Not sure what the implications of this would be, but we could reduce the number of queries if we simplify the forced fits procedure by not querying the database for the forced fits (missing blind detections) but just use the previously found sources in the image.
the database is the slowest part in the whole pipeline, no matter what what backend we use. if we could somehow minimize the number of queries to only inserts of extracted sources and image metadata we could make TraP probably much faster.
Not sure what the implications of this would be, but we could reduce the number of queries if we simplify the forced fits procedure by not querying the database for the forced fits (missing blind detections) but just use the previously found sources in the image.
the database is the slowest part in the whole pipeline, no matter what what backend we use. if we could somehow minimize the number of queries to only inserts of extracted sources and image metadata we could make TraP probably much faster.