[DRILL-7340] Filter is not pushed to JDBC database when several databases are used in the query

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.16.0
• Fix Version/s: 1.18.0
• Component/s: Storage - JDBC
• Assignee: Igor Guzenko
• Reporter: Vova Vysotskyi
• Reviewer: Vova Vysotskyi
• Created: 06/Aug/19 14:03
• Updated: 02/Jan/20 16:49
• Resolved: 02/Jan/20 16:49

Description

For the case when several databases are used in the query, some rules weren't added to the rule set for one of the conventions. It is observed in queries similar to the next query:

select * from mysql.`drill_mysql_test`.person t1
INNER JOIN h2.drill_h2_test.person t2 on t1.person_id = t2.person_id where t1.first_name = 'first_name_1' and t2.last_name = 'last_name_1

Plan for this query is the following:

00-00    Screen
00-01      Project(person_id=[$0], first_name=[$1], last_name=[$2], address=[$3], city=[$4], state=[$5], zip=[$6], json=[$7], bigint_field=[$8], smallint_field=[$9], numeric_field=[$10], boolean_field=[$11], double_field=[$12], float_field=[$13], real_field=[$14], time_field=[$15], timestamp_field=[$16], date_field=[$17], datetime_field=[$18], year_field=[$19], text_field=[$20], tiny_text_field=[$21], medium_text_field=[$22], long_text_field=[$23], blob_field=[$24], bit_field=[$25], enum_field=[$26], PERSON_ID0=[$27], FIRST_NAME0=[$28], LAST_NAME0=[$29], ADDRESS0=[$30], CITY0=[$31], STATE0=[$32], ZIP0=[$33], JSON0=[$34], BIGINT_FIELD0=[$35], SMALLINT_FIELD0=[$36], NUMERIC_FIELD0=[$37], BOOLEAN_FIELD0=[$38], DOUBLE_FIELD0=[$39], FLOAT_FIELD0=[$40], REAL_FIELD0=[$41], TIME_FIELD0=[$42], TIMESTAMP_FIELD0=[$43], DATE_FIELD0=[$44], CLOB_FIELD=[$45])
00-02        HashJoin(condition=[=($0, $27)], joinType=[inner], semi-join: =[false])
00-03          Project(PERSON_ID0=[$0], FIRST_NAME0=[$1], LAST_NAME0=[$2], ADDRESS0=[$3], CITY0=[$4], STATE0=[$5], ZIP0=[$6], JSON0=[$7], BIGINT_FIELD0=[$8], SMALLINT_FIELD0=[$9], NUMERIC_FIELD0=[$10], BOOLEAN_FIELD0=[$11], DOUBLE_FIELD0=[$12], FLOAT_FIELD0=[$13], REAL_FIELD0=[$14], TIME_FIELD0=[$15], TIMESTAMP_FIELD0=[$16], DATE_FIELD0=[$17], CLOB_FIELD=[$18])
00-05            SelectionVectorRemover
00-06              Filter(condition=[=($2, 'last_name_1')])
00-07                Jdbc(sql=[SELECT * FROM "TMP"."DRILL_H2_TEST"."PERSON" ])
00-04          Jdbc(sql=[SELECT * FROM `drill_mysql_test`.`person` WHERE `first_name` = 'first_name_1' ])

DrillJdbcFilterRule wasn't applied for H2 convention and Filter wasn't pushed to H2 database.

This issue may be fixed by specifying JdbcConvention in rules descriptions in Drill DrillJdbcFilterRule and DrillJdbcProjectRule rules and other rules should be fixed in Calcite in the scope of CALCITE-3115.

Issue Links

• is related to: CALCITE-3115 Cannot add JdbcRules which have different JdbcConvention to same VolcanoPlanner's RuleSet.
• relates to: DRILL-7406 Update Calcite to 1.21.0