Cassandra 4 io

This is lightweight cats-effect and fs2 IO wrapper for latest datastax 4.x driver.

Why 4.x?

4.x was re-written in immutable first design, within async first API, optimizations, fewer allocations, metrics improvements, and fully compatible with cassandra 3.x

Goals

• Be safe, type-safe.
• Be fast
  • use minimal allocations
  • minimize resources and abstractions overhead over original datastax driver, which is good

How to use

Cassandra4io is currently available for Scala 2.13 and 2.12.

Add a dependency to your project

libraryDependencies += ("com.ringcentral" %% "cassandra4io" % "0.1.14")

Create a connection to Cassandra

import com.ringcentral.cassandra4io.CassandraSession

import com.datastax.oss.driver.api.core.CqlSession
import cats.effect._

import java.net.InetSocketAddress

val builder = CqlSession
      .builder()
      .addContactPoint(InetSocketAddress.createUnresolved("localhost", 9042))
      .withLocalDatacenter("datacenter1")
      .withKeyspace("awesome") 

def makeSession[F[_]: Async]: Resource[F, CassandraSession[F]] =
  CassandraSession.connect(builder)

Write some requests

package com.ringcentral.cassandra4io.cql introduces typed way to deal with cql queries

Simple syntax

import cats.effect.Sync
import cats.syntax.all._
import com.datastax.oss.driver.api.core.ConsistencyLevel
import com.ringcentral.cassandra4io.CassandraSession
import com.ringcentral.cassandra4io.cql._

case class Model(id: Int, data: String)

trait Dao[F[_]] {
  def put(value: Model): F[Unit]
  def get(id: Int): F[Option[Model]]
}

class DaoImpl[F[_]: Async](session: CassandraSession[F]) extends Dao[F] {

  private def insertQuery(value: Model) =
    cql"insert into table (id, data) values (${value.id}, ${value.data})"
      .config(_.setConsistencyLevel(ConsistencyLevel.ALL))

  private def selectQuery(id: Int) =
    cql"select id, data from table where id = $id".as[Model]
  
  override def put(value: Model) = insertQuery(value).execute(session).void
  override def get(id: Int) = selectQuery(id).select(session).head.compile.last
}

this syntax reuse implicit driver prepared statements cache

Templated syntax

import cats.effect._
import scala.concurrent.duration._
import cats.syntax.all._
import scala.jdk.DurationConverters._
import com.datastax.oss.driver.api.core.ConsistencyLevel
import com.ringcentral.cassandra4io.CassandraSession
import com.ringcentral.cassandra4io.cql._
    
case class Model(id: Int, data: String)
  
trait Dao[F[_]] {
  def put(value: Model): F[Unit]
  def get(id: Int): F[Option[Model]]
}
    
object Dao {
  
  private val tableName = "table"
  private val insertQuery = cqlt"insert into ${Const(tableName)} (id, data) values (${Put[Int]}, ${Put[String]})"
    .config(_.setTimeout(1.second.toJava))
  private val selectQuery = cqlt"select id, data from ${Const(tableName)} where id = ${Put[Int]}".as[Model]

  def apply[F[_]: Async](session: CassandraSession[F]) = for {
    insert <- insertQuery.prepare(session)
    select <- selectQuery.prepare(session)      
  } yield new Dao[F] {
    override def put(value: Model) = insert(value.id, value.data).execute.void
    override def get(id: Int) = select(id).config(_.setExecutionProfileName("default")).select.head.compile.last
  } 
} 

Handling optional fields (null)

By default, cassandra4io encodes Option as a null value. Which is ok for most cases. But in Cassandra, there is a difference between a null value and an empty value. In java driver this difference is represented by BoundStatement#setToNull (default behavior) and BoundStatement#unset (setting an empty field). The main advantage of using unset instead of setToNull is that tombstone will not be created for an empty field.

To use the unset instead of the setToNull for your optional value in a cql interpolators you could add .usingUnset to your optional value. Like in the following example:

import com.ringcentral.cassandra4io.cql._

cql"insert into entities(foo, bar, baz) values (${e.foo}, ${e.bar}, ${e.baz.usingUnset}"

User Defined Type (UDT) support

Cassandra4IO provides support for Cassandra's User Defined Type (UDT) values. For example, given the following Cassandra schema:

create type basic_info(
    weight double,
    height text,
    datapoints frozen<set<int>>
);

create table person_attributes(
    person_id int,
    info frozen<basic_info>,
    PRIMARY KEY (person_id)
);

Note: frozen means immutable

Here is how to insert and select data from the person_attributes table:

final case class BasicInfo(weight: Double, height: String, datapoints: Set[Int])
object BasicInfo {
  implicit val cqlReads: Reads[BasicInfo]   = FromUdtValue.deriveReads[BasicInfo]
  implicit val cqlBinder: Binder[BasicInfo] = ToUdtValue.deriveBinder[BasicInfo]
}

final case class PersonAttribute(personId: Int, info: BasicInfo)

We provide a set of typeclasses (FromUdtValue and ToUDtValue) under the hood that automatically convert your Scala types into types that Cassandra can understand without having to manually convert your data-types into Datastax Java driver's UdtValues.

class UDTUsageExample[F[_]: Async](session: CassandraSession[F]) {
  val data = PersonAttribute(1, BasicInfo(180.0, "tall", Set(1, 2, 3, 4, 5)))
  val insert: F[Boolean] =
    cql"INSERT INTO cassandra4io.person_attributes (person_id, info) VALUES (${data.personId}, ${data.info})"
            .execute(session)

  val retrieve: fs2.Stream[F, PersonAttribute] = 
    cql"SELECT person_id, info FROM cassandra4io.person_attributes WHERE person_id = ${data.personId}"
            .as[PersonAttribute]
            .select(session)
}

More control over the transformation process of UdtValues

If you wanted to have additional control into how you map data-types to and from Cassandra rather than using FromUdtValue & ToUdtValue, we expose the Datastax Java driver API to you for full control. Here is an example using BasicInfo:

object BasicInfo {
  implicit val cqlReads: Reads[BasicInfo] = Reads[UdtValue].map { udtValue =>
    BasicInfo(
      weight = udtValue.getDouble("weight"),
      height = udtValue.getString("height"),
      datapoints = udtValue
        .getSet[java.lang.Integer]("datapoints", classOf[java.lang.Integer])
        .asScala
        .toSet
        .map { int: java.lang.Integer => Int.unbox(int) }
    )
  }

  implicit val cqlBinder: Binder[BasicInfo] = Binder[UdtValue].contramapUDT { (info, constructor) =>
    constructor
      .newValue()
      .setDouble("weight", info.weight)
      .setString("height", info.height)
      .setSet("datapoints", info.datapoints.map(Int.box).asJava, classOf[java.lang.Integer])
  }
}

Please note that we recommend using FromUdtValue and ToUdtValue to automatically derive this hand-written (and error-prone) code.

Interpolating on CQL parameters

Cassandra4IO Allows you to interpolate (i.e. using string interpolation) on values that are not valid CQL parameters using ++ or concat to build out your CQL query. For example, you can interpolate on the keyspace and table name using the cqlConst interpolator like so:

val session: CassandraSession[IO] = ???
val keyspaceName = "cassandra4io"
val tableName    = "person_attributes"
val keyspace     = cqlConst"$keyspaceName."
val table        = cqlConst"$tableName"

def insert(data: PersonAttribute) = 
  (cql"INSERT INTO " ++ keyspace ++ table ++ cql" (person_id, info) VALUES (${data.personId}, ${data.info})")
    .execute(session)

This allows you (the author of the application) to feed in parameters like the table name and keyspace through configuration. Please be aware that you should not be taking your user's input and feeding this into cqlConst as this will pose an injection risk.

References

• Datastax Java driver

License

Cassandra4io is released under the Apache License 2.0.