addBatch / executeBatch does not support on cluster anymore

[den-crane]

final String DB_URL = "jdbc:clickhouse://localhost:8123";
final String USER = "default"
final String PASS = "";

ClickHouseDataSource dataSource = new ClickHouseDataSource(DB_URL);
ClickHouseConnection conn = dataSource.getConnection(USER, PASS);
ClickHouseStatement s =  conn.createStatement();

s.addBatch("drop table if exists aaa on cluster test_shard_localhost");
s.addBatch("create table if not exists aaa on cluster test_shard_localhost(a Int64) Engine=MergeTree order by a ;drop table if exists aaa on cluster test_shard_localhost;");
int[] results = s.executeBatch();
for (int i: results ) {
    System.out.print(i);
}

<groupId>ru.yandex.clickhouse</groupId>
<artifactId>clickhouse-jdbc</artifactId>
<version>0.3.1</version>

import ru.yandex.clickhouse.ClickHouseConnection;
import ru.yandex.clickhouse.ClickHouseDataSource;
import ru.yandex.clickhouse.ClickHouseStatement;


output: 111

<groupId>com.clickhouse</groupId>
<artifactId>clickhouse-jdbc</artifactId>
<version>0.4.0</version>

import com.clickhouse.jdbc.ClickHouseConnection;
import com.clickhouse.jdbc.ClickHouseDataSource;
import com.clickhouse.jdbc.ClickHouseStatement;

Exception in thread "main" java.sql.BatchUpdateException: Query is not allowed in batch update

without on cluster test_shard_localhost -- it works / output: 000

[sondertara]

The version 0.4.1 still has this situation, in the package com.clickhouse.jdbc.internal all PreparedStatement implementations have similar code in the method executeAny :

      try {
                r = executeStatement(builder.toString(), null, null, null);
                if (updateResult(parsedStmt, r) != null && asBatch) {
                    throw SqlExceptionUtils.queryInBatchError(results);
                }
               ...
              
            } catch (Exception e) {
                ...
                if (!continueOnError) {
                    throw SqlExceptionUtils.batchUpdateError(e, results);
                }
                log.error("Failed to execute batch insertion of %d records", counter, e);
            } finally {
                if (asBatch && r != null) 
                    r.close();
                }
                clearBatch();
            }

When sql with on cluster , the updateResult always return response columns, we can add property (continueOnError =true ) to the jdbc connection make it continue.

But I still want to know if this is a bug, or some kind of specification or practice, who can tell me the specific design details?

[zhicwu]

But I still want to know if this is a bug, or some kind of specification or practice, who can tell me the specific design details?

Yes it's a bug. The assumption was DDL should have empty response, which is apparently not true when on cluster was used.

[sondertara]

@zhicwu Thank you for your prompt reply and looking forward to the v1 release. 🧋🧋