Identify computer science algorithm exercises

[kytrinyx]

We currently have at least one, maybe several exercises that asks the learner to implement a specific algorithm.

We discussed whether this type of exercise is a good fit for Exercism, in exercism/discussions#124 and concluded that:

Instead of asking "Write a program that implements erastosthenes sieve." we should ask "Write a program which returns a list of prime numbers, there are a couple of efficient algorithms: erastostenes sive, wheel sieve, etc", perhaps even with links.

We need to identify which exercises target an algorithm rather than targeting what the algorithm does.

The one that started the discussion in sieve, but I'm not sure if that's the only one.

Once we've identified them, we need to open a new issue to implement the new exercise and deprecate the old one.

Exercises identified as too specific

• sieve
• binary-search

[petertseng]

binary-search asks for a specific algorithm.

[ErikSchierboom]

The diffie-hellman exercise is all about a specific algorithm.

[tejasbubane]

luhn is another specific algorithm. Although I am not aware of any similar ones like the sieve example above. Maybe rename luhn to credit-card-validator to remove the specificity?

[Insti]

Are there other algorithms that will pass the tests for diffie-hellman and luhn?
With sieve and binary-search there are.

Is that the difference between a "good" and a "bad" specfic algorithm question?

[ErikSchierboom]

@Insti I think that's it!

[petertseng]

Oh. In that case, it's time to suggest parallel-letter-frequency... because I imagine most tracks' parallel-letter-frequency test suites don't specifically enforce that the implementation is indeed parallel, so a serial solution (whether it is intentionally or unintentionally serial) could pass the tests, right?

(Or should the tests enforce parallelism, and how? Like exercism/haskell#542 ?)

[Insti]

hangman?

[Insti]

binary?
(although it is deprecated.)

[emilianbold]

The core of the matter for me is not about implementing a specific algorithm. It's that the algorithm does not help me improve for that particular language and that sometimes I spend more time on the algorithm than on the language.

Also, I'm not learning how to program, I'm just practicing a new language. Do I really need that many exercises teaching me recursion in this new language? I know it helps to repeat some concepts, but at some point, it's just one exercise after another teaching you nothing new.

One good exercise, "Saddle Points" on Haskell, it's pretty hard to solve with the given method signature. I took it as a challenge and solved it. And I learned a few things about the related APIs. Here the algorithm was easy, it was the hunting for the useful hooks that was educational.

"Parallel Letter Frequency" was hard because you are given no hint on what to use. I read a lot of (some, dubious) articles until I decided to use the simplest function I could use. And it worked.

Right now, after "Robot Name" which was good as I used a MVar I'm offered "Pig Latin" which is pointless, then "Say" which is not much at this point, then "Spiral Matrix" which is just algorithm writing.

I feel like the following exercises are missing a chance to teach me some new language concept or very common language module / function / pattern.

[kytrinyx]

Thanks @emilianbold this makes a lot of sense, and I think you're right.

We definitely need to retire exercises that are not teaching anything.

I don't know whether we know which ones feel like they're not teaching anything though, and this is likely different from language to language.

/cc @iHiD @nicolechalmers not directly related to the redesign, but a question we might want to discuss in order to have some ideas about how to figure out which exercises are not helping.

[stkent]

I would imagine whether or not an exercise teaches anything can also vary from person to person.

[iHiD]

Let's rephrase it to "exercises that are teaching things aligned with the goal of exercism, which is to increase fluency in a language - not to develop computer science skills"

…

On Mon, 3 Jul 2017, 23:40 Stuart Kent, ***@***.***> wrote: I would imagine whether or not an exercise teaches anything can also vary from person to person. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#761 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AARfDDeaKK0NGjfmG5uzEUkFcMlBRqZmks5sKW3DgaJpZM4NAsPD> .