What are additional LetPattern and MatchPattern cases that'd be useful?

[StachuDotNet]

Useful links:

• https://fsharp.github.io/fsharp-compiler-docs/reference/fsharp-compiler-syntax-synpat.html
• https://learn.microsoft.com/en-us/dotnet/fsharp/language-reference/pattern-matching#remarks

[StachuDotNet]

should this ever be legal Dark code?

match [1; 2; 3] with
| first :: second :: [] -> "exactly 2 elements"
| _ -> "not 2 elements"

[pbiggar]

I'd probably think that something based off elipses is better (like rescript does for records). Something like:

match [1,2,3]
| [first, second, ...] -> two or more
| [first, second] -> exactly two
| etc

[StachuDotNet]

should this ever be legal Dark code? (relevant to #4711 right now)
let (a, b, (c, d)) = (1, 2, (3, 4))

[StachuDotNet]

For what it's worth, I think we could support it in the backend, and limit the 'depth' based on something other than the model of the language. (and in the meantime, possibly restrict it altogether in the editor)
For example, an individual canvas could disallow a pattern like this based on stylistic preferences.

[pbiggar]

I think this is fine, not even sure why we'd restrict it (sometimes you have data in that shape!)

[StachuDotNet]

Should this be legal? ("constructor"/enum let pattern)

let opt = Just 1
let (Just a) = opt
a

[pbiggar]

No.

I think in general the following shoudl be legal, but there's issues. Suppose we have the type: type clientID = ClientID of int

Then it makes sense to have: let (ClientID id) = x

But then if you expand the type to type clientID = ClientID of int | NewClientID of int64

Then deconstructing a single variant/constructor becomes a type error suddenly, which is bad.

So I'd propose we don't allow deconstruction of constructors at all right now.