[Haskell-cafe] Ensuring Type Class instances follow the 'rules'

[Ryan Ingram]

Just pretend it was named "Test Driven Development" so it feels more
sexy while you read it :)

  -- ryan

On Fri, May 29, 2009 at 5:05 AM, Hemanth Kapila <saihemanth at gmail.com> wrote:
> Thanks Eugene and wren.
> Serves me right, actually. The one chapter I skipped in RWH was the 11th one
> called "Testing and Quality Assurance" thinking it is too boring.
>
>
> On Fri, May 29, 2009 at 12:44 PM, wren ng thornton <wren at freegeek.org>
> wrote:
>>
>> Eugene Kirpichov wrote:
>>>
>>> Use QuickCheck.
>>
>> Also use SmallCheck or lazy SmallCheck.
>>
>> All of these are automatic tools that allow you to specify laws[1] and
>> automatically generate values for testing the law. QuickCheck generates
>> values randomly, which can be useful but is very often insufficient. Both
>> versions of SmallCheck do exhaustive search of all values down to a bounded
>> "depth" (lazy SmallCheck can often search deeper and faster).
>>
>> If you want to have decent assurances of correctness then use (lazy)
>> SmallCheck. If you're dealing with a really branchy type which makes it
>> prohibitive to search to any reasonable depth, then use QuickCheck in
>> addition. Occasionally QuickCheck can be good for ferreting out really
>> obscure corner cases, but SmallCheck is much better about guaranteeing you
>> get all the basic cases.
>>
>>
>> [1] e.g.
>>
>> prop_compare a b =
>>    compare a b == negateOrdering (compare b a)
>>    where
>>    negateOrdering LT = GT
>>    negateOrdering EQ = EQ
>>    negateOrdering GT = LT
>>
>> --
>> Live well,
>> ~wren
>> _______________________________________________
>> Haskell-Cafe mailing list
>> Haskell-Cafe at haskell.org
>> http://www.haskell.org/mailman/listinfo/haskell-cafe
>
>
> _______________________________________________
> Haskell-Cafe mailing list
> Haskell-Cafe at haskell.org
> http://www.haskell.org/mailman/listinfo/haskell-cafe
>
>