Jefffrey commented on code in PR #20205:
URL: https://github.com/apache/datafusion/pull/20205#discussion_r2787436596
##########
datafusion/functions/src/math/nanvl.rs:
##########
@@ -101,7 +104,49 @@ impl ScalarUDFImpl for NanvlFunc {
}
fn invoke_with_args(&self, args: ScalarFunctionArgs) ->
Result<ColumnarValue> {
- make_scalar_function(nanvl, vec![])(&args.args)
+ let [x, y] = take_function_args(self.name(), args.args)?;
+
+ match (&x, &y) {
+ (ColumnarValue::Scalar(x), ColumnarValue::Scalar(y)) => {
Review Comment:
I think we can do this simpler:
```rust
match (x, y) {
// if x is any scalar nan we return y
(ColumnarValue::Scalar(ScalarValue::Float16(Some(v))), _) if v.is_nan()
=> return y,
(ColumnarValue::Scalar(ScalarValue::Float32(Some(v))), _) if v.is_nan()
=> return y,
(ColumnarValue::Scalar(ScalarValue::Float64(Some(v))), _) if v.is_nan()
=> return y,
(ColumnarValue::Scalar(_), _) => return x, // x is not scalar nan so we
can safely return it, ignore y
// fallback
(ColumnarValue::Array(x), ColumnarValue::Array(y)) => return nanvl(x, y),
_ => unreachable_internal_error(),
}
```
- Basically, if the first argument is scalar we never need to expand either
(we don't need y to be scalar)
##########
datafusion/sqllogictest/test_files/scalar.slt:
##########
@@ -765,11 +765,11 @@ select nanvl(null, 64);
----
NULL
-# nanvl scalar nulls #1
+# nanvl scalar nulls #1 - x is not NaN, so return x even if y is NULL
query R rowsort
select nanvl(2, null);
----
-NULL
Review Comment:
This is interesting; I think we need to fix the array behaviour as well
actually
--
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]
