Skip to content

Remove cudautils.py #10706

Open
Open
@vyasr

Description

@vyasr

Currently the contents of cudautils.py can be grouped into three sets of functionality:

  1. The first set of 7 functions are all targeted at supporting find_first and find_last, which in turn are used respectively by the find_first_value and find_last_value methods of Column objects.
  2. The second set of functions 4 around window sizes are used by the rolling calculations in rolling.py.
  3. The third pair of 2 functions is used for UDF compilation.

The functions in group 3 are necessary but should probably be moved to core/udf/utils.py. The remaining functions should be possible to remove altogether, but they may require some additional functionality to be added to libcudf. For instance, find_first_value and find_last_value may be possible to implement using libcudf's lower_bound followed by an equality check, and similarly find_last_value could be replaced with upper_bound, accounting for all of group 1. For group 2 functions I am not certain if the necessary functionality exists in libcudf's roling aggregations, so we may need to do a little more engineering there first.

Metadata

Metadata

Labels

PythonAffects Python cuDF API.improvementImprovement / enhancement to an existing functionlibcudfAffects libcudf (C++/CUDA) code.

Type

No type

Projects

Status

In Progress

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions