Re: [Drizzle-discuss] Removing Array custom vector allocation for semi-join subqueries

[Michael Widenius]

Hi!

<cut>

>> For certain objects, memroot is a much better way to do it than with
>> C++ deallocators:
>> 
>> - Uses less memory overhead for objects
>> - Less mutex, as you mainly do one free per query and
>> - Much faster malloc;  Mallocs are done in bigger chunks and for many
>> queries you can avoid do any mallocs at all (when the base memory you
>> have is big enough)
>> 
>> memroot is of course not useful for objects that changes size or has a
>> short life time, but it wasn't designed for these.
>> 
>> 

Jim> Monty, aside from minor issues like breaking C++ language semantics, 
Jim> this is extremely dangerous from an engineering point of view.  The 
Jim> problem is this: If an object is properly deleted, its destructor is 
Jim> called.  If it is deleted with the pool, the destructor isn't called, 
Jim> but the memory deleted anyway.  Valgrind isn't going to notice a 
Jim> problem, but the failure to call the destructor may leave memory subtly 
Jim> corrupted.  To work, it requires that the code and/or engineers track 
Jim> which objects need to be explicitly deleted.  Unreliability comes from 
Jim> this sort of problem where an engineer has to understand the entire 
Jim> system to make a successful local change.

I assume you know that we have taken this into account for C++ objects
that are put into the memory root, like items. For these all
destructor's are called before the memory is deallocated.

Jim> If you're concerned about memory allocation/deallocation performance, 
Jim> use a thread based memory allocator that eliminates the mutex.  You get 
Jim> the performance you want leaving C++ semantics intact.  It will lead to 
Jim> cleaner encapsulation, faster code, and a lower induced bug rate.

You can't get it faster than memroot and not much cleaner either...
Bug rate is in the eyes of the coder; A imaginative coder can make
mistakes or break anything and for this the c++ destructors offers no
protection.

As far as I have seen in the past, using memroot give you faster code,
and lower bug rate than when using normal C++ code.

Regards,
Monty