Alexander Plavin

Just my 2c: I would definitely expect `down` to be just like `up`, but in the opposite direction. That is, prefer oldest possible versions, but otherwise always resolve the env...

> must be repeated every time one of the compat entries has to be changed Not just that: a new version of a dependency can restrict versions of its dependencies...

Yeah, especially if `add --weak SomePackage` would not only add `SomePackage = ...` to weakdeps, but also `SomePackageExt = SomePackage` to `[extensions]` and `SomePackageExt.jl` to `ext/` with the empty module...

Basically, this PR makes it possible to run `enclose` for functions with very different scales (think `f(x)` and `1e10 * f(x)`) with the same options: just specify `rtol`! Currently, `tol`...

Wonder if there are any plans for adding multidimensional inrange search? Implementation is very short: https://github.com/JuliaAPlavin/FlexiJoins.jl/blob/00000000ca2fe5eace0b881caa8d2258ca6154ee/src/nearestneighbors.jl#L32-L52 https://github.com/KristofferC/NearestNeighbors.jl/pull/150

Also, what about supporting arbitrary types, at least for BallTree? Stuff like strings or even custom objects can be useful with distance searches. https://github.com/JuliaAPlavin/FlexiJoins.jl/issues/8

I personally don't care much – happy using 0.6 now, just as happy to switch to 0.7/0.8 at some point in the future. Still find it a bit strange that...

> trying to keep performance high on every supported version of Julia is going to add a lot of burden to both the code writer and maintainers On every Julia...

Fixing zero-step ranges seems trivial, should be ready to go now.

I'm generally indifferent to which decision is made regarding this PR (eg, release "breaking" version, or "non-breaking" one). I find it highly improbable that anyone already relies on features from...