A Scala Corrections Library

Is the goal solely to avoid having to explicitly state the type of the target collection? It seems to me that assuming the output collection is likely to be of the same type as the input collection is... frankly... wrong.

Clojure's approach makes a lot more sense to me: Coerce to lazy sequences & chain together O(N) operations by operating on sequences, then require explicit reconstitution of the target type. Using something like (into #{} (f (g (h #{:some :set :items})))) manages to 1) avoid stating the target type (by using an empty set literal) 2) support processing into an existing non-empty collections and 3) avoid the overhead of re-constructing more complex data structures for intermediate steps (at the cost of lazy sequence construction), 4) is trivially extensible to new target types (implement conj) and 5) plays nice with fast transient object construction (implement transient, persistent!, and conj!). Then with reducers and monoidal operations, you can switch from sequential algorithms to parallel algorithms without paying the lazy sequence cost.

Scala's model buys me what exactly? I don't have to write (vec ...) or (into (my-data-structure) ...) or whatever? Seems a steep complexity price to pay for such a trivial and questionable feature.

Am I missing something?