Will C# benefit from Computation Expressions?

[voroninp]

F# has a very nice feature which makes code look nicer compared to what it would be if methods were used directly.

C# linq syntax with from and let is similar to CE, but it's far less expressive.

[svick]

The main uses of computation expressions in F# are addressed by LINQ, async-await and yield return in C#. In other words, where F# has a single flexible system for certain kinds of mini-DSLs, C# has dedicated syntax for each of the primary use cases.

This means if there is some niche use case for F# computation expressions, C# very likely isn't going to have a great syntax for it.

If there is some common use case that's missing, the best option is probably to expand one of the existing syntaxes.

Do you have a specific use in mind?

[svick]

C# already has a dedicated syntax for that: try-catch. I get that there are advantages to Result over exceptions, but it's not a common approach in C# and I don't think C# is going to introduce features specifically to make a different error-handling strategy more convenient.

Though C# is probably going to add discriminated unions in the future (see #113 and #3179), which could make working with Result nicer.

[Joe4evr]

Side-note: It would be useful if, with the support of DUs, there comes a BCL "blessed" version of Result<T, TError> just so that each library doesn't have to define their own and there's good interoperability.

[voroninp]

@svick true, it's not common, but this can also be a survivor bias. It's not convenient to use alternative approaches, so it's not common.

[HaloFour]

I've brought up this conversation on the Discord server before, specifically referencing Haskell do notation and Scala for comprehensions, and there didn't seem to be a lot of interest in such functionality for composing monads in this manner. I'd be curious to see if DUs have an impact on that opinion, especially if the BCL gets a "blessed" Result<T, TError> monad. I could see some language features to bridge the current world of exceptions, but I kinda doubt they'd do something this general purpose.

[Richiban]

some common use case that's missing

I'd argue that the common use case is the ability to write your own. I think C# could invent new functionality that would combine F#'s computation expressions with Kotlin's builders; it would be awesome but probably quite niche.