[jira] [Created] (FLINK-20405) The LAG function in over window is not implements correctly

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

[jira] [Created] (FLINK-20405) The LAG function in over window is not implements correctly

Shang Yuanchun (Jira)
Leonard Xu created FLINK-20405:
----------------------------------

             Summary: The LAG function in over window is not implements correctly
                 Key: FLINK-20405
                 URL: https://issues.apache.org/jira/browse/FLINK-20405
             Project: Flink
          Issue Type: Bug
          Components: Table SQL / Runtime
    Affects Versions: 1.12.0
            Reporter: Leonard Xu


For LAG(input, offset, default) function in over window, it always return current row's input no matter how the offset is set.

After see the codegen code of the function, I think the implementation is not correct and need to correct.
{code:java}

// the offset and default value is never used
public UnboundedOverAggregateHelper$24(java.lang.Object[] references) throws Exception {                                    constant$14 = ((int) 1);            constant$14isNull = false;                                                constant$15 = ((org.apache.flink.table.data.binary.BinaryStringData) str$13);            constant$15isNull = false;                        typeSerializer$19 = (((org.apache.flink.table.runtime.typeutils.StringDataSerializer) references[0]));          }

public void accumulate(org.apache.flink.table.data.RowData accInput) throws Exception {                        org.apache.flink.table.data.binary.BinaryStringData field$21;            boolean isNull$21;            org.apache.flink.table.data.binary.BinaryStringData field$22;            isNull$21 = accInput.isNullAt(2);            field$21 = org.apache.flink.table.data.binary.BinaryStringData.EMPTY_UTF8;            if (!isNull$21) {              field$21 = ((org.apache.flink.table.data.binary.BinaryStringData) accInput.getString(2));            }            field$22 = field$21;            if (!isNull$21) {              field$22 = (org.apache.flink.table.data.binary.BinaryStringData) (typeSerializer$19.copy(field$22));            }                                                if (agg0_leadlag != field$22) {              agg0_leadlag = ((org.apache.flink.table.data.binary.BinaryStringData) typeSerializer$19.copy(field$22));            }                   ;            agg0_leadlagIsNull = isNull$21;                                         }
{code}
 

The question comes from user mail list[1]

[1] http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/FlinkSQL-kafka-gt-dedup-gt-kafka-td39335.html



--
This message was sent by Atlassian Jira
(v8.3.4#803005)