Yonik Seeley <ysee...@gmail.com> wrote: > Hmmm, well something is really wrong for this orders of magnitude > difference. I've never seen anything like that and we should > definitely try to get to the bottom of it.
This might be a wild goose chase, but... Zheng states it is a text field with the content of fairly large documents. This means a high amount of unique values and a gazillion references from documents to those values. When incrementing counters for String faceting, segment ordinal -> index ordinal mapping takes place. Legacy facets has a mechanism where temporary segment-specific counters are used. These are updated directly with the segment ordinals and the mapping to global ordinals is performed after the counting. As far as I can see, JSON Facets does not have this delayed mapping mechanism: Every increment requires a call to the segment->global-ordinal map. With a large field this map cannot be in the fast caches. Combine this with a gazillion references and it makes sense that JSON Facets is slower in this scenario. A factor 20 sounds like way too much though. I would have expected maybe 2. - Toke Eskildsen
