Imported From: https://issues.scala-lang.org/browse/SI-5364?orig=1
Reporter: Tom Dong (tomtung)
Affected Versions: 2.9.1

@paulp said:
This is how it works.

Tom Dong (tomtung) said:
Can you please explain a little bit more? "Flat-mapping a map creates another map" does not make sense to me. To me, flat-mapping is breaking down the original collection, and concatenating the collections it contains to create a new collection. The original map does not contain any map, why should flatMap return a new Map, instead of List, which is what the original map actually contains?

@paulp said:
It doesn't matter what the map contains, or how you personally interpret flatmap - it matters that the operation is invoked on a map, and the result of the operation are elements which make another map. If you want a List, make it a List.

scala> m.toList flatMap (_._2)
res0: List[(Int, Int)] = List((1,11), (1,111), (2,22), (2,222))

Tom Dong (tomtung) said (edited on Jan 10, 2012 8:08:32 AM UTC):
Quote from "Programming in Scala" page 362:

(flatMap) applies the function to each list element and returns the concatenation of all function results.

So this is not my "personal" interpretation.

If you execute:
scala> m.map(_._2)
res10: scala.collection.immutable.Iterable[List[(Int, Int)]] = List(List((1,11), (1,111)), List((2,22), (2,222)))

You get a List of List, and concatenate them gives a List, not a Map.

"Operate on a collection, and get a new collection of the same type" seems to be a nice idea, but I don't think it automatically justifies the resultant semantic confusions. I think the problem here is that flatMapping a Map does not always return a Map. This happens only when the values of the map happens to be collections of 2-tuples - but 2-tuples are not exclusively used as Map elements. IMHO, this inconsistency can be problematic.

@paulp said:
I'm not saying it's perfect, I'm saying this is what is. It's fundamental to the design of the collections and changing it is way beyond what can be addressed in a ticket. Please take it up on one of the mailing lists.

Tom Dong (tomtung) said:
Oh, I see. I thought this were just an implementation issue of the flatMap operation. Thanks anyway.