[xmonad] darcs patch: switch to nested stacks

[David Roundy]

On Sun, Dec 23, 2007 at 01:37:10PM -0800, Don Stewart wrote:
> droundy:
> > Hi all,
> > 
> > Here's a patch that switches to nested stacks in xmonad core.  I haven't
> > converted the test suite, because I've got quickcheck 2 installed, and
> > therefore can't compile the test suite, and I don't intend to uninstall
> > QC2.  If this code works as intended (I haven't tried running it either),
> > then it shouldn't change xmonad behavior.  I haven't adjusted the
> > LayoutClass to handle nested layouts, so there is considerable work to be
> > done (and not by me, I really should be focussing on darcs, but compiling
> > takes a while, particularly when I have to do actual work on my laptop due
> > to a sketchy internet connection).
> > 
> > But I think my ideas are down there.  It's a reasonably simple data
> > structure.  And it ought to be a clean solution for handling the float
> > layer, besides allowing nested layouts to be implemented cleanly.  I
> > haven't implemented full navigation within nested layouts, but that's not
> > hard to implement, and can also be implemented in contrib (or later), if
> > this is accepted.
> > 
> > The other reason I've put so little work on this, is that it's not at all
> > clear whether the core xmonad developers would accept a patch implementing
> > this feature, regardless of how well-tested, well-documented, etc it is.
> > So that's what we really need to hear (we being those of us who want this
> > feature, especially those who would be willing to take over its
> > development).
> 
> Thanks David. We can't look a this for 0.6 (coming out after new
> year's).
> 
> However, I'm heading off on vacation for two weeks, and will play with
> this. We've a long open ticket for fixing the floating layer, so this
> looks like a good start.

Alas, I've since realized that it's not obvious how this'd work out for the
floating layer.  As long as both layers are populated, it's simple:  you've
got two nested nests, and by convention the first one is tiled, the second
floating.  The trouble comes when one of them is unpopulated:  there's no
way to distinguish whether it's the floating or tiled layer that's empty,
since by construction each Nest is non-empty.  One option would be to allow
empty Nests, which would make them much more different from Stacks than the
implementation I sent in.  I suspect that's the "correct" implementation.
Then instead of a Maybe (Nest a) a Workspace would just hold a (Nest a).
And nest2stack would have type Nest a -> Maybe (Stack a).
-- 
David Roundy
Department of Physics
Oregon State University