Re: [Dev-C++] Optimizing using the STL

Thanks for pointing that out..

There is no particular reason for keeping the algorithm.
I just like using less code, this method seems to use more.
(You have the original function and a functor now).

I have another question considering files but I think I
better start another thread for that.

Signed
David Mcken

--- Hossein Haeri <ult...@em...> wrote:
> Hi David,
> 
> > Some interesting ideas there. ...
> 
> Some? Just some? Oh my Goodness! Thanks for that! I could
> except any
> description expect that one.
> 
> > ... You are right I am not too
> > familar with the
> > STL. 
> 
> You will become! For now on, just don't panic when you
> see powerful
> STL codes.
> 
> > ... What is the difference
> > between passing a
> > refrence and passing a const refrence? Is that to
> ensure
> > that the original
> > can never be changed(as would happen of I passed by
> value)?
> 
> Hmmm... Yep in the sense that the original beast wouldn't
> be changed,
> but nop in the sense that there is big differnce between
> passing by
> value and passing by reference: The former copies the
> argument at the
> moment of being called, whilst the later doesn't copy,
> it, in turn,
> works on the same object, by creating a reference out of
> the
> argument.
> 
> > My other comment is that current_players is a member of
> > pc_char. It is
> > static and thus shared by all instances of pc_char. So
> I am
> > guessing the
> > function signature stays the same.
> 
> Well, if you mean the same as what it now is, yes, it
> does. But the
> important point here is that this is the first time I see
> a class
> having a map (in fact, a container) of pointers to
> itself. My
> recommendation is to change this strange design, unless
> you have
> very plausible reasons for adhering to it. It is very
> likely to
> result in sophisticated problems that are all hard to
> debug. (Amongst
> them, one of the most important and vexing ones is the
> "Circular
> Dependancies" problem.)
> 
> 
> > From the looks of it using for_each may not be worth it
> > especially since the
> > function is not finished. 
> 
> I read the bellow changes you want to be made. From my
> viewpoint,
> that doesn't difere much the dicision wheather to use or
> not to use
> for_each. Assuming that you are going to do the changes,
> you still can adhere to for_each. Especially because,
> that way, the only
> change is needed to be done is inside that Functor, and
> no where
> else. (Read on ...)
> 
> > ...
> > void pc_char::write_to_all_pc(const string& to_send,
> bool
> > add_eoln)
> > {
> >   for(map<string, pc_char *>::iterator i =
> > current_players.begin();
> >     i != current_players.end(); ++i)
> >     if((*i).second != this)
> >        (*i).second->write_to_player(to_send, add_eoln);
> > }
> 
> At the moment beeing, the only very subtle bug I can see
> in it is the
> one in the following line:
> 
> >     if((*i).second != this)
> 
> Just think about it for a moment. You are storing
> pc_char* objects
> in that map, right? But you are, probabely, just working
> with a copy
> of the addresses of the original objects, huh? So those
> nifty-looking
> pointers, which are beeing stored in the map, and are
> objects by
> themselves, are copies of the "this" objects, and not the
> same ones!
> Hence, althouge they point to the same location in the
> memory, they
> are not the same, by themselves. Therefore, it is very
> likely that
> the test fails, even if the semantic is right. In fact,
> you should be
> very lucky if the code works. The solution I suggest (if
> it is OK
> under your algorithm), is to replace it with the
> following one.
> 
> if(*(i->second) != (*this))
> 
> This way, you are checking if they are pointing to the
> same location,
> and not if they are the same pointers! (And that's
> presumably what
> you need!)
> 
> > With respect to the books. I will look into getting my
> > hands on a copy.
> 
> Do it, ASAP.
> 
> Regards,
> --Hossein

=====
Signed 
David Mcken

Life Sucks
Live with it

__________________________________
Do you Yahoo!?
Yahoo! SiteBuilder - Free, easy-to-use web site design software
http://sitebuilder.yahoo.com