Thread: Class list. Fast IN&OUT

I have in my application on the client side several lists of data, which are compared, merged and so on finally giving the user final list of available values for inserting. All those lists are sequences that supports bidirectional iterators (class list).
What I have in code:

Code:

In header: typedef list<string, allocator<string> > muteList;
In body: 
muteList tkn_list;
muteList full_list;
muteList diff_list;
 
tkn_list = GetTakenList(ActLev);
full_list = FullMutationList(ActLev);
set_difference (full_list.begin(), full_list.end() ,tkn_list.begin(), tkn_list.end(),inserter(diff_list,diff_list.begin()));

tkn_list is one which is getting its values from DB in TclientDataSet. I’m feeding it with data using While EOF loop.
I believe this is not good and I know this is no fast. Could anyone help me by pointing to the more sound way. TclientDataSet has a Data property of OleVariant type. In my case it is filled with a plain list of strings. I wander if there is a way to copy its data to list sequence (or to get its data as a list sequence) in single operator. May be by casting OleVariant to pointer to list or something like that.
Actually I never used class list. It is not my usual level of programming. Additionally question is: how to send data from list to data controls without loop?
All suggestions will be more than appreciated.
Alex

Did you profile your code?

I don't actually understand what you are getting at. In code part I'm just giving several strings to illustrate what I'm talking about.

You said you believe one part of your code is not fast enough and is slowing the program down. Have you actually profiled to verify that is the case, or are you just guessing?

Programmers are typically very bad at guessing where bottlenecks are, and end up wasting a lot of time optimizing a part of code that doesn't matter anyways.

Always make sure with a profiler that what you believe is slow is actually slow, before diving in and start optimizing.

Well, I'm just guessing. But I know lists are working faster then datasets. Knowledge how to get data in&out lists quickly would show how fast are other ways. I spent considerable amount of time trying to resolve this problem myself but couldn't. Now I rely on the help of this forum.

Well, I'm just guessing. But I know lists are working faster then datasets. Knowledge how to get data in&out lists quickly would show how fast are other ways. I spent considerable amount of time trying to resolve this problem myself but couldn't. Now I rely on the help of this forum.

Yes, a binary search tree (common implementation of set) is probably faster than a list. For large lists, that you need to do a lot of insertions/searches in, very frequently.

The problem is, if your network access takes 99.9% of the time, and whatever you do with the list takes 0.1%, even if you make your list infinitely fast, your program will still only be 0.1% faster. Is it worth your time, risk adding bugs, increased complexity, and reduced maintainability, to get that 0.1% performance increase?

Why not profile your code and find out what REALLY is taking up all the time? Profiling should always be the first step in optimization.

Thank you all for your participation. I cannot not agree with you. But it is a pity that I rest with holes in my education.

It's better to not learn a bad habit than learn a bad habit.

Premature optimization is a very common beginner mistake, that has wasted people countless hours, for exactly nothing, besides bad code.