Create GCoercible class, analogous to GEq
#46
Conversation
Ericson2314
force-pushed
the
gcoercible
branch
2 times, most recently
from
8509f8a to
2915e7b
Compare
Ericson2314
changed the title
GCoercible class, analogous to GEq --- contains #45
Ericson2314
force-pushed
the
gcoercible
branch
2 times, most recently
from
46e2be8 to
6570a3a
Compare
It could be a superclass of `GEq`, but this would be a breaking change and interferes with Safe Haskell, so we refrained from making it one.
Ericson2314
force-pushed
the
gcoercible
branch
from
6570a3a to
dbafb11
Compare
Ericson2314
changed the title
GCoercible class, analogous to GEq --- contains #45GCoercible class, analogous to GEq
| gcoercible (InR x) (InR y) = gcoercible x y | ||
| gcoercible _ _ = Nothing | ||
|
|
||
| instance (GCoercible a, GCoercible b) => GCoercible (Product a b) where |
There was a problem hiding this comment.
This is weird (also in existing GEq instance. Either side would be enough, won't it?
There was a problem hiding this comment.
Huh? It is possible for one to fail and the other to succeed, is it not?
There was a problem hiding this comment.
For TestEquality yes only one is sufficient.
There was a problem hiding this comment.
Yes, I was thinking of TestEquality, and it doesn't seem to have instance for Product so all is fine.
(btw, that's another point not to have TestEquality as a super-class for GEq).
| return Coercion | ||
|
|
||
| instance GCoercible IORef where | ||
| gcoercible x y = |
There was a problem hiding this comment.
I'll be more welcome to add this and STRef instance if the implementation for this function were in base, and I could use Safe (or at least Trustworthy) interface. (This package takes Safe Haskell seriously for time being).
Turns out there is for heterogenous equality for StableName (eqStableName), so figuring out coercion based on STRef equality kind of belongs to base as well.
As maintainer I'll be happy if burden of tracking GHC internals is in GHC-provided lib.
There was a problem hiding this comment.
I just had the same thought. The unsafe function should be where the implementation is, as those must be kept in sync?
No description provided.