[Haskell-cafe] Alternative to newtypes or orphan instances

[Miguel Mitrofanov]

OK. There is no point in having a class C, if there is no functions to work with it. So, this "other package" likely contains some functions that take types (of that class) as input. Like this:

foo :: C a => a -> Int

You don't need class C if you aren't using any functions from the library. So, with your (c), if you try to do

a :: T
a = ...

b :: Int
b = foo a

you'll get a compile error. THAT is the problem.

24.08.2015, 13:28, "Clinton Mead" <clintonmead at gmail.com>:
> There's no new data. Class C already exists in another package. Data T already exists in another different package.
>
> The options are:
>
> a) Create a newtype MyT and an instance MyT of C.
> b) Just create an orphan instance T of C.
> c) Create a new class C1, forward its implementation to C, and add an instance T of C1.
>
> I'm suggesting (c) is best, but I haven't seen this elsewhere, the debate is usually between (a) and (b).
>
> I don't really understand the problems you're proposing with (c), but I'm not sure if that's because I'm misunderstanding you or I'm not explaining myself well.
>
> On Monday, 24 August 2015, Miguel Mitrofanov <miguelimo38 at yandex.ru> wrote:
>> If, for your data, you create an instance of the new class — but not the original one, then you can't use all the machinery that expects input being of the old class. Which is the point.
>>
>> 24.08.2015, 12:55, "Clinton Mead" <clintonmead at gmail.com>:
>>> The original class still exists, I can't see how making a new class based on the old on affects that. Won't existing functions in modules which import the old class instead of the new class continue to work?
>>>
>>> On Monday, 24 August 2015, Erik Hesselink <hesselink at gmail.com> wrote:
>>>> On 24 August 2015 at 09:18, Clinton Mead <clintonmead at gmail.com> wrote:
>>>>> A second approach is an orphan instance. The recommendation here is to put
>>>>> the orphan instances in their own module, so the user can choose to import
>>>>> them.
>>>>>
>>>>> This may works ok if your user is writing an executable. But what if your
>>>>> user is writing a library themselves. But once, you, or your user, directly
>>>>> uses one of the instances, they need to import it, and they pollute the
>>>>> global instance namespace for anyone that uses their package.
>>>>
>>>> For this reason, I think the recommended course of action is to make a
>>>> canonical place for the instance, so that everyone can use it. For
>>>> example, if you have a library 'foo' providing T, and a library 'bar'
>>>> providing C, put the instance in a new package 'foo-bar' (or
>>>> 'bar-foo'). Then everyone can use that one instance, since Haskell is
>>>> built on the assumption that every type has one unique instance per
>>>> class.
>>>>
>>>>> I want to suggest a third option:
>>>>>
>>>>> (3) Copying the class.
>>>>
>>>> This would make a new distinct class, which means you can't call any
>>>> methods which have the original class as the context (f :: C a => a ->
>>>> a) since that class won't exist for type T (you are trying to avoid
>>>> defining that orphan instance). So I don't think this is usable in
>>>> most cases, unless I'm missing something.
>>>>
>>>> Erik
>>> ,
>>>
>>> _______________________________________________
>>> Haskell-Cafe mailing list
>>> Haskell-Cafe at haskell.org
>>> http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe