Am 27.10.2017 um 21:59 schrieb David Raymond:
Also getting 31ms for both...
Thanks for testing that guys...
I was able to get similar timings (about 30msec), when the (quite large) resultset-output was delegated into a file... To avoid doing expensive File-IO in that test, I'd recommend to run it again with the two small Files (SQL1.txt and SQL2.txt), which I've now included in the ZipFile... Here again the Link: http://vbRichClient.com/Downloads/NWind.zip With these two Files, I get the following results (on a Win8.1 machine): sqlite> .read SQL1.txt Run Time: real 0.000 user 0.000000 sys 0.000000 Run Time: real 0.010 user 0.000000 sys 0.000000 Run Time: real 0.000 user 0.000000 sys 0.000000 sqlite> sqlite> .read SQL2.txt Run Time: real 0.000 user 0.000000 sys 0.000000 Run Time: real 0.006 user 0.000000 sys 0.000000 Run Time: real 0.000 user 0.000000 sys 0.000000 sqlite> Note: SQL1-Text is copying into an InMemory-Temp-Table from the Invoices-View - whilst SQL2 is doing the same from "directly given SQL" (the same as used for the ViewDef). The timings in both cases vary about +-1msec - the values as posted represent the median-values. Not really factor 2 (as encountered when copying into an inmemory-resultsets of my wrapper-library) - but nearly so.
Though looking at the explain output it looks for the view/subroutine/subquery version it's spending like extra machine steps copying every result row from one set of registers to another.
Yep, that Extra-copying of the RowValues seems to be the culprit (the Select has a decent Field-Count - and these efforts seem to sum-up over the scanned Records then). Olaf _______________________________________________ sqlite-users mailing list sqlite-users@mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
