Will parallelability be inherited?

Hi, Curious if there's anything special we'll need to do when we make a custom collection subtype in 2.9 to get .par to work on those subtypes. For example I have 22 Seq subtypes lined up for the next release of ScalaTest. They are code-generated so it is hard to point to them in SVN. Here's one of them: class TableFor2[A, B](val heading: (String, String), rows: (A, B)*) extends IndexedSeq[(A, B)] with IndexedSeqLike[(A, B), TableFor2[A, B]] { /** * Selects a row of data by its index. */ def apply(idx: Int): (A, B) = rows(idx) /** * The number of rows of data in the table. (This does not include the heading tuple) */ def length: Int = rows.length /** * Creates a new Builder for TableFor2s. */ override protected[this] def newBuilder: Builder[(A, B), TableFor2[A, B]] = new ArrayBuffer mapResult { (buf: Seq[(A, B)]) => new TableFor2(heading, buf: _*) } /** * Applies the passed property check function to each row of this TableFor2. * *

* If the property checks for all rows succeed (the property check function returns normally when passed * the data for each row), this apply method returns normally. If the property check function * completes abruptly with an exception for any row, this apply method wraps that exception * in a TableDrivenPropertyCheckFailedException and completes abruptly with that exception. Once * the property check function throws an exception for a row, this apply method will complete * abruptly immediately and subsequent rows will not be checked against the function. *

* * @param fun the property check function to apply to each row of this TableFor2 */ def apply(fun: (A, B) => Unit) { for (((a, b), idx) <- rows.zipWithIndex) { try { fun(a, b) } catch { case e => val (aName, bName) = heading throw new TableDrivenPropertyCheckFailedException( FailureMessages("propertyException", UnquotedString(e.getClass.getSimpleName)) + "\n" + " " + FailureMessages("thrownExceptionsMessage", if (e.getMessage == null) "None" else UnquotedString(e.getMessage)) + "\n" + ( e match { case sd: StackDepth if sd.failedCodeFileNameAndLineNumberString.isDefined => " " + FailureMessages("thrownExceptionsLocation", UnquotedString(sd.failedCodeFileNameAndLineNumberString.get)) + "\n" case _ => "" } ) + " " + FailureMessages("occurredAtRow", idx) + "\n" + " " + aName + " = " + a + "," + "\n" + " " + bName + " = " + b + "\n" + " )", Some(e), getStackDepthForPropCheck("TableDrivenPropertyChecks.scala", "forAll"), FailureMessages("undecoratedPropertyCheckFailureMessage"), List(a, b), List(aName, bName), idx ) } } } /** * A string representation of this object, which includes the heading strings as well as the rows of data. */ override def toString: String = stringPrefix + "(" + heading.toString + ", " + rows.mkString(", ") + ")" } I think what should happen when someone invokes .par on this that the result should run in parallel methods like map, etc., that you'd expect to run in parallel. My apply method that does the property check stopping on the first failure would continue to run sequentially, which is what i want, because I don't invoke .par on rows. I'm I correct in my surmise that I won't need to do anything else to get .par to work on TableFor2? Thanks. Bill ---- Bill Venners Artima, Inc. http://www.artima.com