Re: [PATCH] Corrections to Documentation/rbtree.txt

[Ian Abbott <abbotti@mev.co.uk>]

On 20/03/08 18:39, Rob Landley wrote:
> On Thursday 20 March 2008 10:29:57 Ian Abbott wrote:
>> From: Ian Abbott <abbotti@mev.co.uk>
>>
>> The description of the rb_entry() macro in Documentation/rbtree.txt seems
>> incorrect. This patch improves it (hopefully).  Also I changed the example
>> code to call the previous 'my_search()' example instead of an undefined
>> 'mysearch()'.
> 
> I have no objection to the patch (and the my_search thing seems like an 
> obvious typo), but is there a reason to prefer rb_entry() rather than 
> container_of()?  If so, the rationale might be a good thing to add to the 
> documentation...

I don't know the rationale, but all the code I can see uses rb_entry() 
and not container_of().  The only rationale I can think of is that it 
abstracts away from the nodes being embedded in the data a little bit. 
(But not by much - in particular, an implementation of rb trees that 
stored data in the node explicitly would only need a single parameter in 
its rb_entry() accessor.  I like the approach taken in 
include/linux/elevator.h that uses the rb_entry() macro to create a 
specialized accessor macro (rb_entry_rq()) with a single parameter.

-- 
-=( Ian Abbott @ MEV Ltd.    E-mail: <abbotti@mev.co.uk>        )=-
-=( Tel: +44 (0)161 477 1898   FAX: +44 (0)161 718 3587         )=-