dolfin team mailing list archive

[Andy Ray Terrel <andy.terrel@xxxxxxxxx>]

On Thu, Mar 1, 2012 at 9:59 AM, Marco Morandini <morandini@xxxxxxxxxxxxxx>wrote:

> On 03/01/2012 04:51 PM, Andy Ray Terrel wrote:
>
>> On Thu, Mar 1, 2012 at 2:22 AM, Marco Morandini<morandini@aero.**
>> polimi.it <morandini@xxxxxxxxxxxxxx>>wrote:
>>
>>  There are some occurrences of
>>>
>>> container.size() == 0
>>>
>>> scattered throughout the code
>>> (and fewer of container.size() != 0) .
>>> Many of them are harmless, i.e. whenever container
>>> is a vector. Some of them, however,
>>> are for maps of sets (e.g. dolfin/adaptivity/****Extrapolation.cpp:149
>>> ),
>>>
>>> leading to completely unnecessary O(N) operations (racall that size() is
>>> O(N) for a map or a list).
>>>
>>>
>>>  While true for a list, it is not a problem for maps or sets:
>>
>> http://www.cplusplus.com/**reference/stl/map/size/<http://www.cplusplus.com/reference/stl/map/size/>
>> http://www.cplusplus.com/**reference/stl/set/size/<http://www.cplusplus.com/reference/stl/set/size/>
>>
>
> I stand corrected
>
>
>  While a silly implementation of a list will be O(n) most good
>> implementations keep its size around.
>>
>
> but please be aware that the "silly implementation" we are talking about
> is gcc's one:
>
> http://gcc.gnu.org/onlinedocs/**libstdc++/manual/containers.**
> html#sequences.list.size<http://gcc.gnu.org/onlinedocs/libstdc++/manual/containers.html#sequences.list.size>
>
> Marco
>

Yet another reason I hate gcc.  Anywho the foo.empty() is better anyway as
it does represent the idea.

-- Andy