On Thu, Jul 28, 2005 at 05:05:14PM +0200, Juergen Spitzmueller wrote: > Problem description: > From tabular's point of view, there are two copy stacks. The "normal" one for > ordinary text and math copying and tabulars's own stack (namely > paste_tabular_) which holds multiple cell contents. > Now LFUN_PASTE in insettabular just checks if there is a non-empty > paste_tabular_. If yes, its contents will be pasted. It doesn't care whether > the "normal" stack is more recent or not. In fact, it just does not know > that. > So once you have copied several cells, the outside copy stack is not > accessible anymore from within tabular (including the tabular cell's > insettext contents). > > One solution is to tell insettabular whether the ordinary copy stack is newer > than paste_tabular_. The attached patch does exactly that, adding a flag > tabular_stack_dirty_ to Cut and Paste. > > I'm not sure if this is the best approach, or a good approach at least, so > please object.
A proper solution should not use this divided logic but some kind of 'grid' to hold a selection. In most cases this would be just a 1x1 grid but in grid like structures like tables and arrays it'd be more. The grid needs to be only conceptionaly, i.e. we don't necessarily need a grid structure for that. IIRC math already does it that way and uses some kind of serialization of the gid and just stores this serialized data. Andre'
