[ 
https://issues.apache.org/jira/browse/SPARK-30767?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17033294#comment-17033294
 ] 

Maxim Gekk commented on SPARK-30767:
------------------------------------

Also 2.4.4 supports only `SSS` for second fractions. That was fixed by 
SPARK-29904 and SPARK-29949 in 2.4.5. Please, ask Databricks support which DBR 
versions will have those fixes.

> from_json changes times of timestmaps by several minutes without error 
> -----------------------------------------------------------------------
>
>                 Key: SPARK-30767
>                 URL: https://issues.apache.org/jira/browse/SPARK-30767
>             Project: Spark
>          Issue Type: Bug
>          Components: SQL
>    Affects Versions: 2.4.4
>         Environment: We ran the example code with Spark 2.4.4 via Azure 
> Databricks with Databricks Runtime version 6.3 within an interactive cluster. 
> We encountered the issue first on a Job Cluster running a streaming 
> application on Databricks Runtime Version 5.4.
>            Reporter: Benedikt Maria Beckermann
>            Priority: Major
>              Labels: corruption
>
> When a json text column includes a timestamp and the timestamp has a format 
> like {{2020-01-25T06:39:45.887429Z}}, the function 
> {{from_json(Column,StructType)}} is able to infer a timestamp but that 
> timestamp is changed by several minutes. 
> Spark does not throw any kind of error but continues to run with the 
> invalidated timestamp. 
> The following scala snipped is able to reproduce the issue.
>  
> {code:scala}
> import org.apache.spark.sql._
> import org.apache.spark.sql.functions._
> import org.apache.spark.sql.types._
> val df = Seq("""{"time":"2020-01-25T06:39:45.887429Z"}""").toDF("json")
> val struct = new StructType().add("time", TimestampType, nullable = true)
> val timeDF = df
>   .withColumn("time (string)", get_json_object(col("json"), "$.time"))
>   .withColumn("time casted directly (CORRECT)", col("time 
> (string)").cast(TimestampType))
>   .withColumn("time casted via struct (INVALID)", from_json(col("json"), 
> struct))
> display(timeDF)
> {code}
> Output: 
> ||json||time (string)||time casted directly (CORRECT)||time casted via struct 
> (INVALID)
> |{"time":"2020-01-25T06:39:45.887429Z"}|2020-01-25T06:39:45.887429Z|2020-01-25T06:39:45.887+0000|{"time":"2020-01-25T06:54:32.429+0000"}



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

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

Reply via email to