changsun20 commented on issue #14438:
URL: https://github.com/apache/datafusion/issues/14438#issuecomment-2709581340
Hi @eliaperantoni,
After investigating this issue, here are my initial thoughts on
implementation:
The most straightforward approach would be to add a new `span` property to
the `Subquery` struct in `datafusion/expr/src/logical_plan/plan.rs`:
```rust
pub struct Subquery {
/// The subquery
pub subquery: Arc<LogicalPlan>,
/// The outer references used in the subquery
pub outer_ref_columns: Vec<Expr>,
pub span: Option<Span>,
}
```
The span would be extracted in the `parse_scalar_subquery` function within
`datafusion/sql/src/expr/subquery.rs`:
```rust
pub(super) fn parse_scalar_subquery(
&self,
subquery: Query,
// other params...
) -> Result<Expr> {
// other logic...
let span =
Span::try_from_sqlparser_span(subquery.some_way_to_index_the_span_from_the
query());
Ok(Expr::ScalarSubquery(Subquery {
subquery: Arc::new(sub_plan),
outer_ref_columns,
span,
}))
}
```
This span would then be accessible when generating error messages, allowing
us to add diagnostic information at
`datafusion\expr\src\logical_plan\invariants.rs`:
```rust
// This is the original implementation
if subquery.subquery.schema().fields().len() > 1 {
return plan_err!(
"Scalar subquery should only return one column, but found {}: {}",
subquery.subquery.schema().fields().len(),
subquery.subquery.schema().field_names().join(", ")
);
}
```
However, a drawback of this approach is that modifying the `Subquery` struct
will impact 14 other functions in the codebase, making this change less minimal
than ideal (not fulfilling the "as little invasive as possible" requirement).
Please let me know if this is on the right track or if you have any
suggestions. Thank you!
--
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
