Thanks for the suggestion. We have 1.8 million docs in our cluster and it took quite a while to get them in because of all the nested objects and we don't want to have to reprocess them if we can avoid it. Is there a way to add the field to the index and then dynamically update all length fields with a script call or something?