From: Martin Hansson
Date: December 16 2010 1:59pm
Subject: Re: bzr commit into mysql-trunk-bugfixing branch (martin.hansson:3260)
 Bug#58918
List-Archive: http://lists.mysql.com/commits/127080
Message-Id: <4D0A1B2B.2090208@oracle.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 8bit

Øystein Grøvlen skrev 2010-12-16 14.07:
> On 12/15/10 01:38 PM, Martin Hansson wrote:
>> Tor Didriksen skrev 2010-12-15 12.42:
>>> On Tue, Dec 14, 2010 at 1:50 PM, Martin Hansson
>>> <martin.hansson@stripped>wrote:
>>>
>>>> #At file:///data0/martin/bzrroot/refactoring-Field_clone/n-mr-o-t/
>>>> based on
>>>> revid:tor.didriksen@stripped
>>>>
>>>> 3260 Martin Hansson 2010-12-14
>>>> Bug#58918: Add a clone member function to the Field class for better
>>>> duplication
>>>>
>>>> The patch removes all copying of Field objects using memcpy() and
>>>> replaces it
>>>> with virtual clone() and clone(MEM_ROOT*) member functions that rely
>>>> on the
>>>> copy constructor. An operator new (MEM_ROOT*) is also implemented. The
>>>> sole
>>>> purpose of the Field::size_of() member function was to support the
>>>> memcpy and
>>>> it is now removed.
>>>> @ sql/field.cc
>>>> Removal of memcpy() based duplication.
>>>> @ sql/field.h
>>>> New implementation of clone() member functions.
>>>> Removal of size_of() member functions
>>>> @ sql/item.cc
>>>> Removal of memcpy() based duplication.
>>>> @@ -84,8 +79,12 @@ class Field
>>>> Field(const Item&); /* Prevent use of these */
>>>> void operator=(Field&);
>>>> public:
>>>> + /* To do: inherit Sql_alloc and get these for free */
>>>>
>>>>
>>> Why not go all the way :-)
>>> This works like a charm:
>>>
>>> class Field : public Sql_alloc
>>> {
>>> Field(const Item&); /* Prevent use of these */
>>> void operator=(Field&);
>>> public:
>>> uchar *ptr; // Position to field in record
>>> ...
>>> ...
>> I am planning to. In case you didn't notice there's a comment about it.
>> The reason I don't want to do it is because it would work by
>> coincidence, not by design. The class Sql_alloc is -very wrongfully -
>> defined in sql_list.h. We slip that include file in sideways by
>> including mysqld.h which in turn includes sql_list.h. I plan as my next
>> refactoring to create a file sql_alloc.h and then make every class that
>> implements an identical 'new' operator (quite a few of those actually)
>> inherit Sql_alloc instead. Only then will I have gone all the way :-)
>>
>> If you insist, of course I can do it!
>>
>
> Public inheritance?  I think it is questionable whether this is an 
> is-a relationship.  (Ref. "the single most important rule in 
> object-oriented programming with C++" according to Scott Meyers)
>
Nope, it isn't an is-a. But that's a limitation of the C++ language. You 
cannot overload the 'new' operator in a usable fashion without 
inheriting it publicly. But you should really talk to Mats Kindahl about 
these things, he knows much more than I do.

Regards

Martin