cannot insert a range that the start equal with the end

[dwgeneral]

in src/map.rs

/// Panics if range `start >= end`.
    pub fn insert(&mut self, range: Range<K>, value: V) {
        // We don't want to have to make empty ranges make sense;
        // they don't represent anything meaningful in this structure.
        assert!(range.start < range.end);

I think we should allow the range params that the start equal with the end, it's meaningful in real business scence,
would you mind I change the code for this?

[jeffparsons]

Hi @dwgeneral, thanks for getting in touch! 👋

I'm not fundamentally opposed to this change, but it would complicate some other things so I'd want to carefully understand the implications first.

E.g. what would it mean to add a zero-length range at the start of an existing range? Does the existing range's start get moved to the next possible value? I don't think all types you might want to put in the RangeMap even necessarily support that.

Maybe if you tell me more about your use case it would help me understand the best way to approach this.

[jeffparsons]

@dwgeneral I have a work-in-progress PR that introduces a type for what I think matches your use case: #15.

It works, but I have a few things I want to clean up before landing it, as you can see in the "TODO" section on that PR.

[jeffparsons]

@dwgeneral I've just released a new version that supports inclusive ranges: https://docs.rs/rangemap/0.1.6/rangemap/

I know this isn't quite what you were asking for, but do you think it might work for your use case?

I'm still considering allowing zero-length ranges in RangeMap, but it'll be a bit tricky to define an API that supports that but isn't horrible to use. E.g. I don't want to require that your key type supports successor/predecessor functions in the general case, because it's probably extra fiddling around that's not needed for most people, but that would be necessary when inserting zero-length ranges. Maybe just having a separate insert_possibly_zero_length function with the extra trait bounds (TODO: good name 😅) would be enough.

Does that make some kind of sense?

[jeffparsons]

By the way, I'd love to hear about details of your use case if you're able to share. 😃

[Kinrany]

Would it make sense to just ignore empty ranges instead of panicking? This seems equivalent to adding 0 to a number.

[Kinrany]

#112