alan910127 commented on code in PR #15110:
URL: https://github.com/apache/datafusion/pull/15110#discussion_r2005677168
##########
datafusion/optimizer/src/analyzer/type_coercion.rs:
##########
@@ -290,19 +290,72 @@ impl<'a> TypeCoercionRewriter<'a> {
right: Expr,
right_schema: &DFSchema,
) -> Result<(Expr, Expr)> {
+ if let Expr::Literal(ref lit_value) = left {
+ if let Some(casted) =
+ try_cast_literal_to_type(lit_value, op,
&right.get_type(right_schema)?)
+ {
+ return Ok((casted, right));
+ };
+ }
+
+ if let Expr::Literal(ref lit_value) = right {
+ if let Some(casted) =
+ try_cast_literal_to_type(lit_value, op,
&left.get_type(left_schema)?)
+ {
+ return Ok((left, casted));
+ };
+ }
+
let (left_type, right_type) = BinaryTypeCoercer::new(
&left.get_type(left_schema)?,
&op,
&right.get_type(right_schema)?,
)
.get_input_types()?;
+
Ok((
left.cast_to(&left_type, left_schema)?,
right.cast_to(&right_type, right_schema)?,
))
}
}
+fn try_cast_literal_to_type(
+ lit_value: &ScalarValue,
+ op: Operator,
+ target_type: &DataType,
+) -> Option<Expr> {
+ match (op, lit_value) {
+ (
+ Operator::Eq | Operator::NotEq,
+ ScalarValue::Utf8(Some(_))
+ | ScalarValue::Utf8View(Some(_))
+ | ScalarValue::LargeUtf8(Some(_)),
+ ) => {
+ // Only try for integer types (TODO can we do this for other types
like timestamps)?
+ use DataType::*;
+ if matches!(
+ target_type,
+ Int8 | Int16 | Int32 | Int64 | UInt8 | UInt16 | UInt32 | UInt64
+ ) {
+ let opts = arrow::compute::CastOptions {
+ safe: false,
+ format_options: Default::default(),
+ };
+ let array = ScalarValue::to_array(lit_value).ok()?;
+ let casted =
+ arrow::compute::cast_with_options(&array, target_type,
&opts).ok()?;
+ ScalarValue::try_from_array(&casted, 0)
Review Comment:
@findepi thank you for the review. This file is focused on avoiding casts
being applied to the columns, and I was trying to emulate the behavior of
duckdb and postgres, where they always cast the literal and even throw an error
if the literal cannot be casted. So I was wondering whether we should even take
the operator into account here.
--
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: github-unsubscr...@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscr...@datafusion.apache.org
For additional commands, e-mail: github-h...@datafusion.apache.org
