[RFC 1082 Discussion] - uint type abbreviation for FSharp.Core

[cartermp]

This is the discussion issue for RFC 1082.

RFC: https://github.com/fsharp/fslang-design/blob/master/RFCs/FS-1082-uint-type-abbreviation.md
Suggestion: fsharp/fslang-suggestions#818

[abelbraaksma]

The RFC doesn't mention the impact on the cast function. As mentioned in the discussion, I think the uint() function should be introduced as an alias for uint32(). Otherwise it's going to be rather odd that you must write this:

let f x: uint = uint32 x

Instead of this:

let f x: uint = uint x

It would also be nice if you could pair the other remaining primitives similarly, but I understand if that should go into a separate suggestion.

[cartermp]

Thanks @abelbraaksma - updating. I don't think we'd do that for the other primitives here, that would be another suggestion.

[varon]

PLEASE NO! This is going to be incredibly confusing to have uint vs unit.

[Happypig375]

PLEASE YES! This is going to be incredibly helpful to write uint vs uint32.

[davidglassborow]

Would a type alias not currently work ?

[cartermp]

@davidglassborow it is implemented as a type alias + casting functions in the current design. I wouldn't expect that to differ since that's how int is done.

[abelbraaksma]

Would a type alias not currently work ?

That would alias the type, not the cast function.

[krauthaufen]

Hey, personally I'm not a huge fan of the many existing aliases (sbyte/int8/etc.) but when doing this how about short/ushort/long/ulong?

[Happypig375]

uint already exists as a concept today in F#: uint8, uint16, uint32 and uint64.
In contrast, short and long do not exist in F# yet.

[abelbraaksma]

I personally find short/int/long confusing, since historically, and even presently in C/C++, int can mean 32 or 64 bit, long can mean 32 and 64 bit too. Sometimes it depends on word size, and a word can be 16, 32 or 64 bits.

Just to say that I find clarity in the way F# chose the names it did. The only one to remember is (u)int, the rest go by their bit-size-suffix. Adding more variants doesn't add coding clarity (and I'm aware we have both sbyte and int8, float and double and others, I wish there weren't, but that's a fait accompli).

[cartermp]

It's worth noting that there may be some more numeric primitives down the road in .NET. This would probably require corresponding type abbreviations in FSharp.Core.

[abelbraaksma]

I thought I saw an issue about a request for some new primitives like unativeint and the like.

[abelbraaksma]

I think this can be closed? The implementation is complete: dotnet/fsharp#8185.

Note that a comment in the PR specifically says that this doesn't have y wait until F# 5.0, but that this thread is marked for the F# 5.0 milestone.

[cartermp]

@abelbraaksma i'll likely do a round of closing later this year.