I spent a wee while yesterday kicking the tyres on the Play 2.0 framework with Scala. Aside: I think that concurrency/asynchronous processing is really important in the age of multi-core CPUs, and that Akka’s actor-based system is really interesting, and Play uses Akka 2 under the hood, so I’m becoming really interested in becoming proficient in Scala & Play 2. That said, so far I’m struggling to come to terms with Scala’s type system, and it reminds of the struggles I had with ML at University, so this could be an uphill battle!

Anyway, that’s not what this is about. The toy application I’m trying to build (which I’ll share if and when I’m done!) involves storing locations (a latitude and longitude) as decimals. I’ve got my table (stored as an evolution) along the lines of:

CREATE TABLE sites (
  id SERIAL PRIMARY KEY,
  name VARCHAR(255) UNIQUE NOT NULL,
  latitude NUMERIC NOT NULL,
  longitude NUMERIC NOT NULL,
  UNIQUE(latitude, longitude)
);

(I’m using PostgreSQL.) And here’s the first implementation of the model, app/models/Site.scala, which just has a case class to encapsulate the table data, and a singleton object with a finder to return a sequence of site (so that I can list them in a table):

package models
import play.api.db._
import play.api.Play.current
import anorm._
import anorm.SqlParser._

case class Site(
  id: Pk[Long],
  name: String,
  latitude: BigDecimal,
  longitude: BigDecimal
)

object Site {
  val allFieldsParser = {
    get[Pk[Long]]("sites.id") ~
    get[String]("sites.name") ~
    get[BigDecimal]("sites.latitude") ~
    get[BigDecimal]("sites.longitude") map {
      case id ~ name ~ latitude ~ longitude =>
        Site(id, name, latitude, longitude)
    }
  }

  def findAll(): Seq[Site] = {
    DB.withConnection { implicit connection =>
      SQL("SELECT * FROM sites").as(Site.allFieldsParser *)
    }
  }
}

(Totally cargo-culted, BTW.) This means that in the controller I should be able to call Site.findAll() and get back a sequence of sites I can then use to map to HTML fragments. Win.

But I’m getting a compilation error! In the browser, it’s saying could not find implicit value for parameter extractor: anorm.Column[BigDecimal] and pointing to the error being on get(BigDecimal)("sites.latitude") in the parser.

I couldn’t find any documentation on retrieving decimals from a database with anorm (which is mostly why I’m writing this) and looking through the source code didn’t exactly elucidate. However, after a bit of trial and error, I discovered that it works if I add:

import java.math.BigDecimal

at the start of the model file.

So, if you’re having trouble making BigDecimal columns work in your Anorm-based models, remember to explicitly import java.math.BigDecimal. Hopefully this will find its way into Google, making the next Scala newbie’s introduction a little less painful. :)

I’d love if somebody could explain why. Perhaps it would improve my understanding of Scala!

Update Aha, I have a theory. I think that, by default, scala.math.BigDecimal is available in my model’s scope. And scala.math.BigDecimal isn’t a subtype of java.math.BigDecimal, which is what Anorm deals in. So importing the Java version overrides the Scala version and makes the type system happy.

Now I’m wondering: why the two unrelated (in type-land anyway) versions of BigDecimal? Which should my apps be using? Which should Anorm be exposing?