Add sameSmallArray and sameSmallMutableArray

[sjakobi]

Resolves #403.

[sjakobi]

The windows-latest and macos-latest CI jobs fail because primitive's bounds aren't compatible with GHC 9.14 yet. #440 seems to address this.

[chessai]

Can the Eq instance for SmallArray not be similarly cleaned up?

[sjakobi]

As opposed to the instance for SmallMutableArray, the instance for SmallArray can return True, when the arrays are not in the same memory location, but have same the elements. We can use sameSmallArray to enable short-circuiting though! :)

[chessai]

Yeah, short-circuiting was the desired cleanup. Thank you!

[Shimuuar]

There's problem with floating point and lawless Eq instances. With sameSmallArray shortcut
let x=[nan] in x==x returns True but let x=[nan]; y=[nan] in x==y returns False.

[treeowl]

There's problem with floating point and lawless Eq instances. With sameSmallArray shortcut let x=[nan] in x==x returns True but let x=[nan]; y=[nan] in x==y returns False.

Floating point is indeed a problem... Did @sjakobi do some recent pointer equality stuff in unordered-containers, or did @meooow25 do it in containers?

[sjakobi]

Hmm. This is unfortunate, but I think it's acceptable that broken Eq instances don't interact well with the Eq instances of container types. Users can still resort to liftEq (==) if necessary.

[sjakobi]

Did @sjakobi do some recent pointer equality stuff in unordered-containers,

I did some, but I don't think it affects the Eq instances for HashMap or HashSet.

[treeowl]

Did @sjakobi do some recent pointer equality stuff in unordered-containers,

I did some, but I don't think it affects the Eq instances for HashMap or HashSet.

Oh right, some bulk operation stuff. Regardless, we might want to document that floating point NaNs are not suitable for use as keys, If we haven't already.