It actually works fairly well! How big is the dataset? Being more homogeneous works better (gaps in indexes are always tough)

Comments