Disclaimer: I'm trying to learn proper OO programming/design, so I'm pretty new to this stuff.
I guess this is a general design patterns question, but I'll base my example on a game engine or something that renders objects to the display.
Consider the following:
hierarchy http://img31.imageshack.us/img31/9633/diagrame.png
How can this sort of separation between physical objects (e.g., cubes, spheres, etc.) and the rendering mechanism be achieved in an extensible manner?
This design is not set in stone, and perhaps I've got something wrong from the start. I'm just curious as to how a problem like this is solved in real world code.
That would be the Adapter pattern, or it could be implemented as a Strategy pattern.
The renderer should not be extended by the objects which he is supposed to draw. (Just my opinion) an object in your world is NOT a renderer but the renderer uses objects.
So you have maybe:
Interface IRenderer which defines a function draw(BasicObject).
Then your objects just extend BasicObject to be handled by the/a renderer.
As I said just my opinion. :)
Strategy patern it is.
I would use a Visitor pattern here.
Where the Visitor is the renderer and were the Visited is the 3D/Object.
I would also make the 3D/Object a composite.
Related
My task deals with multi-language like (english and hindi). For that I need a common embedding to represent both languages.
I know there are methods for learning multilingual embedding like 'MUSE', but this represents those two embeddings in a common vector space, obviously they are similar, but not the same.
So I wanted to know if there is any method or approach that can learn to represent both embedding in form of a single embedding that represents the both the language.
Any lead is strongly appreciated!!!
I think a good lead would be to look at past work that has been done in the field. A good overview to start with is Sebastian Ruder's talk, which gives you a multitude of approaches, depending on the level of information you have about your source/target language. This is basically what MUSE is doing, and I'm relatively sure that it is considered state-of-the-art.
The basic idea in most approaches is to map embedding spaces such that you minimize some (usually Euclidean) distance between the both (see p. 16 of the link). This obviously works best if you have a known dictionary and can precisely map the different translations, and works even better if the two languages have similar linguistic properties (not so sure about Hindi and English, to be honest).
Another recent approach is the one by Multilingual-BERT (mBERT), or similarly, XLM-RoBERTa, but those learn embeddings based on a shared vocabulary. This might again be less desirable if you have morphologically dissimilar languages, and also has the drawback that they incorporate a bunch of other, unrelated, languages.
Otherwise, I'm unclear on what exactly you are expecting from a "common embedding", but happy to extend the answer once clarified.
Could anyone give me a good use case for ActionScript dynamic classes?
Because it really looks like a bad pratice, for me, in every case.
Anything that uses Proxy must by extension be dynamic. I use Proxy fairly regularly; for example, here's a replacement syntax for ExternalInterface using Proxy.
URLVariables, for example.
You could store the data in a dictionary / object / array too, but you don't gain much in this case, I think, and you cut down some boilerplate.
Well you could use an Object object, but using a dynamic class ensures that it is typed. That's the way I see it, and it's the only reason I would use them. What Juan Pablo is saying is a good reason too.
I have problems with how to design some classes. I have three classes. One superclass, and two subclasses.
One subclass (AnimatedCharacter) is made by flash, and is used to display the object on screen. The other (CharacterPhysics) is made by myself to extend the superclass.
The problem is that the object I use, is of the type AnimatedCharacter, so I can't just put it in a variable of type CharacterPhysics.
What I tried is some sort of Decorator pattern, by giving the object of type CharacterPhysics a reference to the other object. But now I have to override all the methods of the superclass and pass the methodcalls to the reference. Not an ideal situation.
Does someone know how to solve this kind of problem?
alt text http://www.freeimagehosting.net/uploads/7a95f8352c.png
I don't quite understand the purpose of this class structure you describe (the class names confuse me), but in general a few things come to mind that might help you:
Almost always the best solution is to try and rethink your class model by evaluating whether you should for example break up the responsibilities of classes in an alternate way so that you could utilize inheritance and polymorphism in a better way.
"The problem is that the object I use,
is of the type AnimatedCharacter, so I
can't just put it in a variable of
type CharacterPhysics."
If you want to put an AnimatedCharacter into a variable of type CharacterPhysics, the former should extend the latter, or you should have a common interface (or superclass) for both and then type the variable as such. If this is not possible, my opinion is that you should probably try to rethink and refactor your whole class structure, assuming that you have a solid "object-oriented" reason for wanting to do this in the first place ;).
If the above is not possible, there are some other tricks you can evaluate in your context:
The use of mixins can work as a "poor man's multiple inheritance". Derek Wischusen has some examples on how to implement them in AS3 at flexonrails.net.
"Kind of" implementing the decorator pattern with flash.utils.Proxy. The problem with this approach is that you defer a lot of error checking from compile time to runtime, but the good thing is that you don't have to manually write the "proxying" implementations of all of the methods of the "decorated" object, but write just one (callProperty()) instead.
You can interpret a sublass as an instance of a superclass but not vice sersa. Did you state this backwards?
If so, you could use:
vas cp:CharacterPhysics;
...
var ac:AnimatedCharacter = cp As AnimatedCharacter
Off the top of my head, it seems like those 2 should be interfaces which your main class implements
I sometimes end up with a class hierarchy where I have an abstract base class with some common functionality and a couple of implementing classes that fall into two (rarely more) groups which I want to treat differently in some cases. An example would be an abstract tree node class and different branch and leaf implementations where I want to distinguish branches and leaves at some point.
These intermediate classes are then only used for "is-a" statements in flow control and they don't contain any code, although I have had cases where they "grew" some code later.
Does that seem smelly to you? In my tree example, one alternative would be to add isLeaf() / isBranch() abstract methods to the base class and implement those on the intermediate classes, but that didn't seem to be any better to me, really, unless I'd mean to have classes that could be multiple things at once.
To me, using "is-a" tests in flow control is just as smelly as using switch/case. In a good OO design, neither is needed.
Yes, deep inheritance hierarchies are a code smell anyway.
Yup, definitely a code smell -- don't code these empty classes unless you're ready to write that code into it. Think YAGNI (you aint gonna need it) -- don't do it unless you need it already.
Also, have you considered cases wherein these classes are only there to provide abstract methods, or to group them based on capabilities in terms of methods or properties?
If that's the case, maybe what you really need are interfaces, not additional abstract classes?
In general, empty classes are a code smell.
I agree your isLeaf or isBranch methods are a correct alternative.
They add information about the objects , which is helpful.
(This is because, on the super class, you can't express that subclasses are "either leaf or branch").
The two methods with opposite results might also be considered as code duplication.
You could use only one... But I would recommend return an enumerated value LEAF or BRANCH.
A class that doesn't contain any code is definitely a code-smell....
Seems alright to me, if you're going to put new functionality in later.
But if not, an enum is generally used here.
-- Edit
Though I must agree with Ber, that you shouldn't generally be using 'is-a' anyway.
In a web application, is it acceptable to use HTML in your code (non-scripted languages, Java, .NET)?
There are two major sub questions:
Should you use code to print HTML, or otherwise directly create HTML that is displayed?
Should you mix code within your HTML pages?
Generally, it's better to keep presentation (HTML) separate from logic ("back-end" code). Your code is decoupled and easier to maintain this way.
As long as your HTML-writing code is separate from your application logic, and the HTML is guaranteed to be well-formed somehow, you should be okay.
The only code that should be mixed in markup-based pages (i.e, those that contain literal HTML) is the code used for formatting the HTML (e.g., a loop for writing out a list).
There are trade-offs whether you put the code in with the HTML or you use pure code to write the HTML out using quoted string literals.
No, if you want to build good and maintainable software, and to achieve loose coupling.
If I understand the question right, you're asking whether it's a good practice to mix markup with back-end code. No. While this is commonly done, it's still a bad idea.
You should read up on the MVC paradigm, as well as on existing questions on the matter, such as What is the best way to migrate an existing messy webapp to elegant MVC? and Best practices for refactoring classic ASP?
The point is to keep the display logic separate from the rest of the code. In any complex site you'll have code mixed in with your HTML, but the code should be for display purposes only. It shouldn't be doing any complex calculations.
For example, templates will contain loops and conditionals. Plus you'll probably have a library of HTML-specific routines, like printing out an <option> list based on a list object.
Imagine you were writing an application that has two output modes: HTML and something else. How would you write it, to avoid duplicating code? That will probably point you in the right direction.
The HTML that makes up the view has to get sent to the browser in some way. In .net, each server control emits its own HTML markup as part of the page lifecycle. So yes it is OK to use HTML in server side code.
Perhaps you should try following the ASP.net pattern. Create a bunch of controls that represent UI elements and make them responsible for emitting their own HTML based on their state.
Its fugly, and not type safe. But people do it without consequence. I'd prefer using a DOM or, at a minimum, classes designed to write HTML using type safe semantics. Also, its not all that good to mix UI with logic...
If I need methods that generate HTML I usually isolate them in an HtmlHelpers class. That way you keep some level of separation. The ASP.NET MVC Framework does this quite successfully.
If you mean printing out HTML in your code, then no. Unless you have a good reason not to, you should use templates
Even if you think you don't need this now, there's always a good chance you'll need it later. Maybe you want to output in a different format than HTML, or you want different presentation for the same data. You usually have the need for these things further down the road, so it's best to use one from the start.
I hate when developers print() a bunch of html. It's completely unnecessary and looks ugly in any text editor that shows print/echo strings in red.
I agree with everyone else that you should try as hard as you can to separate the HTML/XHTML markup from the application logic. However, sometimes you do need to generate HTML/XHTML in the application logic for various reasons.
In these cases what I have been trying to do is to ensure the bare minimum amount of presentation code is in mixed in with the application logic and try to migrate everything else over to the presentation code. It is worth nothing that is some cases you have situations where you could have everything moved over to the presentation layer, but it might be a bit easier to generate the markup as part of the application logic. In those cases, your best bet is likely to be to go the route that makes the most sense in terms of time.
I don't think there's any excuse for generating HTML inside your business logic. Don't even do it when it's just a "quick fix" or when you'll "go back and fix it later", because that never happens.
To reiterate my position from other questions, using some control logic (conditionals, loops) within HTML to construct it is OK. Do NOT do any data massaging or business logic in the HTML. You have to be disciplined, but it's worth it. Maintenance is much easier if your concerns (like logic and display) are separated.
Ideally you are aiming for a separation of concerns between your presentation (UI) code and your domain (business logic) code.
The reason why you should avoid coupling these two concerns (in either direction) is simple...
You will only have one reason to change a piece of code. whether this is from structural/styling changes in your html design, or from your business rules changing, you should only have to make the change in one place.
To a lesser extent, although many purists would disagree, by sprinkling HTML code through your domain code or vice versa you are creating noise for the next developer who comes along to read/maintain it.
I try to avoid using code to print HTML "directly". It is difficult to maintain, edit, add styles and etc. Some cases like generating an HTML email in the code, I create a text file or HTML file with markers like, [name], [verification code] and etc. I load this from the code and replace those markers. This way, you can edit the style of the email without re-compiling your code. Separating "presentation" and "logic" is a good practice in my opinion.
Mixing code within HTML is generally not a good practice in similar reasons as said in #1. However, I do use code in HTML for things like simple dynamic strings that are displayed multiple times on a page or pages. I think this is better than creating multiple server controls for same exact values to set. Since this is not code "logic" mixed in the HTML, I think this is ok.