Exception during macro expansion: StackOverflowError

[bennythomps]

Making some basic calls, I'm getting exceptions when I compile:

[error] ApiKey.scala:19: exception during macro expansion: java.lang.StackOverflowError
[error]   def create: (ApiKey) => Future[ApiKey] = withSchema {
[error]                                                       ^
{
  import _root_.troy.driver.InternalDsl._;
  import _root_.troy.driver.codecs.PrimitivesCodecs._;
  val prepared = implicitly[com.datastax.driver.core.Session].prepare("\n            SELECT api_key, name, active\n            FROM api.api_keys\n            WHERE api_key = ?\n        ");
  ((apiKey: java.util.UUID) => {
    def parser(row: _root_.com.datastax.driver.core.Row) = ApiKey(column[java.util.UUID](0)(row).as[CDT.Uuid], column[String](1)(row).as[CDT.Text], column[Boolean](2)(row).as[CDT.Boolean]);
    bind(prepared, param(apiKey).as[CDT.Uuid]).executeAsync(ApiKeyRecord.this.session, scala.concurrent.ExecutionContext.Implicits.global).oneOption(scala.concurrent.ExecutionContext.Implicits.global).parseAs(parser)
  })
}

Here's my schema:

CREATE KEYSPACE api WITH replication = {'class': 'SimpleStrategy' , 'replication_factor': '1'};

CREATE TABLE api.api_keys (
  api_key uuid,
  name text,
  active boolean,
  PRIMARY KEY (api_key)
);

Here's my calling code that's "breaking"

class ApiTokenRecord(implicit session: Session) {

  def create = withSchema { (token: UUID, apiKey: UUID, email: String, expiration: DateTime) =>
    cql"""
      INSERT INTO kiko.api_tokens (api_token, api_key, email, expiration_time)
      VALUES ($token, $apiKey, $email, $expiration)
      """.prepared.executeAsync.map(_ => true)
  }
}

Am I doing something wrong?

[tabdulradi]

Hello @bennythomps,

Thanks for reporting this bug.
It is .map(_ => true) that triggered it. A workaround (for now) is to rewrite your example like the following:

class ApiTokenRecord(implicit session: Session) {
  def createStatement = withSchema { (token: UUID, apiKey: UUID, email: String, expiration: DateTime) =>
    cql"""
      INSERT INTO api.api_keys (api_token, api_key, email, expiration_time)
      VALUES ($token, $apiKey, $email, $expiration)
      """.prepared.executeAsync
  }
  def create(token: UUID, apiKey: UUID, email: String, expiration: DateTime) =
    createStatement(token, apiKey, email, expiration).map(_ => true)
}

I'll mark this as bug for now. We are currently replacing the current macro implementation with a cleaner one using Scala meta. So hopefully this bug should vanish soon.

[bennythomps]

Thanks @tabdulradi! What about the .map(_ => true) is breaking it? It's not the permanent solution for me, so what to avoid when shifting will help avoid the problem until the meta stuff is available.

[tabdulradi]

In general avoid writing arbitrary code inside withSchema block. We only tested for specific cases, like setConsistencyLevel, the current Tree transformation code is not robust.

For the specific case of .map with a lambda, even if the transformation got fixed, still won't work due to a bug in Macros.

You can have a peek on the new scala.meta based implementation here