RE: [Dev-C++] 'extern const' and std::vector

[Ioannis V. <iv...@at...>]

> -----Original Message-----
> From: dev...@li...=20
> [mailto:dev...@li...] On Behalf=20
> Of David McKen
> Sent: Thursday, May 01, 2003 10:09 PM
> To: dev...@li...
> Subject: Re: [Dev-C++] 'extern const' and std::vector
>=20
>=20
> The rooms are not nesesarily accessed in sequential order.
> One very used
> command (at least by admins) is the goto command. Syntax:
> goto <vnum or
> player>. This can cause an admin in room #300 to need
> access to room #50000.
> Also ceartain other command allow ordinary players to make
> different jumps
> to rooms that are not in sequential order.
>=20
> The assiging of rooms to the admins in charge of building
> makes it very
> likely that the rooms adjacent to the current room are
> going to be within
> that range (+ or -100) therefore with some exceptions
> (which are not that
> rare) moving from one room to another involves moving up or
> down a few
> vnums.


As Per Westermark said a map seems to be more suitable than vector since =
you
are going to add things in the middle and you want random access. Or a
multimap I would say if every room is beside in more than one rooms.




>=20
> Hmm, I understand about the making the global variable a
> class and actually
> that was my other question. (I guarantee you if it dosen't
> have to e global
> I don't want it to be global, that only increases the
> number of validity
> checks I have to make, eww)
>=20
> Each client is going to be in their own thread. For the
> reading functions
> this should prove no problem (two threads reading the same
> variable doesn't
> cause any problems? Right?) The writing functions (only
> avalible to admins)
> are what will require locking down (Mutex, Critical
> Section, etc.).



Yes you need a locking mechanism (although I do not know much about =
threads
yet).


=20
> One of my original thoughts was to create a vector of
> pointers to each room
> structure. Thereby allowing random room access and also
> allowing the
> numbering system to remain intact.
>=20
> if the vector currently only has rooms 1-100 filled in and
> an admin request
> to goto room #150 to start creating a new area then the
> rooms in between 100
> and 150 will be NULL pointers that can be checked for,


Or in the case of a map only the existing rooms will exist and when
accessing a non-existent room, 0 is returned (if suitable for the type,
which is suitable for pointers).



> if
> an admin request a
> room the does not exist then a template room is created for
> the admin to
> edit to their liking. This has the obvious disadvantage in
> that if an admin
> request room #999999 and the vector only has the rooms up
> to #1000 there
> will be alot of NULL pointers wasting memory.=20



Yes, so map is perfect, it will not create 0-value pointers (NULL is =
better
for C, 0 is better for C++ regarding pointers).



>Given the way
> that these
> numbers are assigned this should not be a problem (with the
> exception of the
> top admins who can load and save any area) most admins are
> forced to remain
> within their assigned range.
>=20
> Any thoughts on this will be appreciated.


As Per said: use map (or multimap if it fits you better).




> BTW we have a project on sourceforge 'SinstradMUD'.
> The current codebase is in the cvs repository in the /src
> directory.
> The (extremly alpha, still deciding what library to use)
> code for the new
> codebase is in the Sinstrad2 directory


For what OS is it going to be?




Ioannis Vranos
=20
* Programming pages: http://www.noicys.freeurl.com
* Alternative URL 1: http://run.to/noicys
* Alternative URL 2: http://www.noicys.cjb.net