It appears I've sparked a bit of controversy. In hindsight, I regret the phrase about "Learning how the computer works as a machine." -- I reality, I meant that C exposes a lower-level, logical model (not physical as the word machine would imply) of the machine.

As has been noted, everything above the microcode layer is an abstraction, technically speaking. So I think the question is really this: At what level of abstraction do you find the optimal position for a beginner to enter into the world of programming?

For the majority of people of the bottom-up mindset, I believe that position to be C (or another small, procedural or even functional language) for 95% of people, with Assembly picking up the remainder. The top-down crowd would be more at home with C# or java or even C++ (though again, I don't recommend it as a first language.)


Why do I believe that C is the optimal beginner language for so many people? I believe that because it introduces programming concepts (storage, conditionals, loops, etc) at a sufficiently high level to be friendly, but also doesn't introduce a lot of high-level concepts of its own in the way that Java, C# or even C++ does. Null-terminated strings are the only abstraction that comes to mind (and I do agree that null-termination was a bad design choice in the long term, probably driven by memory constraints at the time.) While, at the same time, it also exposes the low-level workings of the computer (again, as a logical model not the physical machine.) and facilitating a programmer to create their own higher-level abstractions.

C was, after all, explicitly designed as "a portable assembly language" -- the lowest convenient abstraction above disparate architectures. And while I also believe that a modern reinterpretation of C might benefit by including concepts such as multiple processors and the memory hierarchy, I'm glad C itself does not -- whether by design or glorious accident -- because it could not have served its position as least common denominator if it had.

This is getting out of hand; lets return to the origin of this off-topic argument:

Quote:Original post by ravyne2001
On the other hand, C is a great language to learn how the computer works as a machine, because it provides no high-level abstractions of its own and essentially frees the programmer to do whatever they wish (Hardware/OS restrictions aside.) You can, for example, impliment features in C that are commonly found in OO languages like C++ or C#, such as polymorphism. C can give you a window into the workings of things that you might take for granted in another language. Another plus is that C is a much smaller language than, for example, C++ (in both the number of keywords and the scope of its built-in features) so you can get your head around it more quickly.

Ravyne's statement that "C is a great language to learn how the computer works as a machine" is the genesis of this off-topic debate.

This statement can be interpreted a few ways, such as:
A) Learning C will teach you how a computer works as a machine.
B) If you are learning how a computer works as a machine, then C will aid you in your studies.

Anyone with a degree in computing will know that statement (A) is false, and seeing as Ravyne isn't a newb, it's pretty safe to assume that he intended this statement to be interpreted as something closer to (B), especially when reading the surrounding context of the statement.
However, cshowie assumed that Ravyne actually meant (A) and decided to start a debate about whether this misinterpretation of Ravynes argument is false or not.

Any debate about statement (A) is useless, because noone actually made that statement. Debate based on this controversial statement should be aimed at whether C is a useful aid for students of computer science, not whether C is a useful teacher of computer science!

Nathan Baum then joined the debate, backing up cshowie on his irrelevant argument against the non-existent statement (A). As I already mentioned, these additions to the debate are useless, because there is only one side to this debate (cshowie's view VS cshowie's misinterpretation of ravyne's view).

Baum's comments aimed at me were not actually against statement (B), they were against statement (A). Seeing as I was putting forth an argument in support of statement (B) (that C is lower level than C#, therefore suitable for implementing lower level concepts), and that we all already know that statement (A) is not only wrong, but irrelevant, I found his comments to be of a condescending tone (Just as if during an argument about how to best slice a pizza, your opponent started explaining to you how to tie your shoes as if you were a child).

Ok? So if you're arguing that learning C will not teach you computer science, then you can stop now, because noone has or is supporting that statement.

Quote:Original post by Crypter
They are C concepts, not low level concepts. Strings dont need any null termination. Pointers are used ALOT in low level programming, but only as dereferencing addresses and address arithmetic.

Arrays, pointers and null-termination are not just C concepts. The "null terminated string" may be used by C, but null termination is just a general computing concept that any student will learn at some point.
Therefore if you are a student learning about the general concept of null termination, or arrays, or pointers, then C will be a useful tool to aid you in your studies of these concepts.

Quote:Original post by Crypter
*list of facts*

Noone is disputing this.

Quote:Original post by MaulingMonkey

Quote:

Quote:Original post by Nathan Baum

Quote:
e.g. #1 Learning how to deal with strings correctly in C will teach you a lot more about "how the computer works as a machine" compared to using a C++/C#/Java/etc string class.

No, it'll teach you about how C-style strings work. Given that they work the same on all implementations of C, even when running on machines with wildly differing architectures, that should be a clue that string handling doesn't reveal the secrets of "how the computer works as a machine".

Again, you're not arguing against my point - I didn't say that it would "reveal the secrets", I was trying to illustrate that using lower level constructs teaches you more about the machine than using higher level constructs.

No, it teaches you more about just those "low level" concepts. Are those concepts useful to programming, or actually teach you anything about how computers work? No. They'll show you a sliver of how they worked on a 286, but modern hardware merely emulates that model for backwards compatibility.

Hell, even x86 is a bit of a sham now. Modern CPUs merely decode it into their own internal microformats for actual execution.

Noone ever said that learning C will teach you how a computer actually works.
What I was trying to say is that if you are learning how a computer actually works, then a lower level language, like C, is a more appropriate choice for practising these concepts than a higher level language.

Computers still use arrays, computers still use pointers, computers still use null termination. If you're a computer scientist you will know these basic concepts, to know them you must learn them, to learn them you must use them. An appropriate way to use them is to learn C.

Quote:Original post by MaulingMonkey

Quote:

Quote:Original post by Nathan Baum

Quote:
e.g. #2 You can write a program in C that accepts an arbitrary hexadecimal value from the user and tries to read 17 bytes from that memory address (which may well cause a seg-fault etc...). Higher languages such as C#/Java won't let you do silly low-level things like this because they are *more abstracted*.

C# will let you do silly things like that, not that it matters. The ability to crash your program by probing random memory addresses does not make a language "closer to the metal". BASIC allowed you to probe random memory addresses, yet it was, in many ways, "higher level" than C.

Ok, I'm not that experienced with C# so scratch it from my quote. My point still stands though - AFAIK, Java's version of 'pointers' ('references') can't be assigned to arbitrary memory addresses - the language's abstraction of memory is "more abstract" than C's abstraction of memory.

I thought we were talking about C#, not Java. I mean, it's just in the freakin' thread title. Yes, there are HLLs that will forcibly abstract things even higher, we generally don't have much love for Java around here.

In the context of that quote, I was talking about HLLs in general (such as Java/C#) in an attempt to illustrate that they are more abstracted than lower level languages. It's a fairly simple concept (that higher level == more abstract), but for some reason argumentative people keep trying to disprove it...

Quote:Original post by MaulingMonkey
Forcibly deabstracting things like C does, however, is also the wrong answer. With flexibility comes the ability to do whatever you want whenever it's appropriate. Telling a beginner to use a language that's unsafe by default in it's very design, before mastering even the most basic of concepts of programming -- even the simplest things such as the flow of execution -- is not appropriate.

You're arguing against points that haven't been made! I didn't recommend C to beginners, I just supported ravyne's view that low-level languages are ok to use when learning low-level concepts.

Quote:Original post by ToohrVyk
Lots of info on pointesr

Sorry ToohrVyk, when I said compared C pointers to numbers and HLL pointers to numbers in bubble wrap, I didn't intend you to read that far into it. What I wanted you to focus on was the bubble-wrap part of the analogy (such as garbage collectors, etc...).

I know that you're from Australia, but do you realize that "no one" is, in fact, two words?

Quote:Original post by Hodgman
Ok? So if you're arguing that learning C will not teach you computer science, then you can stop now, because noone has or is supporting that statement.

Most of the points made in favour of C in this thread focused on the "features" of C, such as null-terminated strings, pointers, and similar concepts. Most of the counter-points explained that none of these concepts actually provided any relevant information about how the computer works as a machine. These features may well teach you about some implementation details of some high-level languages (though most HLLs use some tricks that cannot be explained using C concepts, such as tail recursion).

Quote:Arrays, pointers and null-termination are not just C concepts. The "null terminated string" may be used by C, but null termination is just a general computing concept that any student will learn at some point.

I would argue that 1° null-termination is a narrow and rarely used feature (outside of C) that deserves less focus than the more important concepts of type-safety, polymorphism, or concurrency, that 2° C++ iterators are a much better way of learning about pointers, and that 3° arrays are an almost universal feature of programming languages.

Therefore if you are a student learning about the general concept of null termination, or arrays, or pointers, then C will be a useful tool to aid you in your studies of these concepts.

Quote:Original post by Crypter
*list of facts*

Noone is disputing this.

Quote:e.g. #1 Learning how to deal with strings correctly in C will teach you a lot more about "how the computer works as a machine" compared to using a C++/C#/Java/etc string class.

Quote:Noone ever said that learning C will teach you how a computer actually works.

Contradiction.

Quote:What I was trying to say is that if you are learning how a computer actually works, then a lower level language, like C, is a more appropriate choice for practising these concepts than a higher level language.

This, we know. However, you have so far only provided a list of C features that do not actually teach, or help learning, anything useful about the inner workings of the computer. You have mentioned C-strings and pointers, for example, none of which have any relevance or bearing to the way the computer works.

The correct statement, where everyone would agree, is that C teaches you how low-level programming works. But both high-level programming and low-level programming have become completely disconnected of the actual way the computer works.

Quote:Computers still use arrays, computers still use pointers, computers still use null termination.

Actually, programs use these. Computers have no notion of array, pointer or null termination: these are merely programming language concepts, not hardware concepts.

Quote:In the context of that quote, I was talking about HLLs in general (such as Java/C#) in an attempt to illustrate that they are more abstracted than lower level languages. It's a fairly simple concept (that higher level == more abstract), but for some reason argumentative people keep trying to disprove it...

Actually, everyone agrees with this. What we don't agree with is that "lower level == closer to the machine", which isn't the case. See below.

Quote:Sorry ToohrVyk, when I said compared C pointers to numbers and HLL pointers to numbers in bubble wrap, I didn't intend you to read that far into it. What I wanted you to focus on was the bubble-wrap part of the analogy (such as garbage collectors, etc...).

So, let's focus on the bubble-wrap part. Java provides garbage collection while C doesn't. Therefore, the abstraction of the Java memory model is actually closer to the machine than the abstraction of the C memory model, by virtue of the machine using a garbage collection scheme to allocate, deallocate and move objects around. C pointers are in fact implemented as Java references behind-the-scenes, but C programmers never know it. I could even go as far as saying that a compacting garbage collector like that of Java or C# teaches important things about the importance of cache coherency, a subject totally eluded by the C language, as previously noted in this thread.

In short:

• The C language has a very simplistic view of the world.
  
• Higher-level languages have a more complex view of the world.
  
• The computer, as a machine, is extremely complex.

Some of the complex concepts seen in higher-level languages are bound to represent and explain the hardware-level concepts.

Quote:Original post by SLaYaH
I am stuck trying to decide, originally I was only deciding between C or C++ but now theres C# and I'm not sure what to do. If at all possible I'd like to know pros and cons of each and whcih you would suggest to learn coding for an ALL AROUND programmer, not just for gaming but may be part in gaming.

Thanks in advanced.

Oh god, not another one of these threads? Do we not learn people?

Here's some tips: Search the site for C# vs. C++, C# vs. Java, C# vs. Godzilla and so on.. there's a billion of these threads and they all end in tears. Promit did a nice catalogue of the threads a while back.

My bit of advice since you seem to be new:
Pick one, try it, don't like it? Move on and until you find something you're comfortable with. Believe me, there's plenty of information about each language floating around and enough people here proficient in all 3 to give you advice should you need it. In all honesty, it doesn't hurt to know as many languages as you can. It'll give you a broader perspective. At the end of the day you should eventually try and learn all 3 and many more besides if you have the time and patience for it. Just don't bite off more than you can chew.

Good luck.

Quote:Original post by Hodgman
Noone ever said that learning C will teach you how a computer actually works
...
e.g. #1 Learning how to deal with strings correctly in C will teach you a lot more about "how the computer works as a machine" compared to using a C++/C#/Java/etc string class.

This isn't a contradiction...

By "how the computer works as a machine" I meant low level computing concepts in general (as I've already said, multiple times - stop it with the straw men already please!), using the example of all the crap you have to do manually in C compared to a language with proper string support.

In one approach you just use a few function calls, in the other you have to manually use basic computing concepts, therefore, the manual approach requires more knowledge about how computers work at a lower level.

Quote:I'm elaborating further on the cons of your suggestion

That's ok, I just wanted to point out that I never actually suggested that a beginner should choose C over something else (you implied that I did), just that it can be a useful learning tool.

Quote:Active memory as a single integer-addressed whole is **the** bubble-wrap. Garbage collection, however, is clearly a fundamental concept to programming logic. Not one every project needs to use (at least as an automated process), but one nevertheless.

It may be bubble-wrap, but it wasn't THE bubble wrap I was talking about.
It's already been said many times that C is an abstraction (hence a C pointer is going to have bubble-wrap compared to what's really happening), I guess I should have explicitly said that *managed* pointers have *more* bubble wrap than *raw* pointers.

Quote:The fairly simple concept that I'm trying to explain to you is that "totally lacking anything higher level whatsoever" isn't a necessary prerequisite to having low level access.

Ok, I've never disagreed with that.
There's nothing wrong with making your own string class in C# or your own linked list in Java - you can still learn these low level concepts in a HLL. As the above quoted example shows, I was saying that doing them in C (i.e. manually) will teach you more about these low-level concepts than using a string API.
If you're doing them manually in C and manually in C# then there's not much difference (the main one being that you don't have to do it manually in C#).

Quote:C# still uses arrays, pointers, and yes, null termination if you need to support that ungodly legacy hack of a string format. You can do it, just like you could in C. If you're a computer scientist you will know such basic concepts such as garbage collection. To know them you must learn them, to learn them you must use them. An appropriate way to use them is to learn C#.

Yeah, I agree.


So, seeing as C# can do all the (unsafe) things C can do, you could use it to learn all of these old C techniques. However, I (personally) still think it's appropriate to learn them using C for the same reason that you don't conduct survival training in a hotel resort - you do it in the desert.
(On that note, it's probably important for programming students to use some assembly code at some point too, to understand even lower concepts, like how a function call is implemented -- but don't take this as a suggestion that you should build apps in ASM over C#!!!)

Quote:Original post by ToohrVyk
Most of the points made in favour of C in this thread focused on the "features" of C, such as null-terminated strings, pointers, and similar concepts.

As I've already said, I pointed out these "features" as part of an example of the amount of manual work required in C compared to a langage with a good standard library.
I.e. the important part is that you do it manually, not that you use a particular "feature".

Quote:Original post by ToohrVyk
You have so far only provided a list of C features that do not actually teach, or help learning, anything useful about the inner workings of the computer. You have mentioned C-strings and pointers, for example, none of which have any relevance or bearing to the way the computer works.

Again, I've already said that I wasn't talking about the "inner workings of the computer", but the "low level workings of the computer".

Quote:Original post by ToohrVyk
Java provides garbage collection while C doesn't. Therefore, the abstraction of the Java memory model is actually closer to the machine than the abstraction of the C memory model, by virtue of the machine using a garbage collection scheme to allocate, deallocate and move objects around. C pointers are in fact implemented as Java[-style] references behind-the-scenes, but C programmers never know it. I could even go as far as saying that a compacting garbage collector like that of Java or C# teaches important things about the importance of cache coherency, a subject totally eluded by the C language, as previously noted in this thread.

Now this is a counter to the argument about LLLs being closer to the machine than HLLs.
However, my point about doing things manually being a better learning activity than using an API (and C being a good language for this exercise) still stands.

[Edited by - Hodgman on August 30, 2007 2:29:24 AM]

I just want to point out that null terminated strings are a C concept. They have nothing to do with the low level workings, or inner workings, or any of that bullshit. Most languages don't even use null terminated strings; C++, Java, C#, and Pascal/Delphi all store explicit lengths.