Ternary comparison expressions 0 < x < 5

[JohanLarsson]

Comparing a value with a lower and upper bound is common.
Writing x > 0 && x < 5 includes duplication of the x identifier and hence a way to get it wrong.

[canton7]

Just to note, you can write:

x is (>0 and <5)

[Xyncgas]

doesn't work with non constants

[HaloFour]

dotnet/roslyn#136 (comment)

[DavidArno]

@CyrusNajmabadi,

Going to disagree with you on that one. I think x is > y and < z is highly intuitive as it uses existing pattern rules, just with variables rather than constants.

@alrz

Sorry but I really dislike that syntax.

[HaloFour]

@DavidArno

as it uses existing pattern rules

I'd imagine that a significantly smaller segment of the whole C# population is familiar with patterns, let alone relational patterns and their rules.

I think the ability to chain the comparison operators as suggested makes sense and reads nicely. It is technically ambiguous but only in a very niche case that could remain supported without impacting the implementation.

I also think it's worthwhile opening patterns to non-constant expressions, but I think that's a much bigger conversation in general.

[alrz]

I'd imagine that a significantly smaller segment of the whole C# population is familiar with patterns,

Isn't this also a new feature? So the same barrier applies only that patterns are actually a few releases ahead in terms of reach.

[HaloFour]

@alrz

Isn't this also a new feature?

It's not just a matter of which version of the language that you're on but also which features you're more likely to use. I'd suggest that nearly all C# developers have used comparison operators where a significantly smaller percentage have used patterns in general, and a much smaller percentage have used relational operators.

[DavidArno]

@HaloFour,

You are likely right that only a small segment of the C# community have used relational patterns. My thoughts though were along these lines:

The team add the ability to express a range comparison as y < x < z
The team also add support for variables in patterns, so it also becomes possible to express it as x is > y and < z as a side effect

This will then lead to the usual chorus of "why are there two ways of doing the same thing?" So I'd suggest it best to ask the LDM the question "will you ever support variables in patterns?" If no, then the case for supporting y < x < z becomes strong. But if the answer is yes, then why create a second way of doing the same thing?

Of course I appreciate the answer will likely be "maybe" and this whole argument then falls down :)

[CyrusNajmabadi]

Now championed at #4108. Will bring to a future triage to determine how LDM feels about this.

[bernd5]

For me that syntax looks strange - I would prefer to extend relational patterns to work with non - constants.
It looks like you would compare a boolean to something else.