David Bremner <david@tethera.net> writes: > It seems noticeably faster (on the order of 30-50% faster) and the > code is quite a bit simpler to adapt the approach in [1] to only > delete the terms we are going to re-add via indexing. > > This obsoletes the previous series at [2]. It still has all of the > issues mentioned there UI-wise, and the question of the index options > design probably needs more thought. > Some belated testing reveals this implimentation is pretty broken. It probably won't eat your database, but that's only because I forgot to add a call to _notmuch_message_sync. So I'd recommend passing on this for now. The previous approach is probably OK, although I'm going to bash at this fancier approach a bit to see if I can make it work.