On Thu, Sep 12, 2013 at 5:29 AM, Yeb Havinga <yebhavinga@gmail.com> wrote:
> Is the following known behaviour, or should I put some time in writing a
> self contained test case?
>
> We have a function that takes a value and returns a ROW type. With the
> function implemented in language SQL, when executing this function in a
> large transaction, memory usage of the backend process increases.
> MemoryContextStats showed a lot of SQL function data. Debugging
> init_sql_fcache() showed that it was for the same function oid each time,
> and the oid was the function from value to ROW type.
>
> When the function is implemented in PL/pgSQL, the memory usage was much
> less.
>
> I'm sorry I cannot be more specific at the moment, such as what is 'much
> less' memory with a PL/pgSQl function, and are there as many SQL function
> data's as calls to the SQL function, because I would have to write a test
> case for this. I was just wondering, if this is known behavior of SQL
> functions vs PL/pgSQL functions, or could it be a bug?
It sounds like a bug to me, although I can't claim to know everything
there is to know about this topic.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company