[Androsynth]

I used to think: C is a horrible language for beginners, they should learn python due to its simplicity. This will allow them to learn about control structures and program flow in a simple and easy to read environment. But the more I think about it, the more I like C. It gives you the best understanding of what computations are actually being done by your processor; the mental exercises involved with fulling understanding pointers is valuable at the collegiate level (much less so if your learning to program on your own, I believe college is a time to increase your analytical abilities rather than just learn raw facts); It also has the simplest syntax to learn (for a procedural language).

You dont have to explain argv, argc or the details of scanf to beginners. Thats like saying python is complex because you need to explain iterators and generators. You don't need to dig that deep into the language as a beginner, you just need to be able to understand the structure, control flow and be able to get your programs working.

[ufo]

But do we really need to submit beginners to things like segmentation faults, arrays without bound checks and the lack of a reasonable string datatype?

There is no reason to use C as a first language when you can just as well teach all the basic ideas in a more friendly environment and then go through all of K&R in 2 weeks or so.

[Androsynth]

The problem with OOP is that its really not something that should be taught to people who are first learning to program. It is extra complexity and doesn't add any value for beginners. That is my problem with Python.

C has its disadvantages, but you get: 1-a simple syntax 2-good mental exercises in learning, implementing and debugging pointer based data structures 3-a good idea of what is happening at the processor level when a computer program is being run.

You simply don't get that combination in any other language.

There are caveats of course: theres the functional vs procedural argument (which is a higher level argument than this). There could also be a good point that OOP should be the very first thing a student should learn, but only if you teach them a smalltalk based language, not python, java, C++ etc.

For the context of my argument, I chose: 'for teaching first year CS majors their first procedural programming language, what is the best choice?'

[ufo]

Where did you get the idea that I was supporting OOP-first? Its a terrible idea too.

In any case, I think Python is a great procedural language out there for a beginner - you can use the classes as if they were just structs and go a long way with that.

[graeme]

I started learning programming in January using K & R. C is great. Doing so many things manually really teaches you what is going on. It's definitely challenging, but I think I'm learning faster this way.

[cheatercheater]

> But the more I think about it, the more I like C. It gives you the best understanding of what computations are actually being done by your processor

Yes, because C is exactly how computers work inside! SO CLOSE TO THE METAL

> the mental exercises involved with fulling understanding pointers is valuable at the collegiate level

I have learnt about mappings in mathematics in kindergarten. Literally. Then in university we have had exactly one sentence to flesh out the idea in ZFC. You really think C is so mind bending?

[ufo]

And at the same time we have a whole generation of people that don't know what a tail call is. C might be closer to the metal then most languages but it still is a leaky abstraction in some places.

And anyway, the real problem is not C being complicated. Its that the other alternatives are also perfectly reasonable while also being simpler.

[AdamTheCamper]

I believe it is quite close to the metal. When I had my Systems Programming course in first year of college we started with C, and then continued with x86 assembly. One of interesting parts was, when we compiled our C code to assembly and then compared with our hand-written assembly code. Was fun :)

[cheatercheater]

Do you realize that with pipelining, preemptive computation, multiple levels of caching, opcode optimization, NUMA, massive parallelism, and all that other stuff x86 is to hardware what Java is to hardware interrupts?

Answer: you don't because otherwise you would not have posted the above.

[Androsynth]

You're over exaggerating. You can teach students about processor theory, assembly language and C without getting into all the extra stuff.

All the extra stuff is built on that foundation, and learning C helps you understand that foundation better than any other language.

[cheatercheater]

This is not "extra stuff". It is how it works. x86 was made to be simple to hand-code, not to in any way reflect the hardware. And then, decades later, new developments to it are so that it's easy to compile to. The language went through two major paradigms without either having the point of x86 reflecting the hardware it abstracts. SIMD, NUMA, pipelining and caching is exactly how processors work right now. And they're radically different to what happened before the Intel Core family of processors. This is in turn radically different to what happened before the Pentium Pro family of processors. Yet all support the same x86 instruction set. The corollary is that x86 is an abstraction. You're being ignorant if you think x86 is like the opcodes being executed after the whole caching stage has happened.

[Androsynth]

I never said it was mind bending. I said it was a good mental exercise for young engineers to go through to fully understand pointers.

It's one thing to read about pointers, it's another to actually implement a linked list in C.

Again, I want to stress I don't think you should get awarded a PhD for it, but for first year CS majors: it is good practice to actually implement and debug pointer-based data structures.

[cheatercheater]

Yeah, but you're obviously arguing about that it's notable that you can learn about pointers. It's a tiny block of knowledge, really minuscule in comparison to other things which you can learn in the first year, such as the notion of an isomorphism, with assorted examples. I'd say this one can easily fill up about 100x more lab time.

[nivertech]

C is not CLOSE TO THE METAL. It lacks concept of SIMD instructions and barriers/fences. If you want to be CTM learn OpenCL / CUDA (unless you want to learn assembler for actual ISAs or SIMD extensions).

[cheatercheater]

Thanks for spelling it out, I am surprised there are people in this year and age who would not get that "C is close to the metal" is a trite joke.