goldmedal commented on issue #10688: URL: https://github.com/apache/datafusion/issues/10688#issuecomment-2138499681
After some testing, I found it could be an arrow-cast issue. When trying to cast a float32 to float64, the precision is missing. We can easy to reproduce this case like: ``` > select (128.2::float)::double ; +-------------------+ | Float64(128.2) | +-------------------+ | 128.1999969482422 | +-------------------+ 1 row(s) fetched. Elapsed 0.089 seconds. ``` I found it happens in `SimplifyExpressions` rule. `SimpleExpr` will try to evaluate the casting expression. https://github.com/apache/datafusion/blob/3d007608535cb138ae4473ce6305bd4ec8481627/datafusion/common/src/scalar/mod.rs#L2506 `cast_to` will invoke the arrow-cast to get the result. It can also be reproduced like ```rust let expr2 = ScalarValue::Float64(Some(128.2)) .cast_to(&DataType::Float32) .unwrap() .cast_to(&DataType::Float64) .unwrap(); println!("expr2: {:?}", expr2); ------------------------------------ expr2: Float64(128.1999969482422) ``` It happens in `arrow-cast-51.0.0/src/cast.rs` (I'm not sure where is the code placed). ```rust (Float16, Float32) => cast_numeric_arrays::<Float16Type, Float32Type>(array, cast_options), ``` I think it maybe not a DataFusion issue but a arrow-cast issue. @comphead @viirya What do you think? -- 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
