[
https://issues.apache.org/jira/browse/CALCITE-7378?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
zzwqqq updated CALCITE-7378:
----------------------------
Description:
In dialects where `{{{}hasImplicitTableAlias{}}}is` true, even if a table is
explicitly assigned an alias in the original SQL, the `RelToSqlConverter` will
strip it away during the conversion. In scenarios like self-joins, if no
explicit alias is provided, column references will be incorrectly attributed to
the nearest table in the generated SQL structure.
This issue also occurs when handle TPC-H Query 21.
A minimal reproducible example：
{code:java}
select "product_name"
from "product" t1
where "product_id" in (select "product_id"
from "product" t2
where b."product_id" = a."product_id" and t1."product_id" = 2 and
t2."product_id" = 1){code}
after rel-to-sql
{code:java}
SELECT "product_name"
FROM "foodmart"."product"
WHERE "product_id" IN (SELECT "product_id"
FROM "foodmart"."product"
WHERE "product_id" = "product"."product_id" AND "product"."product_id" = 2 AND
"product_id" = 1){code}
I suggest that when a correlated table alias shares the same name as the
current table identifier, the table alias should be required.
was:
In dialects where {{{}hasImplicitTableAlias{}}}is true, even if a table is
explicitly assigned an alias in the original SQL, the {{RelToSqlConverter
}}will strip it away during the conversion. In scenarios like self-joins, if no
explicit alias is provided, column references will be incorrectly attributed to
the nearest table in the generated SQL structure.
This issue also occurs when handle TPC-H Query 21.
A minimal reproducible example：
```
select "product_name"
from "product" t1
where "product_id" in (select "product_id"
from "product" t2
where b."product_id" = a."product_id" and t1."product_id" = 2 and
t2."product_id" = 1)
```
after rel-to-sql
```
SELECT "product_name"
FROM "foodmart"."product"
WHERE "product_id" IN (SELECT "product_id"
FROM "foodmart"."product"
WHERE "product_id" = "product"."product_id" AND "product"."product_id" = 2 AND
"product_id" = 1)
```
I suggest that when a correlated table alias shares the same name as the
current table identifier, the table alias should be required.
> Potential incorrect column attribution in RelToSqlConverter due to implicit
> table alias handling
> ------------------------------------------------------------------------------------------------
>
> Key: CALCITE-7378
> URL: https://issues.apache.org/jira/browse/CALCITE-7378
> Project: Calcite
> Issue Type: Bug
> Components: core
> Reporter: zzwqqq
> Priority: Major
>
> In dialects where `{{{}hasImplicitTableAlias{}}}is` true, even if a table is
> explicitly assigned an alias in the original SQL, the `RelToSqlConverter`
> will strip it away during the conversion. In scenarios like self-joins, if no
> explicit alias is provided, column references will be incorrectly attributed
> to the nearest table in the generated SQL structure.
> This issue also occurs when handle TPC-H Query 21.
> A minimal reproducible example：
> {code:java}
> select "product_name"
> from "product" t1
> where "product_id" in (select "product_id"
> from "product" t2
> where b."product_id" = a."product_id" and t1."product_id" = 2 and
> t2."product_id" = 1){code}
> after rel-to-sql
> {code:java}
> SELECT "product_name"
> FROM "foodmart"."product"
> WHERE "product_id" IN (SELECT "product_id"
> FROM "foodmart"."product"
> WHERE "product_id" = "product"."product_id" AND "product"."product_id" = 2
> AND "product_id" = 1){code}
> I suggest that when a correlated table alias shares the same name as the
> current table identifier, the table alias should be required.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
