Andre Poenitz <[EMAIL PROTECTED]> writes: > You are certainly right when it comes to Model/View separation. The > dimension cache is a 'view thing' (depends e.g. on font size), so it > does not really belong _inside_ the insets. Also, insets that are way > off the visible area do not need it at all. > > What kept me away from an 'external' solution so far was the > perceived cost as we would have a O(log n) vs O(1) lookup for > dimension data. With n being roughly the number of visible insets this > might not even be noticable, though.
Hashtables work wonders in those cases. The standard rb-tree implementation of sets and maps is way slower. /Andreas
