Dmitrii Pogorelov created TEIID-5959: ---------------------------------------- Summary: Teiid doesn't re-optimize select sub-nodes after adding new criteria Key: TEIID-5959 URL: https://issues.redhat.com/browse/TEIID-5959 Project: Teiid Issue Type: Quality Risk Components: Query Engine Affects Versions: 13.1 Reporter: Dmitrii Pogorelov Assignee: Steven Hawkins It looks like Teiid doesn't re-optimize union nodes after adding new criteria to their sub-select nodes. For example, for the query (the user has only role1): {code:sql} -- QUERY 1 select * from "test_tables_pg.test_a" where hasRole('report-role') union all select * from "test_tables_pg2.test_a" where hasRole('role1') ;; {code} Teiid will generate the following Query plan: {code:xml} AccessNode(0) output=[test_tables_pg2.test_a.a, test_tables_pg2.test_a.b] SELECT g_0.a AS c_0, g_0.b AS c_1 FROM test_tables_pg2.test_a AS g_0 LIMIT 100 {code} what is correct because the user has only role1 and Teiid optimizes union removing the first sub-query as its condition in total is false. But for this query: {code:sql} -- QUERY 2 (adds a where condition) select * from ( select * from "test_tables_pg.test_a" where hasRole('report-role') union all select * from "test_tables_pg2.test_a" where hasRole('role1')) a where a = 1 ;; {code} Teiid will propagate the "a = 1" condition in all union sub-queries and generate the following query plan: {code:xml} LimitNode(0) output=[test_tables_pg.test_a.a, test_tables_pg.test_a.b] limit 100 UnionAllNode(1) output=[test_tables_pg.test_a.a, test_tables_pg.test_a.b] AccessNode(2) output=[test_tables_pg.test_a.a, test_tables_pg.test_a.b] SELECT g_0.a AS c_0, g_0.b AS c_1 FROM test_tables_pg.test_a AS g_0 WHERE (FALSE = TRUE) AND (g_0.a = 1) LIMIT 100 AccessNode(3) output=[test_tables_pg2.test_a.a, test_tables_pg2.test_a.b] SELECT g_0.a AS c_0, g_0.b AS c_1 FROM test_tables_pg2.test_a AS g_0 WHERE g_0.a = 1 LIMIT 100 {code} what is incorrect as the {code:xml} SELECT g_0.a AS c_0, g_0.b AS c_1 FROM test_tables_pg.test_a AS g_0 WHERE (FALSE = TRUE) AND (g_0.a = 1) LIMIT 100 {code} still has false condition and logically the query plan for the second query should be the same as for the first one, though it doesn't have an influence on results. I don't know whether it's a bug it not. That's why I created the issue with the "Quality Risk" type. Steps to reproduce: 1. In PostgreSQL create the test_a table using the following scripts: {code:sql} CREATE TABLE public.test_a ( a integer, b integer ); INSERT INTO public.test_a(a, b) VALUES (1, 1); INSERT INTO public.test_a(a, b) VALUES (1, 2); INSERT INTO public.test_a(a, b) VALUES (2, 1); INSERT INTO public.test_a(a, b) VALUES (2, 2); INSERT INTO public.test_a(a, b) VALUES (3, 2); INSERT INTO public.test_a(a, b) VALUES (3, 10); {code} 2. Add PostgreSQL database configuration in standalone-teiid.xml: {code:xml} <datasource jndi-name="java:/test_tables_pg" pool-name="test_tables_pg" enabled="true" use-java-context="true"> <connection-url>jdbc:postgresql://localhost:5432/test?charSet=utf8</connection-url> <driver-class>org.postgresql.Driver</driver-class> <driver>org.postgresql</driver> <pool> <min-pool-size>2</min-pool-size> <max-pool-size>70</max-pool-size> <prefill>false</prefill> <use-strict-min>false</use-strict-min> <flush-strategy>FailingConnectionOnly</flush-strategy> </pool> <security> <user-name>XXXXX</user-name> <password>XXXXX</password> </security> <validation> <check-valid-connection-sql>select 0</check-valid-connection-sql> </validation> <timeout> <blocking-timeout-millis>120000</blocking-timeout-millis> <idle-timeout-minutes>5</idle-timeout-minutes> </timeout> </datasource> <datasource jndi-name="java:/test_tables_pg2" pool-name="test_tables_pg2" enabled="true" use-java-context="true"> <connection-url>jdbc:postgresql://localhost:5432/test?charSet=utf8</connection-url> <driver-class>org.postgresql.Driver</driver-class> <driver>org.postgresql</driver> <pool> <min-pool-size>2</min-pool-size> <max-pool-size>70</max-pool-size> <prefill>false</prefill> <use-strict-min>false</use-strict-min> <flush-strategy>FailingConnectionOnly</flush-strategy> </pool> <security> <user-name>XXXXX</user-name> <password>XXXXX</password> </security> <validation> <check-valid-connection-sql>select 0</check-valid-connection-sql> </validation> <timeout> <blocking-timeout-millis>120000</blocking-timeout-millis> <idle-timeout-minutes>5</idle-timeout-minutes> </timeout> </datasource> {code} 3. Add in test-vdb.xml java:/test_tables_pg configured in the previous step as a data source: {code:xml} <model name="test_tables_pg"> <property name="importer.useFullSchemaName" value="false"/> <property name="importer.tableTypes" value="TABLE,VIEW"/> <property name="importer.importKeys" value="false"/> <source name="test_tables_pg" translator-name="myPg" connection-jndi-name="java:/test_tables_pg"/> </model> <model name="test_tables_pg2"> <property name="importer.useFullSchemaName" value="false"/> <property name="importer.tableTypes" value="TABLE,VIEW"/> <property name="importer.importKeys" value="false"/> <source name="test_tables_pg2" translator-name="myPg" connection-jndi-name="java:/test_tables_pg2"/> </model> {code} 4. add the "role1" configuration: {code:xml} <data-role name="role1" any-authenticated="true" allow-create-temporary-tables="true"> <description>Allow read only</description> <permission> <resource-name>test_tables_pg2.test_a</resource-name> <allow-read>true</allow-read> <allow-execute>true</allow-execute> </permission> <permission> <resource-name>test_tables_pg.test_a</resource-name> <allow-read>true</allow-read> <allow-execute>true</allow-execute> </permission> </data-role> {code} 5. run the following queries and compare their query plans: {code:sql} -- QUERY 1 select * from "test_tables_pg.test_a" where hasRole('report-role') union all select * from "test_tables_pg2.test_a" where hasRole('role1') ;; select hasRole('data','report-role') ;; select hasRole('role1') ;; -- QUERY 2 (adds a where condition) select * from ( select * from "test_tables_pg.test_a" where hasRole('report-role') union all select * from "test_tables_pg2.test_a" where hasRole('role1')) a where a = 1 ;; {code} -- This message was sent by Atlassian Jira (v7.13.8#713008)