scalar const &	lo_replace,
		scalar const &	hi,
		scalar const &	hi_replace,
		rmm::cuda_stream_view	stream = cudf::get_default_stream(),
		rmm::device_async_resource_ref	mr = rmm::mr::get_current_device_resource()
	)

Replaces values less than lo in input with lo_replace, and values greater than hi with hi_replace.

if lo is invalid, then lo will not be considered while evaluating the input (Essentially considered minimum value of that type). if hi is invalid, then hi will not be considered while evaluating the input (Essentially considered maximum value of that type).

Note

: If lo is valid then lo_replace should be valid If hi is valid then hi_replace should be valid

Example:
   input: {1, 2, 3, NULL, 5, 6, 7}
 
   valid lo and hi
   lo: 3, hi: 5, lo_replace : 0, hi_replace : 16
   output:{0, 0, 3, NULL, 5, 16, 16}
 
   invalid lo
   lo: NULL, hi: 5, lo_replace : 0, hi_replace : 16
   output:{1, 2, 3, NULL, 5, 16, 16}
 
   invalid hi
   lo: 3, hi: NULL, lo_replace : 0, hi_replace : 16
   output:{0, 0, 3, NULL, 5, 6, 7}

Exceptions

cudf::logic_error	if lo.type() != hi.type()
cudf::logic_error	if lo_replace.type() != hi_replace.type()
cudf::logic_error	if lo.type() != lo_replace.type()
cudf::logic_error	if lo.type() != input.type()

Parameters

[in]	input	Column whose elements will be clamped
[in]	lo	Minimum clamp value. All elements less than lo will be replaced by lo_replace Ignored if null.
[in]	lo_replace	All elements less than lo will be replaced by lo_replace
[in]	hi	Maximum clamp value. All elements greater than hi will be replaced by hi_replace. Ignored if null.
[in]	hi_replace	All elements greater than hi will be replaced by hi_replace
	stream	CUDA stream used for device memory operations and kernel launches
[in]	mr	Device memory resource used to allocate device memory of the returned column

Returns

Returns a clamped column as per lo and hi boundaries

find_and_replace_all()

std::unique_ptr<column> cudf::find_and_replace_all	(	column_view const &	input_col,
		column_view const &	values_to_replace,
		column_view const &	replacement_values,
		rmm::cuda_stream_view	stream = cudf::get_default_stream(),
		rmm::device_async_resource_ref	mr = rmm::mr::get_current_device_resource()
	)

Return a copy of input_col replacing any values_to_replace[i] found with replacement_values[i].

Parameters

input_col	The column to find and replace values in
values_to_replace	The values to replace
replacement_values	The values to replace with
stream	CUDA stream used for device memory operations and kernel launches
mr	Device memory resource used to allocate the returned column's device memory

Returns

Copy of input_col with specified values replaced

normalize_nans_and_zeros() [1/2]

std::unique_ptr<column> cudf::normalize_nans_and_zeros	(	column_view const &	input,
		rmm::cuda_stream_view	stream = cudf::get_default_stream(),
		rmm::device_async_resource_ref	mr = rmm::mr::get_current_device_resource()
	)

Copies from a column of floating-point elements and replaces -NaN and -0.0 with +NaN and +0.0, respectively.

Converts floating point values from input using the following rules: Convert -NaN -> NaN Convert -0.0 -> 0.0

Exceptions

cudf::logic_error

if column does not have floating point data type.

Parameters

[in]	input	column_view of floating-point elements to copy and normalize
	stream	CUDA stream used for device memory operations and kernel launches
[in]	mr	device_memory_resource allocator for allocating output data

Returns

new column with the modified data

normalize_nans_and_zeros() [2/2]

void cudf::normalize_nans_and_zeros	(	mutable_column_view &	in_out,
		rmm::cuda_stream_view	stream = cudf::get_default_stream()
	)

Modifies a column of floating-point elements to replace all -NaN and -0.0 with +NaN and +0.0, respectively.

Converts floating point values from in_out using the following rules: Convert -NaN -> NaN Convert -0.0 -> 0.0

Exceptions

cudf::logic_error

if column does not have floating point data type.

Parameters

[in,out]	in_out	of floating-point elements to normalize
	stream	CUDA stream used for device memory operations and kernel launches

replace_nans() [1/2]

std::unique_ptr<column> cudf::replace_nans	(	column_view const &	input,
		column_view const &	replacement,
		rmm::cuda_stream_view	stream = cudf::get_default_stream(),
		rmm::device_async_resource_ref	mr = rmm::mr::get_current_device_resource()
	)

Replaces all NaN values in a column with corresponding values from another column.

If input[i] is NaN, then output[i] will contain replacement[i].

input        = {1.0, NaN, 4.0}
replacement  = {3.0, 9.0, 7.0}
output       = {1.0, 9.0, 4.0}

Note

Nulls are not considered as NaN

Exceptions

cudf::logic_error	If input and replacement are of different type or size.
cudf::logic_error	If input or replacement are not of floating-point dtype.

Parameters

input	A column whose NaN values will be replaced
replacement	A cudf::column whose values will replace NaN values in input
stream	CUDA stream used for device memory operations and kernel launches
mr	Device memory resource used to allocate the returned column's device memory

Returns

A copy of input with the NaN values replaced with corresponding values from replacement.

replace_nans() [2/2]

std::unique_ptr<column> cudf::replace_nans	(	column_view const &	input,
		scalar const &	replacement,
		rmm::cuda_stream_view	stream = cudf::get_default_stream(),
		rmm::device_async_resource_ref	mr = rmm::mr::get_current_device_resource()
	)

Replaces all NaN values in a column with a scalar.

If input[i] is NaN, then output[i] will contain replacement.

input        = {1.0, NaN, 4.0}
replacement  = 7.0
output       = {1.0, 7.0, 4.0}

Note

Nulls are not considered as NaN

Exceptions

cudf::logic_error	If input and replacement are of different type.
cudf::logic_error	If input or replacement are not of floating-point dtype.

Parameters

input	A column whose NaN values will be replaced
replacement	A cudf::scalar whose value will replace NaN values in input
stream	CUDA stream used for device memory operations and kernel launches
mr	Device memory resource used to allocate the returned column's device memory

Returns

A copy of input with the NaN values replaced by replacement

replace_nulls() [1/3]

std::unique_ptr<column> cudf::replace_nulls	(	column_view const &	input,
		column_view const &	replacement,
		rmm::cuda_stream_view	stream = cudf::get_default_stream(),
		rmm::device_async_resource_ref	mr = rmm::mr::get_current_device_resource()
	)

Replaces all null values in a column with corresponding values of another column.

If input[i] is NULL, then output[i] will contain replacement[i]. input and replacement must be of the same type and size.

Parameters

[in]	input	A column whose null values will be replaced
[in]	replacement	A cudf::column whose values will replace null values in input
	stream	CUDA stream used for device memory operations and kernel launches
[in]	mr	Device memory resource used to allocate device memory of the returned column

Returns

A copy of input with the null values replaced with corresponding values from replacement.

replace_nulls() [2/3]

std::unique_ptr<column> cudf::replace_nulls	(	column_view const &	input,
		replace_policy const &	replace_policy,
		rmm::cuda_stream_view	stream = cudf::get_default_stream(),
		rmm::device_async_resource_ref	mr = rmm::mr::get_current_device_resource()
	)

Replaces all null values in a column with the first non-null value that precedes/follows.

If input[i] is NULL, then output[i] will contain the first non-null value that precedes or follows the null value, based on replace_policy.

Parameters

[in]	input	A column whose null values will be replaced
[in]	replace_policy	Specify the position of replacement values relative to null values
	stream	CUDA stream used for device memory operations and kernel launches
[in]	mr	Device memory resource used to allocate device memory of the returned column

Returns

Copy of input with null values replaced based on replace_policy

replace_nulls() [3/3]

std::unique_ptr<column> cudf::replace_nulls	(	column_view const &	input,
		scalar const &	replacement,
		rmm::cuda_stream_view	stream = cudf::get_default_stream(),
		rmm::device_async_resource_ref	mr = rmm::mr::get_current_device_resource()
	)

Replaces all null values in a column with a scalar.

If input[i] is NULL, then output[i] will contain replacement. input and replacement must have the same type.

Parameters

[in]	input	A column whose null values will be replaced
[in]	replacement	Scalar used to replace null values in input
	stream	CUDA stream used for device memory operations and kernel launches
[in]	mr	Device memory resource used to allocate device memory of the returned column

Returns

Copy of input with null values replaced by replacement