[SPARK-38363] Avoid runtime error in Dataset.summary() when ANSI mode is on

• Type: Sub-task
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.3.0
• Fix Version/s: 3.3.0, 3.2.2
• Component/s: SQL
• Assignee: Gengliang Wang
• Reporter: Gengliang Wang
• Created: 01/Mar/22 13:54
• Updated: 02/Mar/22 02:54
• Resolved: 02/Mar/22 02:54

Description

When executing df.summary(), Spark SQL converts String columns as Double for the 

percentiles/mean/stddev metrics. 

This can cause runtime errors with ANSI mode on. 

Since this API is for getting a quick summary of the Dataframe, I suggest using "TryCast" for the problematic stats so that the API still works under ANSI mode.

Issue Links

• links to: [Github] Pull Request #35699 (gengliangwang)