pangram 1.4.1.10: Add test case: determine pangram lazily

[sshine]

Fail if the input stream was touched beyond the first occurrence of each letter.

Many standard solutions achieve this property without much effort, and some don't.

This test case notifies the student that a more optimal solution is within reach.

It is commented out by default:

While success-standard passes this test, and a String-based
fail-too-eager example fails as expected, success-text cannot easily
succeed, even if is changed into using Data.Text.Lazy:

One can write a Data.Text.Lazy as LT-specific test:

describe "isPangram" $ do
  it "is lazy" $
    isPangram (LT.fromChunks ["a...z", undefined]) `shouldBe` True

But when the input comes via fromString :: IsString s => String -> s,
the implementation has little control of the chunking. A failed attempt
to circumvent this,

newtype ChunkyLazyText = ChunkyLazyText LT.Text

instance IsString ChunkyLazyText where
  fromString = ChunkyLazyText . LT.fromChunks . map (T.pack . return)

still causes the T.pack invocation on [undefined].

And as @bobdalgleish pointed out in this StackOverflow comment,
I'd actually end up testing fromString. Since "laziness" doesn't mean
the same thing for String as it does for Data.Text.Lazy because
chunking works differently, the reliable alternative is to write tests
for each implementation.

Or, in this case, let string-type specific tests remain commented out.

[petertseng]

Interesting observations about Text-based implementations.

[sshine]

Yes. It's a property of very granulated strings. The main reason to discourage lazy Text appears to have to do with I/O.

Also it's a rabbit hole.