On 5/28/18 13:17, Bruce Momjian wrote: > On Thu, May 10, 2018 at 12:55:23PM -0400, Tom Lane wrote: >> Peter Eisentraut <peter.eisentr...@2ndquadrant.com> writes: >>> I think that last part isn't actually written down anywhere. (It only >>> states the converse.) How about a clarification like this: >> >>> @@ -271,7 +271,10 @@ <title id="sql-declare-notes-title">Notes</title> >>> and not use grouping or <literal>ORDER BY</literal>). Cursors >>> that are not simply updatable might work, or might not, depending on >>> plan >>> choice details; so in the worst case, an application might work in >>> testing >>> - and then fail in production. >>> + and then fail in production. If <literal>FOR UPDATE</literal> is >>> + specified, then the cursor is guaranteed to be updatable, or the >>> + <command>DECLARE</command> command will error if an updatable cursor >>> + cannot be created for the supplied query. >>> </para> >> >> OK by me, except we don't usually use "error" as a verb. Either "fail" >> or "throw an error" would read better IMO. Or you could just stop with >> "guaranteed to be updatable"; I don't think the rest adds much. > > I have done as you suggested and just used the first part; patch > attached and backpatched.
I think we should still add the second part, because it currently doesn't say anything about that a cursor declaration might fail if an updatable cursor cannot be created. -- Peter Eisentraut http://www.2ndQuadrant.com/ PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services