3 ms·
Picking a language to teach programming and concepts in computer science presents an interesting conundrum. Do you pick a commercially relevant language (which
by criddell 6y ago
Picking a language to teach programming and concepts in computer science presents an interesting conundrum. Do you pick a commercially relevant language (which most students want) or go with one that might be better for academic purposes? I probably lean mostly towards the former because time spent learning something like Eiffel feels somewhat wasted.
- qsort 6y agoOn the other hand, there are concepts (closures, objects, pointers) that come more naturally with some languages, and you absolutely need to learn the concepts even if you don't end up using the language. Don't pick one, pick a healthy mix.
- dehrmann 6y agoTo me, the dilemma is between a language that's easy to learn and one that is useful once you learn the language. Most languages have too many warts and footguns to be easy to learn, but learning a toy language isn't good, either, because it's not useful.
- lanstin 6y agoNo one should have a computer science degree and not know multiple languages and multiple programming paradigms. Prolog, lisp or scheme, C, JavaScript, Haskell. The long term task of the professional software developer is to learn new stuff. Helps if your awareness of the cool things you could learn is already broad. I cannot explain how wrong it is to graduate with Java only and think that is sufficient. To be honest, I would also want people to understand the halting problem and Godel’s incompleteness theorem’s proof, as understanding the equivalence of apparent different systems that can handle the integers is very fundamental. And these days some knowledge of distributed theory is also good, the Lamport time clocks paper and so on. Unless the student is ambitious and doesn’t really like programming and just wants to program for a few years and then get into management or something like that (product? bizdev? I’d that a thing?) they will go thru many iterations of interesting new technologies. The brain that enjoys learning mind-bending shit will have a broader conceptual base and a better chance to see how lisp macro type things can help with k8s configs. Or while lazy argument evaluation might not be in your language, having it in your tool box for message processing design or you know evaluation of user preference rule sets against an event might be helpful.
- klibertp 6y agoNo! Eiffel should be the default OOP language for any kind of course. Being forced to think about pre- and postcoditions and invariants by the language is a really good thing for beginner programmers. While the syntax is somewhat verbose (mainly because of using keywords instead of most punctuation) Not teaching about Design by Contract is yet another billion dollar mistake in computer science. There are DbC libraries or frameworks for just about any language, but only very few programmers know about its existence, which leads to criminal underutilisation of the technique in the industry. I can't be sure, but I think there's a lot of thought, at least, behind selecting a particular language for an introductory course on programming. I don't know which language I would choose, but I'd like something that would broaden the horizons of student as much as possible. It could be Mozart/Oz or something similar.
- yesenadam 6y agoI didn't know anything about Eiffel or DbC, just reading about them now. "Eiffel Software applied for trademark registration for Design by Contract in December 2003, and it was granted in December 2004." https://en.wikipedia.org/wiki/Design_by_contract https://en.wikipedia.org/wiki/Design_by_contract Is that possibly something to do with it not commonly being taught about/learnt?
- klibertp 6y agoI have no idea, unfortunately. On the other hand, the trademark doesn't seem to be a problem to guys behind Racket, which has the most advanced contracts system out there; the canonical book for programming classes in Racket, "How to Design Programs", teaches you everything you need to know about DbC without ever mentioning the term itself, IIRC. Even if you don't want to touch Eiffel due to it being proprietary, there are many languages which support DbC you can choose from: already mentioned Racket, Ada, Clojure, Scala, D, Nim, Haxe, Elixir, and many more. Also, there's a difference between DbC and simply using contracts - it's similar to the difference between writing tests and doing TDD. Many people might not be convinced that the benefits of DbC are worth rethinking the development process, and so they dismiss all contracts as "too many asserts, don't need". But, like with tests, you can benefit from a solid test suite even if you don't do TDD; similarly, you can reap a lot of benefits from simply ensuring that precondition violations are caught at the boundary of a subroutine, because it makes it easier to find the reason beind the problem. I'm convinced that in the next 10-15 years we will see gradual typing (with type inference) embracing contracts as a way to extend the guarantees from compile time to runtime. Maybe we will finally get the best of all worlds: rapid prototyping, great performance, and verifiable program correctness.