Does the @geodistsearch use any geospatial indexes (for example, R-trees) for performance?
I am interested in the case where the anchor is constant, and each document has its own latitude / longitude pair, stored in radians.
I tried to figure this out from the Sphinx source code, but could not find any mention of any spatial index. If indexes are not used for geospatial search, how is performance ensured?
Does Sphinx perform a full scan if there are no keywords?
Reference Information. We have a dataset of over 100 million short entries. Some of the newly added items will be stored in latitude / longitude. Millions of entries are added every day. I predict that around 5-10% of newly added records will have location information.
Our goal is to implement a spatial location search with record support for queries like "get all records within 100 meters around the anchor point", "get 100 closest records around the anchor point" and without keyword searches.
Some googling have brought this forum thread back , which suggests using an artificial mesh index for performance. Is this still true?