executeBatch implementation

@serebrserg @AndreevDm

Currently i am working on integration between metabase and ClickHouse.

It seems that clickhouse-jdbc driver doesn’t provide support for formats like: INSERT INTO test.batch_insert (s, i) VALUES (?, ?), (?, ?) and INSERT INTO test.batch_insert (s, i) VALUES (?, 101), (?, 102)

Looks like it is supported by ClickHouse (i’ve wrote some tests to theese formats and fixed 'em), so can it be supported in JDBC?

[lloonngg]

Hi, any update to this issue? Would love to resolve this so that @Badya could hopefully complete a connector from Metabase to Clickhouse. We are using Metabase internally for GreenPlum, and it is a fantastic interface for sql queries and visualization. Being able to incorporate Clickhouse into Metabase would allow us to work on much larger datasets.

[serebrserg]

Hi. There hadn’t been updates. Let me describe my point:

• I agree that implementation differs from standard
• I agree that it would be good to allow different ways to express insert statements
• I think that current “bulk” insert must be encouraged to use and other types including ones in the issue must not, because they are not as effective for ClickHouse.

ClickHouse works effectively with big inserts up to 1M rows and the data for such “bulk” inserts may be compressed for network transfer.

The way I see to make best of both is to turn these complex inserts into simple “bulk” ones with some tricky parsing, which I wouldn’t like to do, or leave current implementation for simple inserts, and “degrade” to copying values groups into query(making a long query with a lot of values groups) for complex ones, which will be less efficient, but easier to implement. I prefer second option, with clear documentation that it is not efficient and is not recommended. Does this sound reasonable for you and do you think it solves the problem? If answer is positive I’ll do this. Anyway, contributions are welcome.