xudong963 edited a comment on issue #1576:
URL:
https://github.com/apache/arrow-datafusion/issues/1576#issuecomment-1013681759
I've found out why it failed.
let's see query 8, there is `when ... then ... else ...`.
```sql
select
o_year,
sum(case
when nation = 'BRAZIL' then volume
else 0
end) / sum(volume) as mkt_share
...
```
In datafusion, the following code to process `when ... then ... else ...`
```rust
fn if_then_else(
bools: &BooleanArray,
true_values: ArrayRef,
false_values: ArrayRef,
data_type: &DataType,
) -> Result<ArrayRef> {
match data_type {
...
DataType::Float64 => if_then_else!(
array::Float64Builder,
array::Float64Array,
bools,
true_values,
false_values
),
}
}
```
In query8, 0 is `Int64`, then during casing to `Float64`, it fails.
```rust
macro_rules! if_then_else {
($BUILDER_TYPE:ty, $ARRAY_TYPE:ty, $BOOLS:expr, $TRUE:expr, $FALSE:expr)
=> {{
let true_values = $TRUE
.as_ref()
.as_any()
.downcast_ref::<$ARRAY_TYPE>()
.expect("true_values downcast failed");
let false_values = $FALSE
.as_ref()
.as_any()
.downcast_ref::<$ARRAY_TYPE>()
.expect("false_values downcast failed");
```
There are two potential approaches to solve the issue
- Let arrow provide implicit conversion.
- Do type conversion in datafusion, such as we convert `true_values` and
`false_values` type to final type in advance.
Any thoughts? @alamb @houqp
--
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]
