> -----Original Message-----
> From: [email protected] [mailto:sqlite-users-
> [email protected]] On Behalf Of Richard Hipp
> Sent: vrijdag 30 augustus 2013 21:41
> To: General Discussion of SQLite Database
> Subject: Re: [sqlite] Inefficient covering index used for Subversion with
> SQLite 3.8.0
>
> On Fri, Aug 30, 2013 at 3:31 PM, Bert Huijben <[email protected]> wrote:
>
> > The analyze on the very small database (which I used for the comparison
> > between 3.7 and 3.8) is:
> >
>
> Thanks for the data.
I can share the/an actual database if that would help. (The wc.db describes
a checkout of a 100% public subversion repository, so there is nothing
secret in it)
For myself it would be interesting to know how I should look at the OR
optimization which was available for our use cases between Sqlite 3.7.12 and
3.7.18.
My question is more like: Should I see this as a light regression that most
likely will be resolved in a future version, or as a buggy assumption in our
code that I should work around? It should be possible via explicit 'INDEXED
BY' clauses and/or breaking the query in separate parts. (Or for future
versions by pre-filling the stat table)
Bert
_______________________________________________
sqlite-users mailing list
[email protected]
http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
