Pangram: Add non-basic characters in the test suite

[sir4ur0n]

I see many students filtering using Char.isLetter , then removing duplicates, and then checking the size is 26.

I think this would be nice to add other characters (e.g. '字') which are letters (Char.isLetter '字' == True) in the test suite to make them realize this is not a good enough function for pangram.

Any opinion? If you agree I can submit a PR.

[petertseng]

Although adding such a case would be in contravention of the Exercism-wide decision to only have ASCII unless the exercise is specifically about dealing with multi-byte encodings ( exercism/problem-specifications#428), individual tracks are allowed to contravene Exercism-wide decisions if those involved (maintainers, proposer, others contributors, etc.) decide it is a good idea.

Since I don't have a specific recommendation in either direction, I leave it to the interested parties to determine if it is a good idea.

[sir4ur0n]

I wasn't aware of this ASCII restriction, thanks for pointing it out, I guess I will stop using the example of Chinese letters then 😅

[sshine]

I think this test case is very good. Since Exercism has a policy against non-ASCII canonical tests, I think we should create a Haskell-specific test case. We may want to use a letter that people are not in doubt about being a letter, e.g. 'Ã' and not e.g. 'ˀ' (glottal stop). :-P

[sshine]

The class of errors you describe are false positives.

But our test suite should also describe false negatives.

1. False positives:

   isPangram :: String -> Bool
   isPangram = (== 26) . length . nub . map toLower . filter isLetter

   succeeds for isPangram (['a'..'y'] ++ "æ") when it should fail.

2. False negatives:

   isPangram :: String -> Bool
   isPangram = (== ['a'..'z']) . sort . nub . map toLower . filter isAlpha

   fails for isPangram (['a'..'z'] ++ "æ") when it should succeed.

Eventually, once exercism/problem-specifications#1518 is resolved, we may be interested in discussing in the problem-specifications repo whether to change the ASCII policy so that Unicode tests can be allowed, but are always listed as optional because they're Unicode.

In the meantime I would prefer that these two tests are added in the Haskell track.

@sir4ur0n: Would you be interested in submitting a PR that adds these two to the Haskell track?