Thank you Theodore. Yes, changing the variable names makes it clearer. Also the fact that only failures are combined on the right hand side... Regards, Andrea Sent with [ProtonMail](https://protonmail.com) Secure Email. -------- Original Message -------- Subject: Re: [Haskell-beginners] Semigroup Instances Local Time: February 6, 2017 5:56 PM UTC Time: February 6, 2017 11:56 PM From: tanuki@gmail.com To: Atrudyjane <atrudyjane@protonmail.com>, The Haskell-Beginners Mailing List - Discussion of primarily beginner-level topics related to Haskell <beginners@haskell.org> Gmail put you in spam. If you haven't figured this out since you asked -- it's a matter of confusing (IMO bad) variable names. Check the data definition: data Validation a b = Failure a | Success b deriving (Eq, Show) Failures are always type a, and successes are always type b. The type variables used in the first line correspond to these. But in the definitions of (<>), they are just local values. The instance could be rewritten like so: instance Semigroup a => Semigroup (Validation a b) where Success x <> Success y = Success x Failure x <> Success y = Success y Success x <> Failure y = Success x Failure x <> Failure y = Failure (x <> y) On Thu, Jan 26, 2017 at 1:55 PM, Atrudyjane <atrudyjane@protonmail.com> wrote: I'm currently studying semigroups and trying to figure out how to determine which type variables need a semigroup instance. Here are a couple of examples from Evan Cameron's github (https://github.com/leshow/haskell-programming-book/blob/master/src/Ch15ex.hs): (1) data Validation a b = Failure a | Success b deriving (Eq, Show) instance Semigroup a => Semigroup (Validation a b) where Success a <> Success b = Success a Failure a <> Success b = Success b Success a <> Failure b = Success a Failure a <> Failure b = Failure (a <> b) * Why doesn't 'b' need an instance of semigroup? (2) newtype AccumulateRight a b = AccumulateRight (Validation a b) deriving (Eq, Show) instance Semigroup b => Semigroup (AccumulateRight a b) where AccumulateRight (Success a) <>AccumulateRight (Failure b) =AccumulateRight (Success a) AccumulateRight (Failure a) <>AccumulateRight (Success b) =AccumulateRight (Success b) AccumulateRight (Failure a) <>AccumulateRight (Failure b) =AccumulateRight (Failure a) AccumulateRight (Success a) <> AccumulateRight (Success b) = AccumulateRight (Success (a <> b)) * Why doesn't 'a' need an instance of semigroup? Thank you, Andrea Sent with [ProtonMail](https://protonmail.com) Secure Email. _______________________________________________ Beginners mailing list Beginners@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/beginners