Access Cassandra From Scala – Hector

I was looking for a convenient way to access Cassandra from Scala for a new version of our app that is currently in development. After a long and tedious afternoon I finally got something working that feels solid, so I thought I’d share my experience here.

The first time I checked the Client Options wiki page there were a couple of projects listed for the Scala language. I wasn’t convinced by any of the alternatives I looked at. Given that Cassandra is such a fast-moving target, I felt that the client library needs to have at least one committed developer (and some community momentum) behind it.

I took a quick look at scromium but I wasn’t convinced with the choice of different token separators. I then looked more deeply into scalandra but just couldn’t make it work with Cassandra 0.6-rc1. I was initially interested in the Akka option since I was already looking at Akka for it’s distributed actors but I was disappointed by its current implementation.

You can use Akka to access Cassandra via the Pluggable Persistence module. According to the documentation you can bypass the Akka STM and use it as a library; they call it the Cassandra Session API. After a lot of tinkering, building Thrift on MacOS X, installing MacPorts to get the missing pieces, etc. I still couldn’t get a simple test to run against Cassandra 0.6-rc1. I later noticed that even Akka developers were considering a shift to Hector, a Java-based library. Given this endorsement I decided to try it myself.

Hector is a thin wrapper over the Cassandra Thrift API. It basically makes it a bit more object-oriented and easy to use from Java.

It turns out that talking to Cassandra from Scala using Hector is very simple, albeit not very Scala-ish. I set up a basic Scala project in Eclipse and added the Hector libraries to the classpath. Then I modified ExampleClient.java and with Cassandra running on the default port it worked like a charm:

import me.prettyprint.cassandra.service.CassandraClientPoolFactory
import me.prettyprint.cassandra.utils.StringUtils._

import org.apache.cassandra.thrift.{Column, ColumnPath}

object CassandraScalaHectorTest {
  def main(args : Array[String]) : Unit = {
    val pool = CassandraClientPoolFactory.INSTANCE.get()
    // var client = pool.borrowClient(Array("cas1:9160", "cas2:9160")
    var client = pool.borrowClient(Array("localhost:9160"))

    try {
      val keyspace = client.getKeyspace("Keyspace1")
      val columnPath = new ColumnPath("Standard1").setColumn(bytes("column-name"))

      // insert
      keyspace.insert("key", columnPath, bytes("value"))

      // read
      val col = keyspace.getColumn("key", columnPath)
      System.out.println("Stored value: " + string(col.getValue()))

      // This line makes sure that even if the client had
      // failures and recovered, a correct
      // releaseClient is called, on the up to date client.
      client = keyspace.getClient()

    } finally {
      // return client to pool. do it in a finally block to
      // make sure it's executed
      pool.releaseClient(client)
    }
  }
}

As you can see this was only slightly modified so it compiles in Scala. It’s a little cleaner and less verbose than the Java version, mostly because of Scala’s type inference, which reduces the need to declare all types.

One of the things I don’t like is the explicit conversions to and from byte arrays. This is exactly the kind of boilerplate that can be done for you by using Scala’s implicit conversions and, overall, it’s just not very idiomatic Scala. Although I’m not a purist and I could live with this solution, I would prefer a cleaner approach. I still haven’t ruled out the possibility of wrapping Hector in Scala and providing a couple of conversions to make it work as expected.

Since I started working on this test the Cassandra Client Options wiki page was updated with a new Scala library called cascal. I haven’t had a chance to try it yet but it looks promising so I will evaluate that next.

Perhaps there’s a silver bullet I haven’t even heard off yet. I’m curious to know how many people are fronting Cassandra with a Scala app.

Advertisement