[
https://issues.apache.org/jira/browse/FLINK-14666?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16969964#comment-16969964
]
Dawid Wysakowicz commented on FLINK-14666:
------------------------------------------
[~yuemeng] let me check if I understand you correctly. What you are saying is
if we have:
cat1
- database2
- kafka_table
user_catalog
- user_db
- v2
A query like:
{code}
USE user_catalog;
USE user_db;
INSERT INTO database2.kafka_table SELECT * FROM v2;
{code}
will not work.
This is expected behavior. Because the query you are actually executing is
{{INSERT INTO user_catalog.database2.kafka_table SELECT * FROM
user_catalog.user_db.v2;}}. Table identifiers according to SQL standard are
always expanded with current catalog and database if needed. The temporary
tables are no exception here. If you want to execute queries cross catalogs you
must fully qualify them.
> support multiple catalog in flink table sql
> -------------------------------------------
>
> Key: FLINK-14666
> URL: https://issues.apache.org/jira/browse/FLINK-14666
> Project: Flink
> Issue Type: Bug
> Components: Table SQL / Planner
> Affects Versions: 1.8.0, 1.8.2, 1.9.0, 1.9.1
> Reporter: yuemeng
> Priority: Critical
> Labels: pull-request-available
> Time Spent: 20m
> Remaining Estimate: 0h
>
> currently, calcite will only use the current catalog as schema path to
> validate sql node,
> maybe this is not reasonable
> {code}
> tableEnvironment.useCatalog("user_catalog");
> tableEnvironment.useDatabase("user_db");
> Table table = tableEnvironment.sqlQuery("SELECT action, os,count(*) as cnt
> from music_queue_3 group by action, os,tumble(proctime, INTERVAL '10'
> SECOND)"); tableEnvironment.registerTable("v1", table);
> Table t2 = tableEnvironment.sqlQuery("select action, os, 1 as cnt from v1");
> tableEnvironment.registerTable("v2", t2);
> tableEnvironment.sqlUpdate("INSERT INTO database2.kafka_table_test1 SELECT
> action, os,cast (cnt as BIGINT) as cnt from v2");
> {code}
> suppose source table music_queue_3 and sink table kafka_table_test1 both in
> user_catalog
> catalog
> but some temp table or view such as v1, v2,v3 will register in default
> catalog.
> when we select temp table v2 and insert it into our own catalog table
> database2.kafka_table_test1
> it always failed with sql node validate, because of schema path in
> catalog reader is the current catalog without default catalog,the temp table
> or view will never be Identified
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
