[ 
https://issues.apache.org/jira/browse/CALCITE-6817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17925840#comment-17925840
 ] 

ZheHu commented on CALCITE-6817:
--------------------------------

I re-test your PR around testUnnestRow(), for the following case.
actualType.toString() results in "RecordType(INTEGER EXPR$0, CHAR(5) EXPR$1)". 
However, if you use actualType.digest, you will get "RecordType(INTEGER NOT 
NULL EXPR$0, CHAR(5) EXPR$1) NOT NULL".
{code:java}
sql("with orders(data) as\n"
        + "  (values (ARRAY[ARRAY[ROW(1, 'Alice'), ROW(2, NULL)], NULL]))\n"
        + "select e.*\n"
        + "from orders, UNNEST(orders.data[1]) as e")
        .type(actualType -> {
          // The inner array that is unnested is nullable in this example
          assertThat(actualType, hasToString("RecordType(INTEGER EXPR$0, 
CHAR(5) EXPR$1)"));
          assertTrue(actualType.isStruct());
          assertThat(actualType.getFieldList().get(0).getType().isNullable(), 
is(false));
          assertTrue(actualType.getFieldList().get(1).getType().isNullable());
        });
{code}

> Add string representation of default nulls direction for RelNode
> ----------------------------------------------------------------
>
>                 Key: CALCITE-6817
>                 URL: https://issues.apache.org/jira/browse/CALCITE-6817
>             Project: Calcite
>          Issue Type: Improvement
>          Components: core
>    Affects Versions: 1.38.0
>            Reporter: ZheHu
>            Assignee: ZheHu
>            Priority: Minor
>              Labels: pull-request-available
>             Fix For: 1.39.0
>
>
> Calcite has the same hehavior with Oracle on returning the null direction if 
> not specified, which means NULLS LAST is assumed for ASC and NULLS FIRST is 
> assumed for DESC.
> However, the default nulls direction is not presented in plan traces. For 
> example:
> {code:java}
> select empno from emp order by empno desc
> {code}
> has the following plan:
> {code:java}
> LogicalSort(sort0=[$0], dir0=[DESC])
>   LogicalProject(EMPNO=[$0])
>     LogicalTableScan(table=[[CATALOG, SALES, EMP]])
> {code}
> It makes RelNode clearer if adding nulls direction, so the above plan will be 
> like:
> {code:java}
> LogicalSort(sort0=[$0], dir0=[DESC-nulls-first])
>   LogicalProject(EMPNO=[$0])
>     LogicalTableScan(table=[[CATALOG, SALES, EMP]])
> {code}



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to