[Haskell-beginners] How to best handle classes of types, interactions and listing of different types?
Silent Leaf
silent.leaf0 at gmail.com
Mon May 23 16:06:52 UTC 2016
Hi! I'm still studying the theory of Haskell, but i also wanted to start
dabbling with guideless coding, all by myself, and it happens experience is
just as much important as theoretical knowledge! I'm not surprise mind you
:P
Say there's a class, many types that can instantiate it. Then, i in fact
need to be able to concatenate (mappend would be ideal!), make lists of
values of different types, all instantiating the class.
I'm a bit lost on this. There are at least two sorts of ways to handle said
group of types, but i'm not certain either is best, and i'm not even very
much satisfied about either. Plus they don't specifically differ as regards
handling of listing or otherwise interacting *between* different instances
of the same class (or alternately, different types born from the same Type
constructor).
First idea i had, was based on making a class indeed, as well i just love
this type of polymorphism (visavis other paradigms). Each type would only
have to instantiate the key metshods, and no need of wrappers to use said
methods, so on this respect it's pretty great.
However, as soon as we want operations between values of different types,
it rather gets tricky.
The only way i know of to make a heterogeneous list is to use
ExistentialQuantificators, create a newtype (or data?) wrapper for the
class, allowing for any type to be in, without parameters.
Now, the problem i have with this, is the wrapping: if i wanna make a list
of types, i gotta wrap all values, and if this list becomes member of a
superlist of the same polymorphic type, i gotta wrap this very sublist
again with the "class wrapper". Maybe i'm imagining problems and i won't
actually need to touch too much to the wrapping if i write the correct
functions or instances, but from where i am, permanent wrapping seems
tiresome, very much in fact removes the purpose of a class: to allow
polymorphic functions without need to wrap the types of inputs/outputs
inside a newtype.
Which lead me to the second solution: forget the class, and make a type
constructor that basically gets a list of key-methods and one field for the
wrapped type. Basically it amounts to OOP, at least it seems so. I'm not
really a big fan of the thing, and eventually i realized it wouldn't help
me more visavis trans-type methods, especially an idea of one or more
monoidal instances for the class, or the type constructor that replaces it.
(Actually I'm not wholly certain as of how I'm supposed to handle Type
constructors instead of classes, as I only know how to make heterogeneous
out of a class, not out of a type constructor, even if its polymorphic
key-methods are fields included in it. (hope i'm being clear enough, don't
hesitate telling me otherwise!)
So, as I'm kinda beginner as regards actual coding, i'm asking advice on
the subject. Maybe the problem is in how i look at the problem, and I
shouldn't even be there trying to do what i try to do. Any alternative
solution or viewpoint is heartily welcome!
I resume my problem once more: how to handle cross-type interactions and
grouping when they only have a class in common, and if possible without
having to use ExistentialQuantificators, unless there's a way to keep the
wrappers far away, invisible, automatic.
Thanks a lot in advance! (Ideas of specialized tutorials are also very much
welcome!)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.haskell.org/pipermail/beginners/attachments/20160523/a743d063/attachment.html>
More information about the Beginners
mailing list