On Jun 24, 2010, at 6:31 AM, Ralph Heinkel wrote: > Hi Michael, > > We have also tried the /*+ FIRST_ROWS(N) */ optimization hint, it > only gave a 25% speed improvement, but the result was still 5 or 7 > times slower than the ROW_NUMBER() OVER approach. > I'll provide benchmark details on Monday, also details about table > (actually a view) layout, indices, etc. > > On Jun 23, 6:04 pm, Michael Bayer <mike...@zzzcomputing.com> wrote: >> >> A full history of this feature is here: >> >> http://www.sqlalchemy.org/trac/ticket/536 >> >> The rationale is based on the bug described in that ticket, as well as that >> we preferred to go with an approach that was recommended by a lead engineer >> at Oracle. >> >> The dialect includes an option to add the /*+ FIRST_ROWS(N) */ directive, by >> specifying the "optimize_limits" keyword to create engine - we originally >> had that in the query in all cases, until some folks chimed in that we >> shouldn't make that decision by default. I don't know if that helps your >> use case. >> >> The previous system can be restored using a @compiles directive. I have >> documented that recipe >> athttp://www.sqlalchemy.org/trac/wiki/UsageRecipes/OracleRowNumberOver. > > I don't understand yet how the @compiles directive works, but I'm > also not that familiar with SA internals. Could you point me to some > URL where this is described?
http://www.sqlalchemy.org/docs/reference/ext/compiler.html -- You received this message because you are subscribed to the Google Groups "sqlalchemy" group. To post to this group, send email to sqlalch...@googlegroups.com. To unsubscribe from this group, send email to sqlalchemy+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/sqlalchemy?hl=en.