From: Warren Young
Date: October 24 2007 11:21am
Subject: Re: Memory management using Res and Row
List-Archive: http://lists.mysql.com/plusplus/7115
Message-Id: <471F2ABD.2040104@etr-usa.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

Gary Anderson wrote:
> When the application exits, I seem to have a lot of 
> un-released memory (leaks) that appear to be holding items read from the 
> database during processing.  

Whenever you think you have a leak, it's essential that you take the 
next step, which is to correlate memory use with some work unit.  Does 
memory use go up as a function of time, or query count, or rows 
retrieved, or what?  If it remains fixed, it's "overhead", not a "leak".

There is known overhead in the MySQL C API library, and a memory 
debugger is going to show it as a "leak" because it doesn't know better. 
  There's a way to release it -- see the C API docs -- but since most 
programs have to access the database for their entire run time, there's 
not much point in doing it; any platform MySQL runs on is going to have 
decent memory management, so it releases unfreed memory at exit.  As a 
result, we don't bother to expose this as a feature in MySQL++.

> The manual indicates that a ResUse object 
> contains a purge() method to release the resources held by the object, 

This isn't generally needed.  It's more an implementation detail than 
anything else.  A quick glance at the source shows that ~ResUse() calls 
it, so if you have a leak due to not calling purge(), you really have a 
leak due to not destroying your ResUse objects.

> but I can't seem to find similar methods for Result, 

Result derives from ResUse.

> ...Row or Query.  Do such methods exist?

You mean other than the destructors?

> If not, how is this allocated memory freed?

The same way memory is freed in any C++ program.