From: Warren Young
Date: March 5 2009 7:32pm
Subject: Re: Confirmation about thread un-safety/safety
List-Archive: http://lists.mysql.com/plusplus/8440
Message-Id: <49B028B6.9000703@etr-usa.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

Jim Wallace wrote:
> Thanks for the reply. Just to confirm, from reading the link, if I use
> store/store_next, I can give the StoreQueryResult to another thread
> since the first is finished consuming the results and the entire result
> set is in the StoreQueryResult.  (Looking at the code, that looks safe
> since I'm just calling vector::at())

That sounds right, particularly with v3.0+, where there are fewer hidden 
links between MySQL++ data structures and the low-level C data 
structures returned by the MySQL C API library.

That said, beware that I'm not writing from experience, since, as you 
know, I believe threads are evil.  I only use them for applications so 
trivial that they cannot fail.

(Not that you asked, but my company's primary product achieves 
concurrency through message-passing among independent single-threaded 
processes.  Each process uses async I/O to keep it responsive.  We don't 
need no steenking threads.)

You might want to download the svn version of MySQL++ and build the user 
manual.  I've just finished reworking the threads chapter and the 
tutorial section on concurrent queries to clarify several things.