[
https://issues.apache.org/jira/browse/SPARK-44714?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Xinyi Yu updated SPARK-44714:
-----------------------------
Description:
Current LCA resolution has a limitation, that it can't resolve the query, when
it satisfies all the following criteria:
# the main (outer) query has having clause
# there is a window expression in the query
# in the same SELECT list as the window expression in 2), there is an lca
This is because LCA won't rewrite plan until UNRESOLVED_HAVING is resolved;
window expressions won't be extracted until LCA in the same SELECT lists are
rewritten; however UNRESOLVED_HAVING depends on the child to be resolved, which
could include the Window. It becomes a deadlock.
*We should ease some limitation on the LCA resolution regarding to having, to
break the deadlock for most cases.*
For example, for the following query:
{code:java}
create table t (col boolean) using orc;
with w AS (
select min(col) over () as min_alias,
min_alias as col_alias
FROM t
)
select col_alias
from w
having count > 0;
{code}
It now throws confusing error message:
{code:java}
[UNRESOLVED_COLUMN.WITH_SUGGESTION] A column or function parameter with name
`col_alias`
cannot be resolved. Did you mean one of the following? [`col_alias`,
`min_alias`].{code}
The LCA and window is in a CTE that is completely unrelated to the having. LCA
should resolve in this case.
was:
Current LCA resolution has a limitation, that it can't resolve the query, when
it satisfies all the following criteria:
# the main (outer) query has having clause
# there is a window expression in the query
# in the same SELECT list as the window expression in 2), there is an lca
This is because LCA won't rewrite plan until UNRESOLVED_HAVING is resolved;
window expressions won't be extracted until LCA in the same SELECT lists are
rewritten; however UNRESOLVED_HAVING depends on the child to be resolved, which
could include the Window. It becomes a deadlock.
*We should ease some limitation on the LCA resolution regarding to having, to
break the deadlock for most cases.*
For example, for the following query:
create table t (col boolean) using orc;
with w AS ( select min(col) over () as min_alias,
min_alias as col_alias FROM t
)select col_alias from whaving count(*) > 0;
It now throws confusing error message:
[UNRESOLVED_COLUMN.WITH_SUGGESTION] A column or function parameter with name
`col_alias`
cannot be resolved. Did you mean one of the following? [`col_alias`,
`min_alias`].
The LCA and window is in a CTE that is completely unrelated to the having. LCA
should resolve in this case.
> Ease restriction of LCA resolution regarding queries with having
> ----------------------------------------------------------------
>
> Key: SPARK-44714
> URL: https://issues.apache.org/jira/browse/SPARK-44714
> Project: Spark
> Issue Type: Sub-task
> Components: SQL
> Affects Versions: 3.4.2, 3.5.0
> Reporter: Xinyi Yu
> Priority: Major
>
> Current LCA resolution has a limitation, that it can't resolve the query,
> when it satisfies all the following criteria:
> # the main (outer) query has having clause
> # there is a window expression in the query
> # in the same SELECT list as the window expression in 2), there is an lca
> This is because LCA won't rewrite plan until UNRESOLVED_HAVING is resolved;
> window expressions won't be extracted until LCA in the same SELECT lists are
> rewritten; however UNRESOLVED_HAVING depends on the child to be resolved,
> which could include the Window. It becomes a deadlock.
> *We should ease some limitation on the LCA resolution regarding to having, to
> break the deadlock for most cases.*
> For example, for the following query:
> {code:java}
> create table t (col boolean) using orc;
> with w AS (
> select min(col) over () as min_alias,
> min_alias as col_alias
> FROM t
> )
> select col_alias
> from w
> having count > 0;
> {code}
>
> It now throws confusing error message:
> {code:java}
> [UNRESOLVED_COLUMN.WITH_SUGGESTION] A column or function parameter with name
> `col_alias`
> cannot be resolved. Did you mean one of the following? [`col_alias`,
> `min_alias`].{code}
> The LCA and window is in a CTE that is completely unrelated to the having.
> LCA should resolve in this case.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
