Question on speed of set operations

Question on speed of set operations

[reinert]

Dear all,

I try to make an "Include" procedure similar to

procedure Include(Container: in out Set; New_Item: in Element_Type);

(under Containers.Ordered_sets) and where one includes 
(eventually replaces) elements in a set instead of including 
one element at a time. The intention is to make the user code 
clean/readable and efficient.

So what is the best alternative among the following tree procedures
for the above purpose (in terms of *speed*):

procedure Include(Container: in out Set; New_Items: in Set) is
begin
   for e of New_Items loop
       Container.Include(e);
   end loop;
end include_set;

procedure Include(Container: in out Set; New_Items: in Set) is
begin
   Container.Difference(New_Items);
   Container.Union(New_Items);
end include_set;

procedure Include(Container: in out Set; New_Items: in Set) is
begin
   Container := New_Items or Container
end include_set; 

Could there be other/better alternatives? And/or am I trying
to reinvent the wheel?

reinert

Re: Question on speed of set operations

[Simon Wright]

reinert <reinkor@gmail.com> writes:

> So what is the best alternative among the following tree procedures
> for the above purpose (in terms of *speed*):

If I were you I would do some benchmarking.

Re: Question on speed of set operations

[Randy Brukardt]

"Simon Wright" <simon@pushface.org> wrote in message 
news:lyh8bl2s3u.fsf@pushface.org...
> reinert <reinkor@gmail.com> writes:
>
>> So what is the best alternative among the following tree procedures
>> for the above purpose (in terms of *speed*):
>
> If I were you I would do some benchmarking.

Agreed. I doubt that there is a portable answer; it would depend on the 
internal implementation of the container type (which is purposely left to 
the implementation).

                                         Randy.

Re: Question on speed of set operations

[reinert]

> >
> >> So what is the best alternative among the following tree procedures
> >> for the above purpose (in terms of *speed*):
> >
> > If I were you I would do some benchmarking.
> 
> Agreed. I doubt that there is a portable answer; it would depend on the 
> internal implementation of the container type (which is purposely left to 
> the implementation).
> 
>                                          Randy.

And depending on the size of "New_Items" (as compared to the "Container"), I assume? 

Seems like I end up with the first alternative (since, for me, "New_Items" is small. 

reinert