vincev commented on code in PR #5106:
URL: https://github.com/apache/arrow-datafusion/pull/5106#discussion_r1097431852
##########
datafusion/expr/src/utils.rs:
##########
@@ -739,6 +739,35 @@ pub fn from_plan(
Ok(plan.clone())
}
LogicalPlan::DescribeTable(_) => Ok(plan.clone()),
+ LogicalPlan::Unnest(Unnest { column, schema, .. }) => {
+ // Update schema with unnested column type.
+ let input = Arc::new(inputs[0].clone());
+ let nested_field = input.schema().field_from_column(column)?;
+ let unnested_field = schema.field_from_column(column)?;
+ let fields = input
+ .schema()
+ .fields()
+ .iter()
+ .map(|f| {
+ if f == nested_field {
+ unnested_field.clone()
+ } else {
+ f.clone()
+ }
+ })
+ .collect::<Vec<_>>();
+
+ let schema = Arc::new(DFSchema::new_with_metadata(
+ fields,
+ input.schema().metadata().clone(),
+ )?);
+
+ Ok(LogicalPlan::Unnest(Unnest {
+ input,
+ column: column.clone(),
+ schema,
+ }))
+ }
Review Comment:
Ah yes, I had to add this code because tests were failing due to a
projection push-down when calling an aggregate function like count. At this
point the child schema has only the count field while the unnest schema has all
the fields in the initial projection.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
