On 3/11/15 3:57 PM, Tom Lane wrote:
> Alvaro Herrera <alvherre@2ndquadrant.com> writes:
>> But autovacuum is still manufacturing a VacuumStmt by hand. If we want
>> to get rid of that, I think it'd work to have a new
>> ExecVacuum(VacuumStmt, params) function which is called from
>> standard_ProcessUtility and does just vacuum(rel, relid, params).
>> Autovacuum on the other hand can call vacuum() without having to
>> construct the parse node.
>
> Why would we want to get rid of that? A struct is a handy and legible
> way to pass a pile of parameters. I doubt it would be an improvement for
> vacuum() to grow a long list of separate parameters.
We're not exactly getting rid of it; Thomas' patch adds a second struct
that deals with detailed vacuum parameters that are not actually present
in VacuumStmt. These are things that are specific to autovac but not
manual VACUUM. But the patch in it's current form still have autovac
building a somewhat bogus VacuumStmt.
What's being proposed is to expose VacuumStmt (which only makes sense
for VACUUM) only where it's needed, and use VacuumParams everywhere
else. In particular, this means autovac will just deal with VacuumParams
and will no longer build a fake VacuumStmt.
--
Jim Nasby, Data Architect, Blue Treble Consulting
Data in Trouble? Get it in Treble! http://BlueTreble.com