Provided implementation for Set.exactlyOne and Set.tryExactlyOne

[yreynhout]

I only had the following insight after submitting this PR - I probably can't just change / extend the api surface of FSharp.Core. Still, it felt like this was missing. Some guidance on how to proceed would be appreciated.

[dnfadmin]

All CLA requirements met.

[abelbraaksma]

I assume you're adding this for parity with lists, sequences etc, right? Or is there a language suggestion? Personally, I think it's good to have it (parity, I mean).

I assume the reason tests fail is because if the surface area (the public functions) changes, they will fail. It's to protect against accidental additions of public functions, that would otherwise go unchecked.

[cartermp]

@yreynhout you can see the test failures here: https://dev.azure.com/dnceng/public/_build/results?buildId=712967&view=ms.vss-test-web.build-test-results-tab&runId=22108046&resultId=101213&paneView=debug

[yreynhout]

@abelbraaksma indeed, felt a bit weird that I had to either write custom functions or convert to a list to get the functionality.

[yreynhout]

@cartermp Thanks for the hint. Are you suggesting I simply change the surface area in the test file(s)?

[cartermp]

The surface area changes are necessary to make this mergeable. But I think that some new tests for the Set module are also in order here.

[abelbraaksma]

FYI, the related suggestion for Seq and friends was: fsharp/fslang-suggestions#137. It should perhaps be paired with the suggestion to extend the surface area of Set: fsharp/fslang-suggestions#803.

I don't know if it is needed, but if so, I can help write up a short RFC for this. Or we could perhaps extend this PR to the other functions mentioned in fsharp/fslang-suggestions#803.

[yreynhout]

@cartermp I'm more than willing to add tests but failed to find a suitable place. Any pointers as to where you'd like me to add them would be useful - I'm a newbie in this codebase. I do hope I can confine the tests to cover only what I'm adding.

@abelbraaksma If there is protocol to be followed I'll take any suggestions on how to proceed. I can copy the seq related RFC for tryExactlyOne and work my way from there.

[abelbraaksma]

Sounds good, I'm not sure what the protocol is here, but my guess would be that changes to the API need an RFC, so that people are aware it exists, and/or can comment on the design.

[KevinRansom]

@yreynhout , there is supposed to be a test that fails when you change the surface area of FSharp.Core. But yes, you can add api's to FSharp.Core, we just have to figure out if we want to make a change.

There is a Documented Change Request process here:https://github.com/fsharp/fslang-suggestions/issues

And here is an example: fsharp/fslang-suggestions#873

[KevinRansom]

@yreynhout ,
I am closing this PR, because the API has not yet been approved in principle at: https://github.com/fsharp/fslang-suggestions/issues

Please pursue the RFC process mentioned and above and then reactivate this PR for consideration once the API additions have been approved in principle.

Thanks

Kevin

[abelbraaksma]

This could be part of the Set module expansion issue: fsharp/fslang-suggestions#1361. Linking for posterity ;).