Hacker News new | ask | show | jobs
by joshuata 3382 days ago
It was even worse starting in C++. Almost half the code in any project was opaque black magic. What is char and why do I need it? Why do I need `using namespace std`? Why do I need to create a Pair in order to insert into a Map?
2 comments

I don't think it's worse in C++. You can make purely imperative C-like programs in C++ with essentially no boilerplate (with the exception of maybe needing io from the standard library), and build things up as they go.
While I still am a big fan of C++, I'd attest to the fact that is a worse language to teach beginners. Funnily enough, that's not so much because of the language, but because of its legacy.

I learned programming with Borland Turbo Pascal and later Delphi and have to admit I struggled with the OOP mindset at first. The according chapters of the book I used were filled with contrived examples that left me saying "I can also do this the old way!". Apart from this, I found the book to be very well written, though.

Then, I tried learning C++ (using a Book frome the same series) with Borland C++ Builder. Boy, was that a mistake. I didn't get around to understand why I had to keep using -> instead of . everywhere (before, I almost never used a pointer anywhere), the whole delete-dance was a nightmare and the Borland tooling was so filled with magical clicks that generated event handlers that I never understood how the whole thing really worked (you never got to see the whole code). Headers vs. .cpps was a similar disaster, and compiler errors could only be met with comparing your code with the example on a symbol by symbol basis[1].

So, while C++ is a nice language (far better than Java IMO), starting in it was a nightmare. The problem here is that you have to explain way too many concepts at the beginning (which still holds true even with C++17), and since APIs like the Borland GUI libraries or win32 are based on so much cruft and ages old crap, you'll never get to the point where a beginner can say "Look, I've drawn a circle!".

Honestly, if you ask me, programming courses should best just start with a modern version of Python. But even then you'll have trouble setting up on Windows and getting over the whole UI frameworks disaster.

[1]: This kills experimentation. That's the absolutely last thing you want to do if you're trying to teach a motivated kid how to program.

Edit:

Also note: Often, the mindset seems to be that "beginner" means someone at a university. In my opinion, that's a fallacy too. I started out when I was 11 - I'd never even had any decent english course in school back then. Think of that, and you'll start to see the flaws in starting out with some way too complicated concepts. We should get rid of the notion that a learner has to "push through" things (especially at the start!). It will be to the benefit of society.

The problem there seems to be that you were learning an IDE/Gui builder and c++ at the same time. I still think I language that exposes pointers directly (c, c++, others?) are far better than something like python. Pointers force you to understand memory, allocations, the heap and stack, etc. There are way too many professional programmers that don't understand them.

  The problem there seems to be that you were learning an IDE/Gui builder and c++ at the same time. 
Even from a modern persective, I'd disagree with that. Yes, the IDE was a pain point, but it existed because the alternative is much worse. You don't want to start out explaining Makefiles to anyone (even if you start with VS you'll be spending an hour setting up instead of programming).

  Pointers force you to understand memory, allocations, the heap and stack, etc. 
Yeah, and we'll both agree that this is crucial knowledge for any professional programmer. But if you're starting out (we're talking first language here), it's in the way. You don't even know what a loop is yet. You're at least years away from professional.

Also, not everyone who programs wants to become a software developer. Sysadmins don't need pointers, but a readable alternative to shell scripts is nice (I know such people, they exist). A journalist who wants to make a flashy chart on the internet doesn't need to know pointers. A biologist also doesn't - but (s)he can certainly benefit from programming. Have you seen what Jupyter can be used for?

Edit:

As a side note, I also see much value in starting with a more functional language than C++.

Lexical scope really affects your thinking. If you can get to that at the end of the first course, you'll have taught something tremendously useful. I've been working at a local university of applied sciences until last year, and you really don't want to know how many students there struggle with a C++ lambda expression.

Again, to reiterate: I'm a fan of C++. I still cringe at most other imperative languages. But I see its weaknesses (especially in teaching).

> Also, not everyone who programs wants to become a software developer. Sysadmins don't need pointers, but a readable alternative to shell scripts is nice (I know such people, they exist). A journalist who wants to make a flashy chart on the internet doesn't need to know pointers. A biologist also doesn't - but (s)he can certainly benefit from programming. Have you seen what Jupyter can be used for?

In context, this is about CS classes, while the others might do some programming they shouldn't need to do computer science. Also, although being easier, is jupyter doing anything not being done in emacs years ago?

>You don't want to start out explaining Makefiles to anyone (even if you start with VS you'll be spending an hour setting up instead of programming)

To anyone on the path of becoming a professional programmer, yes, it's worth a few minutes to write a trivial makefile. It's worth teaching how to compile from the command line and how to batch those commands with a DSL is really simple. Again, there are way to many programmers who have no idea what their IDE is doing, which really isn't that much.

> Yeah, and we'll both agree that this is crucial knowledge for any professional programmer. But if you're starting out, it's in the way. You don't even know what a loop is yet. You're at least years away from professional.

c/c++ don't force this straight away though, you can easily do for loops before you begin managing memory. In any language memory should be understood before arrays are.

> Also, although being easier, is jupyter doing anything not being done in emacs years ago?

While I already think it's worthwhile that you can sidestep emacs' setup and learning curve for uninterested students, you might be interested in how it is already used for collaboration and teaching.

One list of examples might be this:

https://github.com/jupyter/jupyter/wiki/A-gallery-of-interes...

If you're teaching math or physics to a bunch of students who know Python and SciPy, I would assume Jupyter notebooks are an amazing way to distribute course material.

> In context, this is about CS classes, while the others might do some programming they shouldn't need to do computer science.

We seem to read the same statement differently. I actually had a CS101 class in highschool (also in Pascal, I didn't learn much there ;-) ). While all students there wanted to go in the engineering/science direction, I think only a hand full of us ended up in CS. Granted, this might be due to terminology.

On the other hand, I work with people who studied "business informatics", which at a bachelor's degree has the same base classes. I can guarantee you none of them use a pointer in their daily work (again, they do more in the sysadmin direction).

> To anyone on the path of becoming a professional programmer, ...

Yes, but not in the first lecture. Not before they know what a loop is. Many don't even know the command line (yes, I was also stunned to see that).

On the other point, I entirely agree :). I could make a huge list of frustrations I have with professionals not being willing or able to introspect how things work.

> c/c++ don't force this straight away though, [...]

As I said, it's not so much of the language anymore, more its legacy. C++ has stuck around quite a while, and this results in an abundance of old and bad APIs.

And with the better ones, you'll be able to leave pointers aside, but end up ignoring templates instead... Also, already the output operator on STL streams means you'll have a lot of explaining (or ignoring) to do.

Additionally, if you really are already teaching computer science students, I'd say you'll give them a better time explaining pointers and so on to them after they already know a language they can certainly use in the industry - in which case the idea of using Java to start out still has a point...

Edit:

This basically leads back to the original argument of the article: If you change the starting language for beginners, do it for their learning progression. If you're changing it to C++ for C++'s sake, you won't be doing your students a favor, stick with Java. If you can tolerate telling your students "hey, look, this magical thing is actually that feature here", you can also tell them later on that C++ is basically much like Java except for some details. I think language switching isn't as hard as it's made out to be if you're teaching the right concepts.

It would of been better to start you with C then transitioned to C++ later on or skip it all together and go directly to java/ocamel/haskell then come back to C++ at a later date.