Re: STL map or hash map using struct as data and find it
On Dec 31 2007, 4:20 pm, kl <kjell.ll...@gmail.com> wrote:
On 31 Dec, 15:52, James Kanze <james.ka...@gmail.com> wrote:
On Dec 31, 1:51 pm, kl <kjell.ll...@gmail.com> wrote:
On 31 Dec, 13:21, James Kanze <james.ka...@gmail.com> wrote:
On Dec 31, 11:48 am, kl <kjell.ll...@gmail.com> wrote:
[...]
I will improve the the structure or even add it into a class
(try to avoid that as it is very simple lookup table with one
purpose of function to translate an ip address to a country).
However the structure is working fine currently and has worked fine
for while
There's nothing necessarily wrong with using a struct here; as I
explained, given your performance considerations, and the fact
that the table is not updated once initialized, I'd seriously
consider a using a C style array of POD struct and memory
mapping it. With the precautions that I explained, to avoid
errors.
but of course improvments can be done and I'm open to support
cross platforms had some plans to port my app to Linux one day
and it is today a Windows app but I don't want use any
Microsoft propertiary classes such as MFC or other stuff hence
that's why I'm looking into STL.
Isn't DWORD a Microsoftism? I've never seen it used on a Unix
machine. (In fact, the only place I've seen it used in real
code, other than in the Windows API, is on IBM mainframes and on
the old Interdata 8/32. Where it was an 8 byte entity.)
More generally: unless the struct consists entirely of char and
char[] (possibly signed or unsigned), you'll have to consider
that if you write it as a binary memory image, you'll only be
able to read it as a binary memory image with the same program
which wrote it (or a program compiled with exactly the same
compiler, compiler version and options). From what I
understand, this is probably an acceptable restriction. Whether
it is worth the bother depends on the application; if you're
writing a server that is started once per day, or less
frequently, it's probably not worth the bother---reparsing the
CVS each start-up is not an issue. If its a small service
application, invoked each time it is used, start-up definitely
is an issue, and using the binary memory image is definitly the
way to go. Even if you don't use memory mapping, a single read
for the entire table will be a lot faster than any other
solution (and is very simple as well).
It is developed in C but consider to port it over to C++ as it
is a more modern and...
That explains the use of FILE*:-).
(Honestly, for this level of I/O, I'd define a very slim wrapper
around the basic system level functions, and use it. You'd have
to port it to Linux, but neither FILE* nor iostream are really
designed for effective binary I/O.)
--
James Kanze (GABI Software) email:james.kanze@gmail.com
Conseils en informatique orient=E9e objet/
Beratung in objektorientierter Datenverarbeitung
9 place S=E9mard, 78210 St.-Cyr-l'=C9cole, France, +33 (0)1 30 23 00 34