Initialize cluster via connection URL

[selevit]

Hi, and thanks for the great library!

Wanted to mention that it would be nice to be able to initialize the ClickhouseCluster class via a URL like clickhouse//:user:password@host.... The clickhouse_driver library that you use actually supports it via a Client.from_url static method.

Unfortunately, clickhouse_driver does not yet expose the "parse url" API, and I need to parse the URL manually to pass it to ClickhouseCluster.

[zifter]

Hey,

Thank you for the idea.

I made the requested changes, #25

You can find the necessary changes in the latest release 0.7.0

[selevit]

Hello, thanks a lot for implementing this!

There is a minor issue that I wanted to report. Now it's possible to pass a default database for migration into ClickhouseCluster (via a URL), but ClickhouseCluster.migrate requires db_name to be passed. Would be nice to make this parameter optional.

Because clickhouse_driver does not yet expose their URL parsing API (I also created a ticket in their issues).

It means that to supply db_name to ClickhouseCluster I need to parse the URL myself, i.e. reimplement this logic of Client.from_url. It doesn't make a lot of sense, since clickhouse-migrations supports the db URL now.

[zifter]

#26 is merged
@selevit can you check the latest release 0.7.1?

[selevit]

Big thanks for the quick fix!

It worked for me. Just a minor – IMO it looks a bit weird that it's required to pass the db name explicitly as None, like this.

cluster.migrate(
    db_name=None,  # use default db
    migration_path=MIGRATIONS_DIR,
    ...
)

[zifter]

@selevit yes, looks weird, but I don't want to make breaking changes in API