[
https://issues.apache.org/jira/browse/CALCITE-7106?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Shengkai Fang updated CALCITE-7106:
-----------------------------------
Description:
Recently, I am using Flink to support a VECTOR_SEARCH function. An similar
example is
```
SELECT * FROM orders, LATERAL TABLE(TOPN( TABLE Shipments, orders.productid))
```
But after sql2rel phase, the output tree is
```
LogicalProject(ROWTIME=[$0], PRODUCTID=[$1], ORDERID=[$2], ORDERID0=[$3],
ROWTIME0=[$4], RANK_NUMBER=[$5])
+- LogicalJoin(condition=[true], joinType=[inner])
+- LogicalTableScan(table=[[CATALOG, SALES, ORDERS]])
+- LogicalTableFunctionScan(invocation=[TOPN($1, $1)],
rowType=[RecordType(INTEGER ORDERID, TIMESTAMP(0) ROWTIME, BIGINT RANK_NUMBER)])
+- LogicalProject(ORDERID=[$0], ROWTIME=[$1])
+- LogicalTableScan(table=[[CATALOG, SALES, SHIPMENTS]])
```
At my first glance, it should use Correlate rather than Join here. After debug,
I find the Blackboard#lookupExp tries to use `inputs` to build expression if
inputs is not null. However, in this case, inputs are not members of the
current scope.
was:
Recently, I am using Flink to support a VECTOR_SEARCH function. An example is
```
SELECT * FROM orders, LATERAL TABLE(VECTOR_SEARCH( TABLE Shipments,
orders.productid))
```
But after sql2rel phase, the output tree is
```
LogicalProject(ROWTIME=[$0], PRODUCTID=[$1], ORDERID=[$2], ORDERID0=[$3],
ROWTIME0=[$4], RANK_NUMBER=[$5])
LogicalJoin(condition=[true], joinType=[inner])
LogicalTableScan(table=[[CATALOG, SALES, ORDERS]])
LogicalTableFunctionScan(invocation=[VECTOR_SEARCH($1, $1)],
rowType=[RecordType(INTEGER ORDERID, TIMESTAMP(0) ROWTIME, BIGINT RANK_NUMBER)])
LogicalProject(ORDERID=[$0], ROWTIME=[$1])
LogicalTableScan(table=[[CATALOG, SALES, SHIPMENTS]])
```
At my first glance, it should use Correlate rather than Join here. After debug,
I find the Blackboard#lookupExp tries to use `inputs` to build expression if
inputs is not null. However, in this case, inputs are not members of the
current scope.
> Can not generate correlated Polymorphic Table Function
> ------------------------------------------------------
>
> Key: CALCITE-7106
> URL: https://issues.apache.org/jira/browse/CALCITE-7106
> Project: Calcite
> Issue Type: Improvement
> Components: core
> Affects Versions: 1.40.0
> Reporter: Shengkai Fang
> Priority: Major
>
> Recently, I am using Flink to support a VECTOR_SEARCH function. An similar
> example is
>
> ```
> SELECT * FROM orders, LATERAL TABLE(TOPN( TABLE Shipments, orders.productid))
> ```
>
> But after sql2rel phase, the output tree is
> ```
> LogicalProject(ROWTIME=[$0], PRODUCTID=[$1], ORDERID=[$2], ORDERID0=[$3],
> ROWTIME0=[$4], RANK_NUMBER=[$5])
> +- LogicalJoin(condition=[true], joinType=[inner])
> +- LogicalTableScan(table=[[CATALOG, SALES, ORDERS]])
> +- LogicalTableFunctionScan(invocation=[TOPN($1, $1)],
> rowType=[RecordType(INTEGER ORDERID, TIMESTAMP(0) ROWTIME, BIGINT
> RANK_NUMBER)])
> +- LogicalProject(ORDERID=[$0], ROWTIME=[$1])
> +- LogicalTableScan(table=[[CATALOG, SALES, SHIPMENTS]])
> ```
>
> At my first glance, it should use Correlate rather than Join here. After
> debug, I find the Blackboard#lookupExp tries to use `inputs` to build
> expression if inputs is not null. However, in this case, inputs are not
> members of the current scope.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
