#7775 closed feature request (wontfix)
Mark intentionally omitted type class instances
Reported by: | Lemming | Owned by: | |
---|---|---|---|
Priority: | normal | Milestone: | |
Component: | Compiler (Type checker) | Version: | 7.7 |
Keywords: | instance warning | Cc: | |
Operating System: | Unknown/Multiple | Architecture: | Unknown/Multiple |
Type of failure: | None/Unknown | Test Case: | |
Blocked By: | Blocking: | ||
Related Tickets: | Differential Rev(s): | ||
Wiki Page: |
Description
In GHCi I get
Prelude> 2+3::Char <interactive>:2:2: No instance for (Num Char) arising from a use of `+' Possible fix: add an instance declaration for (Num Char) In the expression: 2 + 3 :: Char In an equation for `it': it = 2 + 3 :: Char
In many cases users now actually try to implement the instance Num Char, but this is certainly not a good idea. I would like to tell the user whenever an instance is intentionally unimplemented and why it is missing. Like so
{-# NOINSTANCE Num Char where "There is no sound arithmetic on characters." #-}
The syntax after NOINSTANCE should be like in an instance declaration. It shall be treated like a type class instance, e.g. it shall be checked whether the same instance already exists, or whether it is orphan, or whether it is not Haskell 98, and it should require to enable according extensions like MultiParamTypeClasses. If on resolution a non-instance is chosen then the compiler should print the custom warning "There is no sound arithmetic on characters." instead of "probable fix: add instance".
Change History (4)
comment:1 Changed 7 years ago by
comment:2 Changed 7 years ago by
difficulty: | → Unknown |
---|---|
Resolution: | → wontfix |
Status: | new → closed |
Thanks for the suggestions. However, I think that now that the misleading advice has been removed, it's not worth implementing the pragma, so I'm closing this ticket.
comment:3 Changed 6 years ago by
Version: | 7.6.2 → 7.7 |
---|
There is still a difference between intentionally and unintentionally omitted instances. I find it important to document for me and for others why I have omitted an instance. Also without the "possible fix" message people keep asking why an obvious instance is missing. "Why is this parser not Monoid?" - maybe because there are different ways of viewing it as Monoid. "Why is this identifier type not Num?" - maybe because it makes no sense to multiply identifiers. Or even worse, people do not just ask but start writing (conflicting) orphan instances.
The "possible fix" message was already removed in 7.6 (#7222).