[
https://issues.apache.org/jira/browse/CALCITE-5230?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17578095#comment-17578095
]
Julian Hyde commented on CALCITE-5230:
--------------------------------------
{quote}It doesn't have to, but it might be beneficial to make it available to
change the return type for those who want it.{quote}
Let's think about this in user-space, i.e. specification, not implementation.
Does currently Calcite's PERCENTILE_CONT currently derive the correct result
type in all cases? If not, what should it be?
> PERCENTILE_DISC return type derivation
> --------------------------------------
>
> Key: CALCITE-5230
> URL: https://issues.apache.org/jira/browse/CALCITE-5230
> Project: Calcite
> Issue Type: Bug
> Reporter: Itiel Sadeh
> Assignee: Itiel Sadeh
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.32.0
>
> Time Spent: 10m
> Remaining Estimate: 0h
>
> In the following expression:
> {code:java}
> PERCENTILE_DISC ( numeric_literal ) WITHIN GROUP ( ORDER BY
> order_by_expression) {code}
> the return type of `PERCENTILE_DISC` aggregate function is hard-coded as
> `DOUBLE`,
> while in Postgresql and MSSQL it is determined by the `order_by_expression`
> type.
> Currently there is no way to derive the correct type of `PERCENTILE_DISC`
> function from the `order_by_expression` type, as this expression is not an
> argument for the `PERCENTILE_DISC` function.
> Calcite is creating the following tree from this expression:
> {code:java}
> SqlCall(operator: WITHING_GROUP, operands: [SqlCall( operator:
> PERCENTILE_DISC, operands: [numeric_literal], order_by_expression]){code}
> for correct type derivation, the order_by_expression should be an operand of
> the PERCENTILE_DISC call
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
