[SPARK-49739][SQL] Throw Custom error condition for invalid options in SaveIntoDataSourceCommand class #48188
Conversation
| @@ -46,6 +46,10 @@ case class SaveIntoDataSourceCommand( | |||
| override def innerChildren: Seq[QueryPlan[_]] = Seq(query) | |||
|
|
|||
| override def run(sparkSession: SparkSession): Seq[Row] = { | |||
| if (options.values.exists(_ == null)) { | |||
There was a problem hiding this comment.
How about keys? Should we check them too?
There was a problem hiding this comment.
No, the keys cannot be null since Map[String, String] can't accept null keys.
| }, | ||
| "NULL_VALUE" : { | ||
| "message" : [ | ||
| "Values of keys and values in `map()` must be string, but got null in value." |
There was a problem hiding this comment.
Users can build values dynamically, and could be not clear which one is null. Let's help them in troubleshooting, and provide key which has the null value.
| @@ -2752,6 +2752,11 @@ | |||
| "message" : [ | |||
| "A type of keys and values in `map()` must be string, but got <mapType>." | |||
| ] | |||
| }, | |||
| "NULL_VALUE" : { | |||
There was a problem hiding this comment.
How about to re-use the existing error condition: NULL_DATA_SOURCE_OPTION?
There was a problem hiding this comment.
i actually missed this condition, it makes more sense
|
Master branch is guarded from the |
|
Closing the pr because #46955 solves it. |
What changes were proposed in this pull request?
NullPointerException was thrown when invalid options were provided in the run function of SaveIntoDataSourceCommand up to now. That is because this map is propagated here, and java.util.Properties throws NPE if key/value is null. In map, a key cannot be null, but a value can, so that is why this NPE is possible to produce.
I propose we send the client an appropriate error condition. I've created a new error condition
INVALID_OPTIONS.NULL_VALUE. Options are checked, and the error is thrown if there exists an entry in options map where the value is null.Why are the changes needed?
To send the client an appropriate error.
Does this PR introduce any user-facing change?
Yes, in case invalid options are sent, we throw
INVALID_OPTIONS.NULL_VALUEinstead of NullPointerException.How was this patch tested?
Wrote a unit test.
Was this patch authored or co-authored using generative AI tooling?
No.