feat: Implement the bitwise_not in NotExpr #5902
Conversation
github-actions
bot
added
core
| @@ -622,6 +622,17 @@ async fn test_not_expressions() -> Result<()> { | |||
| ]; | |||
| assert_batches_eq!(expected, &actual); | |||
|
|
|||
| let sql = "SELECT not(1), not(0)"; | |||
| let actual = execute_to_batches(&ctx, sql).await; | |||
There was a problem hiding this comment.
maybe lets move it to scalar.slt sqllogic test?
There was a problem hiding this comment.
I agree it would be nice to move to https://github.com/apache/arrow-datafusion/blob/main/datafusion/core/tests/sqllogictests/test_files/scalar.slt but we can do that potentially as a follow on PR
There was a problem hiding this comment.
Follow on PR to port the tests: #5968
| Ok(DataType::Boolean) | ||
| fn data_type(&self, input_schema: &Schema) -> Result<DataType> { | ||
| // Ok(DataType::Boolean) | ||
| let data_type = self.arg.data_type(input_schema)?; |
There was a problem hiding this comment.
That looks breaking imho... I would propose to have the own BitwiseNotExpr instead of making NotExpr universal
There was a problem hiding this comment.
As @alamb suggestion, I embed bitwise_not into NotExpr. But I agree that this way will break NotExpr original struct. I think I should further discuss the implementation in the issue #5700
| @@ -622,6 +622,17 @@ async fn test_not_expressions() -> Result<()> { | |||
| ]; | |||
| assert_batches_eq!(expected, &actual); | |||
|
|
|||
| let sql = "SELECT not(1), not(0)"; | |||
| let actual = execute_to_batches(&ctx, sql).await; | |||
There was a problem hiding this comment.
I agree it would be nice to move to https://github.com/apache/arrow-datafusion/blob/main/datafusion/core/tests/sqllogictests/test_files/scalar.slt but we can do that potentially as a follow on PR
* feat: make NotExpr support integer Bitwise_not * feat: make NotExpr support integer Bitwise_not --------- Co-authored-by: RT_Enzyme <52275903001@stu.ecnu.edu.cn>
This reverts commit 87e5adc
…ache#6599) This reverts commit 87e5adc
Which issue does this PR close?
Closes #5700. Because the SQL dialect use
Generic, sqlparser can't parse operator~as unary Bitwise_not. But we can usenotto complement integer instead of operator~in SQL.Rationale for this change
What changes are included in this PR?
Extending Expr::Not to work with integral types so that it can evaluate Bitwise_not expression.
Are these changes tested?
Pass Integrated tests.
Add United tests for bitwise_not integral types.
Are there any user-facing changes?
Yes. The
notcan NOT boolean type and Bitwise_not integer in sql.