Re: Overlays as an AA-tree

[Stefan Monnier]

> Ideally, I think, this last detail should be of no consequence and the
> implementation should only be limited by the number of overlays and not
> by where in the buffer they occur.

Luckily, it's not currently a problem we really face.  Instead the
problem we face is how to handle a large number of overlays in a buffer,
but where any given buffer position is only covered by a modest number
of overlays.

The problem of handling efficiently a large number of overlays covering
a given position seems pretty hard to solve, actually.

> I think I'm going to relax now.

BTW, in your new overlay representation, how many bytes (or words) does
an overlay cost (in the master tree, if I count right, an overlay costs
3 Lisp_Misc objects, so that should be 3x 6words, i.e. 72 bytes on
a 32bit system)?

I'd expect the new representation to be no larger (the current
representation has a lot of redundancy).  But I'm curious because,
depending on the answer, the whole Lisp_Misc thingy might want to
be reconsidered.


        Stefan