Re: A few questions about desktop.el

[Richard M. Stallman]

    It's quite useful, and it'll be great the day we can make it
    save/restore window and frame configurations. But I don't think
    there's an easy way to turn a window configuration into elisp and back

It is straightforward to do walk-windows and record the size and
contents of each leaf window.  It's necessary in addition to record
the structure.  I think this can be deduced from that info as follows:

Notice when consecutive windows (in the standard window ordering) have
the same left and right edges, and when they have the same top and
bottom edges.  In those cases, you have vertical or horizontal
siblings.  So make a list of them and replace them with an item
that corresponds to the combination of them.

Do this over and over until you're left with one "window",
and you've reconstructed the whole tree.

We could add a primitive which constructs a window configuration
from a list of window sizes and contents.  If the above algorithm
fails to reduce the specified data to a single window, it would
signal an error.

(We could conceive of making window configurations transparent data
using this method.  This would make save-window-configuration slower;
would it be enough slowdown to cause a problem?  Perhaps nowadays
it would not be a problem.)