Op 13 sep 2015, om 15:17 heeft Luuk het volgende geschreven: > > > On 13-09-15 14:57, E.Pasma wrote: >> The rowcounts were: >> >> 1) 999 - aurel's original version is 3.8.9 >> 2) 1000 - Luuks non-recursive version, same for my own trial >> 3) 837 -the version with temp table >> 4) 838 - DRH;s version >> >> and there is nothing to worry here. >> The main difference is explained by a mistake in the format >> parameter to the strftime function. In (1) and (2) that was '%%w' >> which is meaningles. In (4) this is changed to '%w' and that >> effectively filters out the sundays. Thanks DRH for the comments >> that clarify this intention. >> The remaining difference is because in (3) the final date is not >> incuded as start date on a seperate line. Still it is already there >> as end date. >> > > i don't like Sundays.... > (https://www.youtube.com/watch?v=fl6I-D8GXqA) > > sqlite> select count(*) from securities where > strftime('%w',calc_date)<>0 ; > 6301006 > Run Time: real 3.282 user 2.796875 sys 0.437500 > sqlite> select count(*) from securities where strftime('% > %w',calc_date)<>0 ; > 6301006 > Run Time: real 3.187 user 2.640625 sys 0.546875 > sqlite> > > ;) Still nothing to worry. I just learned that the strftime function returns '0' for Sundays, not 0, and that 0<>'0'.
That leaves us with the topic of recursive aggregate queries.
