Re: [RFC] supporting an alternative layout management algorithm

[Paolo Bonzini <bonzini gnu org>]

Matthias Clasen wrote:
> On Tue, Sep 2, 2008 at 1:24 PM, Behdad Esfahbod
>> These changes of your to GTK+ are quite similar to the "ExtendedLayout" work
>> that Mathias Hasselmann has done previously, though that's not merged into
>> GTK+ yet.  He also introduces an interface that adds height-for-width and
>> width-for-height methods for sophisticated size allocation.  That's a more
>> generic approach than yours from what I understand.  See:
>>
>>  http://live.gnome.org/MathiasHasselmann/NewLayoutManager
>>
>> Can you give it a test and verify that it is flexible enough for your usecase?
> 
> Yes, please.
> 
> We'll take a serious effort at integrating the width-for-height work
> in the next cycle, so making sure that it is flexible enough for your
> needs would be very valuable.

Yes, it seems feasible.  I think I can implement the flow container by
making a GtkBox subclass that asks for its size like this:

    __________
   |  B1      |
   |__________|
   |  B2  |
   |______|_
   |  B3    |
   |________|_
   |  B4      |
   |__________|

... but can trade "height for width" to pack more items on the same
line.  This probably is not covered by Matthias' work, and it should be
a worthwhile addition to the first GTK+ release that has extended layout.

And I think the GtkManagedLayout would be replaced simply by GtkViewport
-- but I don't see it on the list on the wiki page, and it might be that
GtkViewport needs to be made aware of extended layouts.

I'll check out the branch but anyway I'll ask, can the code be
"injected" into a current GTK+ release like I'm doing now or does it
touch the very innards of GTK+?

Thanks for the very quick turnaround,

Paolo