Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Reelya]

That looks like C#, which is strictly object-oriented. You can't have code outside classes.

Someone got to it before me, but it's def Java.

[MagmaMcFry]

The idea behind having all code in classes is way more than just simplification. It also adds flexibility in every aspect other than writing code.

Don't you think that statement is contradictory?
My experience is that in many instances using object orientation makes you add tons of boilerplate that could be easily avoided when using some other paradigm which offers everything you need for less code and better readability.

Flexibility comes from having more tools available, not from having one tool which supposedly does everything.

I said "every aspect other than writing code". Compilation, distribution, dynamically loading classes at runtime, these things. And technically, the only negative aspect of forcing everything to be in classes is that you have to write "public static" in front of all your otherwise-global methods and variables.

[Antsan]

I don't get how having everything object oriented would make these things more flexible. I'm not even sure how compilation can be "more flexible" (unless you mean programmable compilers, in which case object orientation doesn't help in the slightest).
Common Lisp has dynamic loading of everything, including functions and class definitions. It even can do redefinition of everything at runtime.

And technically, the only negative aspect of forcing everything to be in classes is that you have to write "public static" in front of all your otherwise-global methods and variables.

No, it also forces you to think in a very specific way about programming. The object metaphor can be pretty imposing and that is a bad thing when you're thinking about something which isn't neatly representable in a class hierarchy.

It's not like the programming world is divided into "functions" and "objects".

[Orange Wizard]

Personally I quite like having everything in classes, but that's mostly because it appeals to my deranged OCD.

It doesn't make the code any more or less flexible. It makes you think in a specific way about namespaces, which changes how the code is structured on the macro-level but don't really do anything beyond that.

Also, it makes things neater.

[Antsan]

It makes you think in a specific way about namespaces, which changes how the code is structured on the macro-level but don't really do anything beyond that.

Seeing how there's stuff beyond classes and functions in other languages, I'd say you're wrong. Of course most language don't make use of anything but data structures and functions.

If you like neat code, try Haskell some time. You'll look at OOP and wonder how you could ever think it was neat.

[LoSboccacc]

more a case of having this or that particular hammer.

look a piece of code has normally 4 ports:

- a data on stack coming from above
- b data you put on stack for code below you (includes returns)
- c data that sits on the heap
- d data that sits remotely (disk, db, stuff)

a programming style is mostly about managing the complexity of this.

oop has a thing for partitioning the heap, so that a method only has to deal with minimal c, you know everything that touches a particular subset of c, and thus you mostly care about a and b - possibly you limit complexity with only taking two port at once (like, a dao class would work with a and d or b and d, but not c) but that's just good habit and optional to the style.

functional programming is mostly focusing on having everything flowing from a to b with minimal impact from other ports to minimize complexity of processing data, and treats c and d mostly as start/end of a particular code flow

the language you code in doesn't really prevent you to use this or that style. heck I've seen beautiful oop design in c.

the more you think about code and data flow, the less you need to take on a particular style as a religion, as opposed to use the one that solves the problem at hand more easily (many complex data structure in memory? many transformation to be done on a large dataset? need to handle multiple synchronized updates on many datasources?)

[Orange Wizard]

If you like neat code, try Haskell some time. You'll look at OOP and wonder how you could ever think it was neat.

Could you say it's...

* Orange Wizard puts on sunglasses

Objectively better?

More seriously, OOP vs. not-OOP boils down to personal preference. S'not really a big deal.

[Moghjubar]

Not quite, especially in cases where you need optimization, you have to break OOP practices (while few people these days gain benefit and work with low level... and most programs wouldn't get written if they all had to be that low... the savings from a properly optimized program are tremendous)

[jaked122]

Besides, you can do non-oop in java. It just results in static classes, methods, attributes, etc. Make them public and they are very nearly the same as declaring a global variable or function. It's just less convenient to call them. More text.

[i2amroy]

Not quite, especially in cases where you need optimization, you have to break OOP practices (while few people these days gain benefit and work with low level... and most programs wouldn't get written if they all had to be that low... the savings from a properly optimized program are tremendous)

Except so much work has been done on the C/etc. compilers that in the vast majority of cases the compiler will be able to optimize even unoptimized code beyond anything you could do. I don't have the link here, but I remember at least one article I've read before by a compiler guru who pointed out that in many cases when you are working in C or C++ we've reached the point where your manual optimizations in lots of cases end up actually slowing down the code because the compiler has to bend over backwards to jump through the "optimization" hoops that you set up rather than just using the best optimized solution.

It's not always true, of course, but compilers are starting to get smart enough to be able to optimize your logic much better than you ever could manually these days, and that's only becoming more true as time passes and people put ever more amounts of work to do that.

Funnily enough, in a fair number of cases this also applies to an advantage of higher level languages over lower level ones, since it lets the compiler do more optimizations that have been laboriously compiled over the years, rather than forcing you to manually program in what you (hope) is the best optimization of them all. (Of course this is countered somewhat by the fact that lower level languages have been around longer and thus in general tend to have better compilers, but it's still something to keep in mind).

[Gatleos]

Man SFGUI is really sparse on documentation. I've been trying to connect a callback function to an "OnExpose" event, which is supposed to trigger every time a GUI element is rendered... but it doesn't. I've already seen the authors of the library telling a confused user that some things in the library aren't fully implemented, as if that makes it okay for the library to not TELL you this. Now I'm wondering if that's the case here.

GUI programming is hard.

[Telgin]

Not quite, especially in cases where you need optimization, you have to break OOP practices (while few people these days gain benefit and work with low level... and most programs wouldn't get written if they all had to be that low... the savings from a properly optimized program are tremendous)

Except so much work has been done on the C/etc. compilers that in the vast majority of cases the compiler will be able to optimize even unoptimized code beyond anything you could do. I don't have the link here, but I remember at least one article I've read before by a compiler guru who pointed out that in many cases when you are working in C or C++ we've reached the point where your manual optimizations in lots of cases end up actually slowing down the code because the compiler has to bend over backwards to jump through the "optimization" hoops that you set up rather than just using the best optimized solution.

It's not always true, of course, but compilers are starting to get smart enough to be able to optimize your logic much better than you ever could manually these days, and that's only becoming more true as time passes and people put ever more amounts of work to do that.

This is all true and something that I live by.  It does remind me of an interesting anecdote from recent experience though.  I was surprised to discover that GCC 4.8 at least isn't able to optimize a simple for loop that checks the .size() condition of a vector or similar standard library class so that the .size() check is moved out of the loop condition and replaced with a simple cached copy of the value.  That is,

Code: [Select]

vector<int> something;

// Insert a lot of things into something

for (unsigned int i = 0; i < something.size(); i++) {
    // Do things
}

should be translated to something like:

Code: [Select]

vector<int> something;

// Insert a lot of things into something

unsigned int size = something.size();

for (unsigned int i = 0; i < size; i++) {
    // Do things
}

The reason, of course, is that the compiler doesn't know easily if the loop body changes something that would cause .size() to return a different value.  Even with relatively simple code that only reads from the vector though, it won't treat the size as an invariant.  I've actually saved several % runtime in a computationally heavy program by performing that optimization manually.

[Reelya]

Has anyone used the Box2D library in C++. I need to get this hooked up in an SFML demo by tomorrow, and nothings actually moving so I don't know where to start. I made a body, added the stuff listed in the tutes, apply a force and step forward in the world each frame, but there's no visible effect on the object I'm trying to move.

[monkey]

Is the object's density non-zero ?

[jaked122]

Has anyone used the Box2D library in C++. I need to get this hooked up in an SFML demo by tomorrow, and nothings actually moving so I don't know where to start. I made a body, added the stuff listed in the tutes, apply a force and step forward in the world each frame, but there's no visible effect on the object I'm trying to move.

This sounds like a silly issue.


After each step do you update the coordinates of the sfml objects that are represented in the box2d objects, those don't seem to be coupled normally, so perhaps you ought to update the positions of your drawing primitives to the positions of your physics primitives.


Sorry that I can't be more specific, but that sounds like the sort of problem I would have. Is that it?