Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!ucsd!ucbvax!tut.cis.ohio-state.edu!cs.utexas.edu!uunet!mcsun!sunic!tut!ra!tucos!moj
From: scc@cl.cam.ac.uk (Stephen Crawley)
Newsgroups: comp.lang.clu
Subject: Re: Recursive iterator in CLU?
Keywords: iterators
Message-ID: <934@scaup.cl.cam.ac.uk>
Date: 3 Oct 89 22:53:12 GMT
References: <730@tuvie>
Sender: moj@tucos.UUCP
Distribution: inet
Organization: U of Cambridge Comp Lab, UK
Lines: 47
Approved: moj@cs.utu.fi
X-Posting-Number: 49

Ulrich Neumerkel writes:
> CLU's explicit concept of iterators seems at the first glance preferable
> to the rather ad hoc way iterators are defined in oo -languages. (As
> a method of an object, or as a friend-class.) Instead of handling
> the whole state of the iterator explicitely, one can imagine the yield
> statement as yust a "print" statement which will pe "piped" to the caller.
> But this works for simple examples only. If you want to traverse e.g. a 
> binary tree you have to call within the iterator the same iterator for 
> both left and right subtree. In my opinion it would be more intuitive 
> to use the "print" methapher for all kinds of calls.

Recursive iterators are no problem in CLU.  For example: 

tree_node = cluster [T: type] is tree_walk ...

rep = struct [
  left:		node_or_leaf,
  right:	node_or_leaf]
  
node_or_leaf = oneof [
  a_node:	tree_node,
  a_leaf:	T]

tree_walk = iter (node: cvt) yields (T)
  tagcase node.left
    tag a_node (n: tree_node):
      for leaf: T in tree_walk(n) do yield (leaf) end
    tag a_leaf (leaf: T):
      yield (leaf)
    end
  tagcase node.right
    tag a_node (n: tree_node):
      for leaf: T in tree_walk(n) do yield (leaf) end
    tag a_leaf (leaf: T):
      yield (leaf)
    end
  end
  
..

end tree_node

Frankly, I don't see what is non-intuitive about this use of recursion.  
The CLU compiler may do a poor job of optimising recursive iterators ... 
but that is a different issue entirely.

-- Steve