On Sun, Sep 28, 2008 at 11:22:02PM +0200, Pavel Sanda wrote: > hi, > > its constant frustration for me that it is not possible to keep > outliner tree view in some persistent state. i very often need to > uncollapse only few branches in my view to see the structure and move > between these places. the document is big enough that full > uncollapsing is not usable. > > it drives me crazy to uncollapse all these branches again and again > after each update of the outliner, so i would like to have some > persistent feature of outliner. > > i though about possible solutions: > > 1. to detect changes in structure of the document and update toc view > only in such a case. most simple would be some caching mechanism of > the last structure. > > 2. to remember the way tree was uncollapsed and after each rebuilt > make uncollapse it in the same way if the structure around haven't > changed to much. this start to suggest some diff-like algorithm and > will be almost impossible to get completely right. > > 3. provide some 'freeze' checkbox with handling of the outliner trying > to move in already deleted/changed parts of document. update will on > demand of refresh button. > > i tend for 3. solution but would like to hear your opinions.
I'd vote for 2. The tree items can be represented as 'path', the 'tree state' would be a set of such paths. All that would be needed would be a way to compute such a 'path' that's reasonably stable or easily 'fixable' under the document translations that are typically with the outliner. One such option would be the first few words from the section headings, or some kind of checksum etc... Andre'
