nucleotide-count: Nucleotides have a misleading type

[isovector]

The function nucleotideCounts :: String -> Either String (Map Char Int) I'd suggest has the wrong type. Nucleotides are certainly an enum and not a Char. I worry the exercise as it stands today discourages people from thinking about data modeling and instead plays into a bias to use strings more often than they should be.

[petertseng]

I agree -- an enum would make sense so the type system can help make clear there are only four possible values at each position.

I would also take this opportunity to address what was said in #626 (comment)

While I think that enforcing invariables in the type system is a great idea, sometimes it is awkward to do that in Haskell. I don't think we should push all the exercises in this direction when it feels artificial in the language.

About nucleotide-count, using length-parameterized types is idiomatic in Idris, but it is still considered type-hackery in Haskell, which is an advanced topic and would confuse students.

I still think there would be value in using a data for nucleotide-count; we don't have to do the Idris-style length-parameterisation.

I believe the files needed to edit to resolve this issue are the following:

• https://github.com/exercism/haskell/blob/master/exercises/nucleotide-count/test/Tests.hs
• https://github.com/exercism/haskell/blob/master/exercises/nucleotide-count/src/DNA.hs
• https://github.com/exercism/haskell/blob/master/exercises/nucleotide-count/examples/success-standard/src/DNA.hs

[sshine]

I agree. But if we use a

data Nucleotide = A | C | G | T

then what's the point in having a Map Nucleotide Int when you could just have a

data NucleotideCount = NC { _a :: Int, _c :: Int, _g :: Int, _t :: Int }

but then you wouldn't need the Nucleotide data type if the original input is a String.

Perhaps one should keep the Map Nucleotide Int since that adds complexity.

[sshine]

Should the tests keep asserting that the map contains all four nucleotides when a nucleotide occurred zero times?

[isovector]

@sshine I personally wouldn't let the NucleotideCount described above through in a real code review---it doesn't compose well with the ecosystem. Eg, writing Nucelotide -> NucelotideCount -> Int requires a pattern match that the compiler won't help you write correctly. When in doubt, use the definitions that are the hardest to get wrong.

[sshine]

@isovector Do you mean something other than what I did in PR #714?

[isovector]

@sshine nope, #714 looks great to me!