[jira] [Created] (SPARK-39193) Improve the performance of inferring Timestamp type in JSON/CSV data source

["Gengliang Wang (Jira)" <ji...@apache.org>]

Gengliang Wang created SPARK-39193:
--------------------------------------

             Summary: Improve the performance of inferring Timestamp type in JSON/CSV data source
                 Key: SPARK-39193
                 URL: https://issues.apache.org/jira/browse/SPARK-39193
             Project: Spark
          Issue Type: Improvement
          Components: SQL
    Affects Versions: 3.3.0
            Reporter: Gengliang Wang
            Assignee: Gengliang Wang


When reading JSON/CSV files with inferring timestamp types `.option("inferTimestamp", true)`, the Timestamp conversion will throw and catch exceptions. As we are putting decent error messages in the exception, the creation of the exceptions is actually not cheap. It consumes more than 90% of the type inference time. 

We can use the parsing methods which return optional results instead.

Before the change, it takes 166 seconds to infer a JSON file of 624MB with inferring timestamp enabled.

After the change, it only 16 seconds.



--
This message was sent by Atlassian Jira
(v8.20.7#820007)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org