From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
---|---|
To: | Dimitri Fontaine <dfontaine(at)hi-media(dot)com>, pgsql-hackers(at)postgresql(dot)org, "Pavel Stehule" <pavel(dot)stehule(at)gmail(dot)com>, "Gregory Stark" <stark(at)enterprisedb(dot)com>, "Robert Haas" <robertmhaas(at)gmail(dot)com>, "Martijn van Oosterhout" <kleptog(at)svana(dot)org> |
Subject: | Re: WIP patch: convert SQL-language functions to return tuplestores |
Date: | 2008-10-31 16:16:34 |
Message-ID: | 5036.1225469794@sss.pgh.pa.us |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
With the attached patch, SQL functions support returning the results of
INSERT/UPDATE/DELETE RETURNING clauses. An INSERT/UPDATE/DELETE
statement is always executed to completion before returning (including
firing any related triggers or rules), so we always materialize the
RETURNING output. When the function result comes from a SELECT, we
continue to use value-per-call mode when being called from
ExecMakeFunctionResult (thus preserving the prior behavior if the
calling query doesn't execute to completion). When called from
ExecMakeTableFunctionResult, we materialize the output and return the
tuplestore in one call. This is also the same behavior as before,
since ExecMakeTableFunctionResult would've filled a tuplestore anyway;
but it's noticeably faster because a lot of call/return and executor
entry overhead is eliminated.
This is code-complete but I haven't touched the docs yet. Barring
objections, I plan to commit when I've finished fixing the docs.
regards, tom lane
Attachment | Content-Type | Size |
---|---|---|
sql-functions-3.patch.gz | application/octet-stream | 14.9 KB |
From | Date | Subject | |
---|---|---|---|
Next Message | Joshua D. Drake | 2008-10-31 16:17:23 | Re: Enabling archive_mode without restart |
Previous Message | David Fetter | 2008-10-31 16:10:05 | Re: Window Functions: patch for CommitFest:Nov. |