Date: Tue, 5 Jun 2007 01:23:12 -0400
From: mark@mark.mielke.cc
To: Jason Lustig <lustig@brandeis.edu>
Cc: pgsql-performance@postgresql.org
Subject: Re: Question about SQL performance
Message-ID: <20070605052312.GA16345@mark.mielke.cc>
References: <F84FCD04-CB11-4CA2-838A-0ABA75A7F1A2@brandeis.edu>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <F84FCD04-CB11-4CA2-838A-0ABA75A7F1A2@brandeis.edu>
User-Agent: Mutt/1.5.14 (2007-02-12)

On Mon, Jun 04, 2007 at 11:18:30PM -0400, Jason Lustig wrote:
> I have some questions about the performance of certain types of SQL  
> statements.
> 
> What sort of speed increase is there usually with binding parameters  
> (and thus preparing statements) v. straight sql with interpolated  
> variables? Will Postgresql realize that the following queries are  
> effectively the same (and thus re-use the query plan) or will it  
> think they are different?
> 
> 	SELECT * FROM mytable WHERE item = 5;
> 	SELECT * FROM mytable WHERE item = 10;
>
> Obviously to me or you they could use the same plan. From what I  
> understand (correct me if I'm wrong), if you use parameter binding -  
> like "SELECT * FROM mytable WHERE item = ?" - Postgresql will know  
> that the queries can re-use the query plan, but I don't know if the  
> system will recognize this with above situation.

Although they could use the same plan, it is possible that using the
same plan is non-optimal. For example, if I know that 99% of the table
contains item = 5, but only 1% of the table contains item = 10, then
the 'best plan' may be a sequential scan for item = 5, but an index scan
for item = 10.

In the case of a prepared query, PostgreSQL will pick a plan that will
be good for all values, which may not be best for specific queries. You
save parsing time and planning time, but may risk increasing execution
time.

> Also, what's the difference between prepared statements (using  
> PREPARE and EXECUTE) and regular functions (CREATE FUNCTION)? How do  
> they impact performance? From what I understand there is no exact  
> parallel to stored procedures (as in MS SQL or oracle, that are  
> completely precompiled) in Postgresql. At the same time, the  
> documentation (and other sites as well, probably because they don't  
> know what they're talking about when it comes to databases) is vague  
> because  PL/pgSQL is often said to be able to write stored procedures  
> but nowhere does it say that PL/pgSQL programs are precompiled.

I think you can find all of these answers in the documentation, including
my comments about prepared queries. Does it matter if the program is
precompiled? I believe it is, but why would it matter?

Are you addressing a real performance problem? Or are you trying to avoid
issues that you are not sure if they exist or not? :-)

Prepared queries are going to improve performance due to being able to
execute multiple queries without communicating back to the
client. Especially for short queries, network latency can be a
significant factor for execution speed.

Cheers,
mark

-- 
mark@mielke.cc / markm@ncf.ca / markm@nortel.com     __________________________
.  .  _  ._  . .   .__    .  . ._. .__ .   . . .__  | Neighbourhood Coder
|\/| |_| |_| |/    |_     |\/|  |  |_  |   |/  |_   | 
|  | | | | \ | \   |__ .  |  | .|. |__ |__ | \ |__  | Ottawa, Ontario, Canada

  One ring to rule them all, one ring to find them, one ring to bring them all
                       and in the darkness bind them...

                           http://mark.mielke.cc/