[SPARK-7683] Confusing behavior of fold function of RDD in pyspark

• Type: Sub-task
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 1.3.1
• Fix Version/s: 2.0.0
• Component/s: PySpark
• Labels: releasenotes
• Target Version/s: 2.0.0
• External issue URL: https://github.com/apache/spark/pull/5587
• Assignee: Sean R. Owen
• Reporter: Ai He
• Created: 16/May/15 23:44
• Updated: 19/Jan/16 09:35
• Resolved: 19/Jan/16 09:34

Description

This will make the “fold” function consistent with the "fold" in rdd.scala and other "aggregate" functions where “acc” goes first. Otherwise, users have to write a lambda function like “lambda x, y: op(y, x)” if they want to use “zeroValue” to change the result type.

Issue Links

• is duplicated by: SPARK-12595 fold should pass arguments to op in the correct order
• relates to: SPARK-6416 Document that RDD.fold() requires the operator to be commutative
• links to: [Github] Pull Request #10771 (srowen)