Culture shock

[Whovian]

I've got quite a bit of C++ down (thought there is no way I know everything), and have some minor experience in Java and C#.

So, for anyone who knows another language in addition to C++, what would you say is the hardest bit for transition either way?

From Java to C++, I'd say Memory Management/Pointers, Templates, and Multiple Inheritance.

From C++ to Java, I'd say the lack of MI, lack of Enums, Memory Management, and everything being declared in a class. In fact, the one thing I can't wrap my head around is why Sun decided to force everything to be declared in a class.

[Cubbi]

The hardest transitions are cross-paradigm transitions: every little bit is different. Look at Haskell or Scheme, Prolog or Erlang, APL or J

the one thing I can't wrap my head around is why Sun decided to force everything to be declared in a class

That's the legacy of the original OOP schism betwen Simula's "objects model a physical world" and Smalltalk's "everything is an object". C++ and Java chose different object models.

[zb0S216C]

Whovian wrote:

"So, for anyone who knows another language in addition to C++, what would you say is the hardest bit for transition either way?"

I went from C to C++. Does that count? If so, definitely templates - have you seen the error messages? It's like staring at sun:

Error Message wrote:

"std::basic_istream<_CharT, _Traits>& std::basic_istream<_CharT, _Traits>::ignore(std::streamsize, typename _Traits::int_type) [with _CharT = char, _Traits = std::char_traits<char>]|"

...not informative at all.

Everything else is easy enough.

Whovian wrote:

"I can't wrap my head around is why Sun decided to force everything to be declared in a class."

I don't think it's anything to do with the fact that it's an object-oriented language.

_Wazzak

[hanst99]

Going from C++ to Java, i was at first like "yay garbage collector! no more explicit destruction!" - which of course only worked til I needed deterministic destruction anyways. Damn.

[chrisname]

hanst99 wrote:

til I needed deterministic destruction

Why did you need that?

[hanst99]

Destroying OpenGL textures before the context is released.

[rapidcoder]

Destroying OpenGL textures before the context is released.

It is not really much harder than in C++. Just need to provide a manual destroy/close method, similar to C++ destructors. And if you know a little of reflection and the new Java 7 try-with-resources block, you can even accomplish exactly the same behaviour as RAII in C++.

[Moschops]

If so, definitely templates - have you seen the error messages?

Try switching to the LLVM compiler with the Clang++ front. The error messages are an order of magnitude more helpful.

[hanst99]

It is not really much harder than in C++. Just need to provide a manual destroy/close method, similar to C++ destructors. And if you know a little of reflection and the new Java 7 try-with-resources block, you can even accomplish exactly the same behaviour as RAII in C++.

But that's exactly the problem - I need the order to be deterministic. The context must be released AFTER the textures have been. So you need explicit destruction. Which is what I said (I never said it was problematic to implement in Java, it was just that I came to Java with the expectation that the garbage collector would somehow magically do everything for me).

[MrHutch]

Going from C++ to Obj-C is a bit meh.

Syntax is funny, straying from dot-notation more often than not. Conceptually, it's a bit similar to Java in some ways, despite being a subset of C.

I thought method calling and definitions were particularly odd.

This call looks pretty funky at first glance. ;-)
[myObject Add: 5 AddTo: 10];

[zb0S216C]

Moschops wrote:

"Try switching to the LLVM compiler with the Clang++ front."

Thanks for the suggestion :) I'll look it up.

_Wazzak

[hanst99]

I have heard that Obj-C takes stuff from Smalltalk so... this is a block that sends the message Add:AddTo:(5,10) to myObject upon execution?

[MrHutch]

Yeah, pretty much.

The C++ equivalent would be something along the lines of:

myObject.Add(5,10);

The 'AddTo' part is optional. You can give parameters tags to remind you what they are when you're calling them.

It's weird not using dot-notation. Kind of easy to spot method calls, though, as they'll always be inside of square brackets. There's minor exceptions where Obj-C classes can form their own setter/getter methods and you'd use dot-notation, but that's another story. :-)

[hanst99]

It would look like

| myObject |
myObject := MyObjectClass new.
myObject add:5 to:10

in Smalltalk.

Square brackets are used for defining closures in Smalltalk, like

1 to: 100 do: [:i | Transcript show: 'Hello World! ',i,'th time!'].

(Squeak/Pharo Smalltalk)

[Whovian]

iHutch105 wrote:

despite being a subset of C

I didn't know that! Wow. (EDIT: I mean superset)

[hanst99]

Obj-C isn't a subset of C, but I think that was just a slight slip up.

[Whovian]

O rly? http://en.wikipedia.org/wiki/Objective-C#Syntax

[chrisname]

hanst99 is right - Objective-C is not a subset of C, it's a superset of C.

[Whovian]

Ah. Sorry, I also typoed that. =P I seem to be prone to typoes.

[moorecm]

In fact, the one thing I can't wrap my head around is why Sun decided to force everything to be declared in a class.

It's really no different since you can break encapsulation at will.

I went from C to C++.

I would guess that this is the hardest transition there is. It's particularly tricky because it's a mindset change more than a syntax change. It also leads a lot of developers to thinking that they have successfully moved to C++ when they haven't transitioned at all.


Going from something like Java or Python to C++ would be difficult because of the lack of standardized/built-in libraries to do everything under the sun.