Concatenating

Files
file	concatenate.hpp
	Concatenate columns APIs.

Functions
rmm::device_buffer	cudf::concatenate_masks (host_span< column_view const > views, rmm::cuda_stream_view stream=cudf::get_default_stream(), rmm::device_async_resource_ref mr=cudf::get_current_device_resource_ref())
	Concatenates views[i]'s bitmask from the bits [views[i].offset(), views[i].offset() + views[i].size()) for all elements views into an rmm::device_buffer More...
std::unique_ptr< column >	cudf::concatenate (host_span< column_view const > columns_to_concat, rmm::cuda_stream_view stream=cudf::get_default_stream(), rmm::device_async_resource_ref mr=cudf::get_current_device_resource_ref())
	Concatenates multiple columns into a single column. More...
std::unique_ptr< table >	cudf::concatenate (host_span< table_view const > tables_to_concat, rmm::cuda_stream_view stream=cudf::get_default_stream(), rmm::device_async_resource_ref mr=cudf::get_current_device_resource_ref())
	Columns of tables_to_concat are concatenated vertically to return a single table. More...

Detailed Description

Function Documentation

concatenate() [1/2]

std::unique_ptr<column> cudf::concatenate	(	host_span< column_view const >	columns_to_concat,
		rmm::cuda_stream_view	stream = cudf::get_default_stream(),
		rmm::device_async_resource_ref	mr = cudf::get_current_device_resource_ref()
	)

Concatenates multiple columns into a single column.

Exceptions

cudf::logic_error	If types of the input columns mismatch
std::overflow_error	If the total number of output rows exceeds cudf::size_type

Parameters

columns_to_concat	Column views to be concatenated into a single column
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 single column having all the rows from the elements of columns_to_concat respectively in the same order.

concatenate() [2/2]

std::unique_ptr<table> cudf::concatenate	(	host_span< table_view const >	tables_to_concat,
		rmm::cuda_stream_view	stream = cudf::get_default_stream(),
		rmm::device_async_resource_ref	mr = cudf::get_current_device_resource_ref()
	)

Columns of tables_to_concat are concatenated vertically to return a single table.

column_view c0 is {0,1,2,3}
column_view c1 is {4,5,6,7}
table_view t0{{c0, c0}};
table_view t1{{c1, c1}};
...
auto t = concatenate({t0.view(), t1.view()});
column_view tc0 = (t->view()).column(0) is {0,1,2,3,4,5,6,7}
column_view tc1 = (t->view()).column(1) is {0,1,2,3,4,5,6,7}

Exceptions

cudf::logic_error	If number of columns mismatch
std::overflow_error	If the total number of output rows exceeds cudf::size_type

Parameters

tables_to_concat	Table views to be concatenated into a single table
stream	CUDA stream used for device memory operations and kernel launches
mr	Device memory resource used to allocate the returned table's device memory

Returns

A single table having all the rows from the elements of tables_to_concat respectively in the same order.

concatenate_masks()

rmm::device_buffer cudf::concatenate_masks	(	host_span< column_view const >	views,
		rmm::cuda_stream_view	stream = cudf::get_default_stream(),
		rmm::device_async_resource_ref	mr = cudf::get_current_device_resource_ref()
	)

Concatenates views[i]'s bitmask from the bits [views[i].offset(), views[i].offset() + views[i].size()) for all elements views into an rmm::device_buffer

Returns an empty buffer if the column is not nullable.

Parameters

views	Column views whose bitmasks will be concatenated
mr	Device memory resource used for allocating the returned memory
stream	CUDA stream used for device memory operations and kernel launches

Returns

Bitmasks of all the column views in the views vector