Mon Feb 21 09:17:37 EST 2011

Back to the Stack

* The "lexical" Staapl spin-off has come full circle[1]: I should
  switch to a _stack_ architecture for describing directed acyclic


What needs to be done is to make this mesh with the rest of Staapl.

The key choice is to separate the internal representation that
supports higher order functions, from the "user interface" that allows
the use of lexical variables on top of this.

The stack representation has explicit "dup".  Recovery of sharing
information is not possible in abstract-interpretation based Haskell
implementations; it requires a CPS-style approach.

The stack interface is already threaded.

To make this explicit, the sawtooth algorithm can be used as an
application pull.

[1] entry://../meta/20110126-092935