Which Coding Framework for Mac OS X ?

DrStrangeLoop asks: "I am in the progress of getting into coding for Mac OS X, and I am wondering which GUI framework/language i should focus on. Apparantly, there are at least three options: the Cocoa Objective-C API [I don't want to learn Objective-c, but it seems that's how Apple wants me to code], the Cocoa Java API [gets compiled to PPC binaries, lots of APIs available [think Google], but absolutely no decent documentation to be found] and Swing Java classes [look 'n feel of Cocoa, but portable]. However, the most important feature for me is a clean and easy IPC with BSD layer processes. I figure sockets will work with all options, but what about the other mechanisms? Any suggestions?" Update: 10/13 22:08 GMT by C :For those curious about the Cocoa/Carbon debate, you can find an article that discusses this very issue, here. Thanks to the folks over at Freenode's #macdev for providing the link.

Comments

You also have Carbon

[akac]

Carbon is a good API to begin with. You can use C++ or C and access all the nice GUI of OS X.

Cocoa just makes things a lot easier.

I like to think of it this way (though technically its wrong, its a nice way to think of it): Win32 == Carbon; .NET == Cocoa

learning objective c

[Dougal]

To be honest, I'd say get over it and learn Objective C. For an experienced programmer it won't take you long to get to grasps with the basics of Objective C, and Cocoa is really easy to work with. Who knows, you might even grow to like it :) At first I thought Objective C was weird, but now I really like it.

Get a beginners book and work your way through it. I recommend Cocoa Programming for Mac OS X by Aaron Hillegass, published by Addison Wesley, ISBN 0201726831. This seems to be what most people learn from.

Cocoa all the way

[petrilli]

As someone who grew up around Macs, and then moved to NeXT, I'd have to say that if you know C, and understand OO principles (prefferably in the Smalltalk model, not C++), then you can pick up Objective-C in an afternoon. It's really just some extensions to C. Carbon is a great set of APIs if you've got 10 years of Toolbox experience on the Mac, but otherwise, it's much harder to learn the ins and outs.

Cocoa, because of it's past with NeXTstep, has a lot of emphasis on rapid-prototyping and dynamicism. Lots of delegation, lots of easy stuff to do to get an effective solution. As for Java, I'd have to follow Aaron Hilldegrass' advice... don't. Not on the Mac, not unless your goal is cross-platform. If you're writing for the Mac, write in Objective-C, C++, or even REALbasic, but not Java. There's simply no good reason.

Having written nearly identical programs many years ago for Mac, PC and NeXTstep, I'd say this. Mac and PC (Mac = traditional Toolbox on OS7.x) and Win32 are roughly equivelent to get the work done. Different, but one isn't necessarily easier. The NeXTstep app was implemented in 1/3 of the time, and that included me learning NeXTstep.

Nothing wrong with Objective C

[zulux]

I love C++, templates, and think nothing of derfrrencing a pointer to a vector of objects that are stored in a crazy containter that I cobbeled together last night while high on cough syrup and mountain dew:

However -

Objective C kicks ass for GUI apps. It's easy to learn, and hides some of c's rough edges, and it's fast to compile. Make no mistake - nobody's going to code the next operatig system in it - but for GUI apps, take this good tool and run with it.

Re:Cocoa all the way - Mod Parent Up

[Anonymous Coward]

The above poster hit the nail on the head.

Now, for my own two cents regarding Java+Cocoa: Don't do it. Some friends and I programmed a Jabber-based instant messenger in Java and Cocoa before OSX was officially released. Working with the beta frameworks was a nightmare, for one simple reason: Nobody else was doing anything with Java and Cocoa. To my knowledge, there are only a handful that work with it now.

Like other posters have said, if you're just starting out on a platform, get over your fear of learning a new language. Objective C + Cocoa on OSX and C# + Windows Forms on WXP are the way to go. Also, Carbon, like MFC on Windows, is only really useful if you've got a lot of experience with the older Mac operating systems. There's a lot of cruft in there that's not really necessary and can be a pain to learn. You should find yourself developing a lot cleaner code a lot faster if you take the time to learn the newer tools available.

Have fun. =)

Obligatory resume (on topic because it lists the project!... riiight): http://www.cs.hmc.edu/~awooster/resume.htm

Cocoa / Obj-C

[Sahib!]

Definitely, learn Cocoa and Objective C if you can, just for the experience of realizing how much better it is that most of your other choices. With InterfaceBuilder, you can completely abstract your UI from the core of your code.

Since you say you want to use the BSD layer, I suggest making a command-line version of this core code first (you can do this in C or C++), since this will be immediately more portable to other Unices. Once that is done, tying that code in with the UI you build in InterfaceBuilder is simple.

Also, there are some pretty interesting native IPC libraries in Mach. If you don't mind your code being tied to Mach, you should check that out as well.

Oh, did I mention that Apple's developer tools and documentation are free for download?

Java Cocoa

[BlueGecko]

the Cocoa Java API [gets compiled to PPC binaries, lots of APIs available [think Google], but absolutely no decent documentation to be found]

First, Java Cocoa is not native. Your application will still run on the Java VM. In my mind, it fills a very narrow purpose: it allows you to quickly turn a 100% Java application into an application that behaves exactly like a truly native OS X application and which can take advantage of the niceties that provides (for example, Services support, Sheets, Quartz, etc.). Secondly, I hear the complaint about the lack of Java API documentation all of the time, and I honestly don't understand it. Admitting that I've been working in NEXTSTEP and later OpenStep and then Cocoa for a very long time, Apple provides a detailed Java API reference that has class-for-class and method-for-method documentation exactly on par with the ObjC documentation and also provides a tutorial for Java developers that literally produces exactly the same application as the ObjC Cocoa tutorial. You need merely do both tutorials to see how exactly the two APIs line up. The only somewhat legitimate complaint is that Apple examples typically come as ObjC, but how hard is it, honestly, to figure out that [aSpellingChecker ignoreWord: @"bodacious" inSpellDocumentWithTag: myTag] becomes aSpellingChecker.ignoreWord("bodacious", myTag)? The rules for ObjC-to-Java conversion of APIs and code are very straight-forward:

1. Any code relating to autorelease pools may be deleted and ignored (at least in basic code; by the time you need to make use of manual garbage collection in Java Cocoa (multithreaded code in the GUI being a big one), bluntly, you'll already know enough that it won't bother you)
2. Java class names remain identical to their ObjC counterparts, and are in either com.apple.cocoa.foundation if they are part of the ObjC FoundationKit and part of com.apple.cocoa.application if they are part of the ObjC AppKit. (QuickTime is in its own API and really is a separate discussion, since "QuickTime for Java" has existed for several years now and is more a part of Carbon than Cocoa.)
3. Method names for Java always consist of the first part of the ObjC identifier and always take the same parameters (except that NSStrings become java.lang.String). For example, (void)NSSpellChecker>>ignoreWord:(NSString *)inSpellDocumentWithTag:(int) becomes void NSSpellChecker.ignoreWord(String, int).

All that said, I would strongly recommend you take the time to learn ObjC Cocoa. I use Java Cocoa when I am grafting a Cocoa UI onto Java code, but ObjC is still faster (though the gap is narrowing) and I still honestly believe the language is still a bit more productive due to things such as Categories, Smalltalk-style method names, and other things. If you know C and Java and are familiar with the concepts of reference-counted garbage collection, learning ObjC will take you the better part of a morning.

Comment removed

[account_deleted]

Comment removed based on user account deletion

Re:Mozilla

[Fished]

People responding seem to be missing a crucial point: I am not talking about a web application here. I am talking about using the mozilla framework to develop a local application. This is actually how Mozilla, including Mail, Messenger and Composer, is developed.

If you haven't looked at it closely, don't knock it. Take a look at Creating Applications with Mozilla, from O'Reilly (of course.)

Re:The Master Of C

[Jorrit]

The difficulty of learning a language has little to do with the number of keywords it has. It is more a matter of how those keywords (and operators) are used together to enable programmers to do things. C is not that hard but in some cases it can get complex compared to other languages (even if those other languages have more keywords). Pointer arithmetic is one area in which many people struggle. Especially if you come from a programming language (like Java, ...) which doesn't have that.

C++ adds a number of keywords to the C language but not that much. Even so it is a highly complex language with lots of special constructs and exceptions.

Greetings,

Re:The Master Of C

[orthogonal]

Re C and pointers:

Understanding pointers is a challenge for many C newbies. It helps if you remember that a pointer is just a number, usually a variable but sometimes a constant, which represents an address in memory. It helps even more if you can remember that that address in memory is just an offset from the start of the memory (note this only works on architectures with flat addressing; it gets a bit hairier on, say, DOS).

Since C is a pass by value language, passing a pointer (or anything else) means passing a copy of a value. In the case of a pointer, that means a copy of the address, so now both the caller and the called function know that some memory exists, and can (in the abscence of const) modify it.

Since the pointer is just a number, it can be added to. Adding to a pointer actually means that the address is incremented by the amount of the addition times the size of the thing being pointed to. In other words, the compiler takes some pains to ensure one is alsays pointing to the start of what should be an object, and not within the guts of the object. So if I declare a a pointer to a 2 byte sized object, adding one to the pointer will point me to the next two byte oject in memory, and not to the second byte of the first object.

Naturally, the programmer must allocate the memory to be pointed to. The pointer is merely an address, it's not the memory being addressed. Declaring a pointer gives the programmer a place to write the address down; it does not create the mmemory: the fucntion call malloc ("memory allocate" ) "creates" the memory.

I beleieve the OP noted he came from an assembly background, which means that he's used pointers (untyped ones at that) so it may be less of a challenge.

The other C pitfalls for the unwary are the declaration structure, and (related) to this, the way arrays are handled. Several primers exist on decoding declarartion syntax; in a nutshell, declarations should be read from the inside out, left to right until a right parenthesis forces reading right to left. The declarartion:

int ( *f )( char ) ;
is read "f is a pointer to a function taking a char argument and returning an int.

In C, arrays are not first class objects, and so when passed to functions they "decay" to pointers to the first element of the array. An array name can be treated like a pointer, and indexing into an array is defined by the language to be exactly equivalent to dereferencing an array plus the index:
Given
char array[ 4 ] ;
array[ 0 ] is just *( array + 0 ); and array[ 3 ] is just *( array + 3 ). array[ 4 ] is *( array + 4 ) and is an illegal (well, undefined) dereferencing of the "element" one past the end of the array.

I have said too much for a post, so I'll end now. The point is, as long as one remembers a few key ideas: pass by value, pointers are numbers for addresses, and declarations need to be read inside-out, C is rather straigtforward.

Re:The Master Of C

[Frymaster]

yes, the basic *concept* of pointers is easy and the implimentation as outlined in the hello world calibre demo code that comes with most tutorials is simple... but the reality of memory management in non-trivial apps is quite different. why do you think there are so many memory profilers and leak-detectors on the market?

in obj-c memory mgmt is much better handled. when you make an object a "reserve" is put on it. when yr done with the object, you "release" it. you can add as many reserves as you want and release them when you wish. when the reserve count finally reaches zero, the object is deallocated. brilliant!

the skinny on obj-c and memory is here

if you find you have objects deallocating prematurely and are losing track of yr reference counting, get the object meter from omni group (makers of omniweb). there's a free demo license that works quite well for the budget-challenged.