Apache DataFusion SQL Query Engine
feat: include source field context in cast errors (#24920)
Huaijin4f7ce268f8a5
diff --git a/datafusion/physical-expr/src/expressions/cast.rs b/datafusion/physical-expr/src/expressions/cast.rsindex 565d396dad..845b9608d9 100644--- a/datafusion/physical-expr/src/expressions/cast.rs+++ b/datafusion/physical-expr/src/expressions/cast.rs@@ -330,7 +330,22 @@ impl PhysicalExpr for CastExpr {fn evaluate(&self, batch: &RecordBatch) -> Result<ColumnarValue> {let value = self.expr.evaluate(batch)?;- value.cast_to(self.cast_type(), Some(&self.cast_options))+ value+ .cast_to(self.cast_type(), Some(&self.cast_options))+ .map_err(|error| {+ let source = self+ .expr+ .return_field(batch.schema().as_ref())+ .ok()+ .filter(|field| !field.name().is_empty())+ .map(|field| format!("field '{}'", field.name()))+ .unwrap_or_else(|| format!("expression '{}'", self.expr));+ error.context(format!(+ "Failed to cast {source} from {} to {}",+ value.data_type(),+ self.cast_type()+ ))+ })}fn return_field(&self, input_schema: &Schema) -> Result<FieldRef> {@@ -769,7 +784,11 @@ mod tests {let expression =cast_with_options(col("a", &schema)?, &schema, Decimal128(6, 2), None)?;let e = expression.evaluate(&batch).unwrap_err().strip_backtrace(); // panics on OK- assert_snapshot!(e, @"Arrow error: Invalid argument error: 123456.79 is too large to store in a Decimal128 of precision 6. Max is 9999.99");+ assert_snapshot!(e, @r"+ Failed to cast field 'a' from Decimal128(10, 3) to Decimal128(6, 2)+ caused by+ Arrow error: Invalid argument error: 123456.79 is too large to store in a Decimal128 of precision 6. Max is 9999.99+ ");// safe cast should return nulllet expression_safe = cast_with_options(col("a", &schema)?,@@ -1088,25 +1107,51 @@ mod tests {#[test]fn invalid_cast_with_options_error() -> Result<()> {- // Ensure a useful error happens at plan time if invalid casts are used+ // Ensure a useful error happens at runtime if invalid casts are usedlet schema = Schema::new(vec![Field::new("a", Utf8, false)]);let a = StringArray::from(vec!["9.1"]);let batch = RecordBatch::try_new(Arc::new(schema.clone()), vec![Arc::new(a)])?;let expression = cast_with_options(col("a", &schema)?, &schema, Int32, None)?;let result = expression.evaluate(&batch);- match result {- Ok(_) => panic!("expected error"),- Err(e) => {- assert!(- e.to_string()- .contains("Cannot cast string '9.1' to value of Int32 type")- )- }- }+ let error = result.expect_err("expected error").strip_backtrace();+ assert_eq!(+ error,+ "Failed to cast field 'a' from Utf8 to Int32\n\+ caused by\n\+ Arrow error: Cast error: Cannot cast string '9.1' to value of Int32 type"+ );Ok(())}+ #[test]+ fn invalid_cast_with_empty_field_name_uses_expression() {+ let schema = Schema::new(vec![Field::new("", Utf8, false)]);+ let batch = RecordBatch::try_new(+ Arc::new(schema.clone()),+ vec![Arc::new(StringArray::from(vec!["9.1"]))],+ )+ .expect("valid record batch");+ let expression = cast_with_options(+ col("", &schema).expect("valid column"),+ &schema,+ Int32,+ None,+ )+ .expect("valid cast expression");++ let error = expression+ .evaluate(&batch)+ .expect_err("expected error")+ .strip_backtrace();+ assert_eq!(+ error,+ "Failed to cast expression '@0' from Utf8 to Int32\n\+ caused by\n\+ Arrow error: Cast error: Cannot cast string '9.1' to value of Int32 type"+ );+ }+#[test]fn field_aware_cast_preserves_target_field_semantics() -> Result<()> {// Target field metadata should be preserved exactly (no merging with source).diff --git a/datafusion/sqllogictest/test_files/errors.slt b/datafusion/sqllogictest/test_files/errors.sltindex 5d3ddd43f0..55aa405ae3 100644--- a/datafusion/sqllogictest/test_files/errors.slt+++ b/datafusion/sqllogictest/test_files/errors.slt@@ -42,7 +42,7 @@ statement errorSELECT sin(c1) FROM aggregate_test_100# cast_expressions_error-statement error DataFusion error: Arrow error: Cast error: Cannot cast string 'c' to value of Int32 type+statement error DataFusion error: Failed to cast field 'c1' from Utf8View to Int32\ncaused by\nArrow error: Cast error: Cannot cast string 'c' to value of Int32 typeSELECT CAST(c1 AS INT) FROM aggregate_test_100# aggregation_with_bad_arguments