开发者

advice on an oo planing

开发者 https://www.devze.com 2023-02-21 00:33 出处:网络
I am trying to develop an object oriented representation of a finite element model. The finite element model consists of some output files(outputs of a commercial Finite element program) in binary for

I am trying to develop an object oriented representation of a finite element model. The finite element model consists of some output files(outputs of a commercial Finite element program) in binary format which keep certain data related to the finite element model (for people who are not in the field, these data correspond to nodes of the model( coordinates of the nodes), elements of the model, element connectivity for the element representations(which elements are associated with which nodes) and element matrix information). These data could also be represented generally as some vectors of ints and double and some other array structure if you call so.

I was thinking of making a general class FeModel and associating the above mentioned files as members of this class and work this way, of course these files would also be objects representing these files. However, the files that keep the information are not a part of the FeModel class that I was thinking on, at least conceptually. Since the idea does not relate to the real world representation, I thought that there should be a better way. Those files are just there to keep the information.

I am now thinking on the option of just reading the necessary information from these files in the FeModel class by creating suitable memb开发者_如何学JAVAer functions and build what I want this way so that the interface is more or less a minimal one. However, on the other hand dividing this task into different classes representing the above mentioned files and using those as members in the FeModel class does not look also like a bad option to me. What are the decision criteria in these situations? I am aware of the fact that a problem could be solved in many different ways but are there some sort of guidelines to follow in similar kinds of cases where one hesitates between some options?

Greetz, U.


My that's a bit of a wall of text. It could use a bit of filtering/rewriting.

As I understand it you have several files representing your 3D elements etc.

I nice OO way of doing it (at least in my mind) would be to have a separate class for each type of file you want to load, pass the filename in the constructor and load in the data, and have accessible members to access the data.

But without further information, structure to what you're asking it's a bit difficult to say.


You probably want to read about SOLID principles of OOP. In my opinion, it's all about making you design less fragile in the face of changes. In few words, you should strive to the design which is changes-friendly: minor changes domain in model should be reflected by minor and local changes in the design and the implementation.

Don't forget that C++ is not OO-language, but rather language that supports several paradigms (including OO) and other paradigms could fit better to task at hand. If you have access to Bjarne Stroustrup excellent "The C++ Programming Language (Special Edition)", read through Part IV ("Design using C++") to get the idea of roles of classes and other C++ concepts.


An OO solution focuses on behavior, not data. You haven't told us what behaviors you need so we can't help with an OO solution.

That said, find your invariants and create a class for each related set of invariants and you should end up with a decent modular solution.

I suggest you read the book Object-Oriented Design Heuristics by Arthur Riel.

0

精彩评论

暂无评论...
验证码 换一张
取 消