[
https://issues.apache.org/jira/browse/SPARK-52407?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Christopher Boumalhab updated SPARK-52407:
------------------------------------------
Labels: datasketches (was: )
> Add Native Support for Apache Theta Sketches
> --------------------------------------------
>
> Key: SPARK-52407
> URL: https://issues.apache.org/jira/browse/SPARK-52407
> Project: Spark
> Issue Type: New Feature
> Components: PySpark, SQL
> Affects Versions: 3.4.4
> Reporter: Christopher Boumalhab
> Priority: Minor
> Labels: datasketches
> Original Estimate: 672h
> Remaining Estimate: 672h
>
> *Add Theta Sketches Support to Spark SQL*
> This proposal aims to integrate Apache DataSketches' Theta Sketches into
> Spark SQL and the DataFrame APIs. While Spark already includes support for
> HLL sketches via {{hll_sketch_*}} functions, Theta Sketches provide
> additional capabilities not covered by HLL, such as set operations
> (intersection, difference) and value aggregation by key.
> *Motivation:*
> * Enable scalable and memory-efficient approximate set operations for
> large-scale analytics.
> * Support high-cardinality distinct counts, approximate aggregations over
> batch and streaming data, and set-based operations across datasets.
> * Leverage Apache DataSketches, which is already an Apache incubating
> project and a dependency within Spark.
> *Proposed Features:*
> * New aggregate functions for Theta Sketches, including:
> ** {{theta_sketch_agg(col)}} — build Theta sketches
> ** {{theta_union(sketch1, sketch2)}} and {{theta_union_agg(sketch_col)}} —
> union operations
> ** {{theta_intersection(sketch1, sketch2)}} and
> {{theta_intersection_agg(sketch_col)}} — intersection operations
> ** {{theta_difference(sketch1, sketch2)}} — difference operation
> ** {{theta_sketch_estimate(sketch)}} — estimate cardinality
> * Similar functions to support Tuple Sketches, prioritized after Theta
> sketches are integrated.
> *Implementation Notes:*
> * Follow naming and design conventions established by existing HLL sketch
> UDFs.
> * Engage with the Apache DataSketches community for technical guidance and
> cross-project synergy.
> This enhancement will enable Spark users to perform advanced approximate
> analytics with improved performance and scalability, complementing existing
> approximate functions.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
