On Mon, May 23, 2016 at 12:01 PM, Jeff Davis <pgsql@j-davis.com> wrote:
On Fri, May 20, 2016 at 1:41 PM, David G. Johnston <david.g.johnston@gmail.com> wrote: > How does the relatively new FILTER clause play into this, if at all?
My interpretation of the standard is that FILTER is not allowable for a window function, and IGNORE|RESPECT NULLS is not allowable for an ordinary aggregate.
So if we support IGNORE|RESPECT NULLS for anything other than a window function, we have to come up with our own semantics. > We already have "STRICT" for deciding whether a function processes nulls. > Wouldn't this need to exist on the "CREATE AGGREGATE"
STRICT defines behavior at DDL time. I was suggesting that we might want a DDL-time flag to indicate whether a function can make use of the query-time IGNORE|RESPECT NULLS option. In other words, most functions wouldn't know what to do with IGNORE|RESPECT NULLS, but perhaps some would if we allowed them the option.
Perhaps I didn't understand your point?
The "this" in the quote doesn't refer to STRICT but rather the non-existence feature that we are talking about.
I am trying to make the point that the DDL syntax for "RESPECT|IGNORE NULLS" should be on the "CREATE AGGREGATE" page and not the "CREATE FUNCTION" page.
As far as a plain function cares its only responsibility with respect to NULLs is defined by the STRICT property. As this behavior only manifests in aggregate situations the corresponding property should be defined there as well.
