Skip to content

Spark map lookup is O(n) #54646

@DanSkells

Description

@DanSkells

In my day job, we encountered performance issues when we were performing lookups in a map with > 1,000,000 element. After profiling, we realised that the map lookup is a linear scan, and looking to either optimise this or understand if some other library or other programming metaphor exists within Spark. The use case is insertion and lookup, without any need to remove elements. Happy to discuss / work with maintainers if appropriate to address the issue. I have seen the OpenHashMap implementation, but this seems to be a Spark internal implementation, so not sure if this would be an appropriate base to work from.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions