Initial diamond text support in test suite

[chiroptical]

Add Data.Text support for diamond from #841

@sshine Should I just modify the README.md or add a .meta/hints.md similar to https://github.com/exercism/haskell/blob/master/exercises/bob/.meta/hints.md?

[sshine]

Thanks for the PR!

You seem to have uncovered a pretty significant problem of #841:

Test suites with property tests don't overload as easily as when we only need to compare for equality.

I can think of three approaches, one of which you have picked:

• Write a thin ToString class, as you did.
• Use an existing low-dep package like string-conversions to do the same. (It does pull in bytestring for no purpose beyond letting people solve the exercise with that, too. This exercise only has byte-sized characters in input and output.)
• Consider going Text-only for this exercise and future exercises in a similar situation. Since diamond is currently unlocked by hello-world, maybe this has implications, and maybe that's okay as long as the stub has the right imports; this isn't set in stone, and it's a principal decision either way.
• (Edit: In the spirit of my co-maintainer) Or something else entirely.

I am inclined to let you decide, but we can give @petertseng a couple of days to respond.

[sshine]

@barrymoo: What would you prefer?

[petertseng]

By the way, just so my position is known:

I agree this is unfortunate. I'm tentatively open to any of the proposed solutions, and don't have any to add at this moment.

I think we're waiting for something to be added to the README, yes?

[sshine]

Yes, I'm sorry, I completely forgot that. @barrymoo:

• Decide which option you prefer to go with
• Add a .meta/hints.md that resembles e.g. Port raindrops test-suite to support Data.Text #853
• At the track repo's root, configlet generate . --only diamond
• Commit and push both of the updated .meta/hints.md and README.md

This will merge the hint with README.md in such a way that passes CI.

[chiroptical]

@sshine

Decide which option you prefer to go with

I really don't think this is a decision I should make. I am split between string-conversions and Text-only. Many folks just starting out (including me) might prefer to use String because they are comfortable with it. On the other hand, learning and using Text is likely what you will encounter in the "real" world.

[chiroptical]

The text package actually depends on bytestring anyway (http://hackage.haskell.org/package/text). I think this makes me lean towards the string-conversions package.

[sshine]

@barrymoo: I'd go with string-conversions then.

[chiroptical]

I'll update this PR tonight with string-conversions and metadata.

[sshine]

Hi Barry, thanks for your work.

I've pushed two small changes to the feature branch.

I think this is ready to go. As per our recent custom, I'll give other maintainers 48 hours to interject.

---

(The rest of this is partly a note to myself, partly a recognition of the pain points you experienced in contributing.)

I saw parts of your twitch stream and collected notes where my specification was unclear to remind myself to document and/or provide this information in the future.

• .meta/hints.md is optional and can be created if it doesn't exist.
• Hard to find examples of hints in exercises: Caused by a bug we're fixing in #864.
• Saying diamond :: Char -> Maybe [Text] in the hint twice does seem superfluous. Let's keep it like this for now. We can revisit changing the Data.Text hint into a shorter form for all exercises for which this applies.

As for:

• "Do I really wanna put configlet in bin/?" I had mixed feelings.
• "problem-specifications: Path not found": Kudos for assuming this was an Exercism repo.

I should have told you that bin/ensure-readmes-are-updated.sh fetches configlet and clones the problem-specifications repo for you; I forgot to tell you this, because I use configlet directly; the advantage of bin/ensure-readmes-are-updated.sh is that it automatically downloads dependencies if they're not already available.

• Your quest to explore the CI log: You initially overlooked the error and stopped at the second occurrence of red color, which happened to be a false negative. (This is addressed in #783, but not solved yet.)

[sshine]

As a side note unrelated to the completion of this pull request, string-conversions has two interfaces, the other one of which is in the module Data.String.Conversions.Monomorphic in which there's a function toString

toString :: ConvertibleStrings a String => a -> String

that fixates the output type. Some people (such as the author of the string-fromto package) go even further and eliminate the use of type classes here. I've not personally experienced annoyance at type-inference ambiguity when converting strings, but I probably haven't written enough Haskell.

A comparable example you may have run into is when you write x ^ 10 and GHC will say:

$ stack ghc -- -Weverything Pow.hs
...
Pow.hs:5:15: warning: [-Wtype-defaults]
    • Defaulting the following constraints to type ‘Integer’
        (Integral b0) arising from a use of ‘^’ at Pow.hs:5:15-20
        (Num b0) arising from the literal ‘10’ at Pow.hs:5:19-20
    • In the first argument of ‘print’, namely ‘(2 ^ 10)’
      In the expression: print (2 ^ 10)
      In an equation for ‘main’: main = print (2 ^ 10)
  |
5 | main = print (2 ^ 10)
  |               ^^^^^^

because "what 10???"

$ stack ghci
λ> :t 10
10 :: Num p => p