Daniel Karrasch

It's hard for me to estimate the impact of this proposal without seeing a draft. In some sense, I think the proposal is consistent with @nalimilan's pairwise implementation in `StatsBase.jl`....

IIRC, the 3-arg dot is slower than using BLAS twice (which I just confirmed with a quick benchmark for `Symmetric` matrices). Its main purpose is to avoid intermediate allocations. Or...

Oh, I was benchmarking large matrices (dimension several hundreds). I wasn't sure what the typical use case for `Mahalanobis` is.

But `colwise` and `pairwise` have very different meanings here, independent of whether data points are viewed as columns or rows. `colwise` is meant to compute distances only between corresponding data...

I guess we should keep `colwise(dist, x, y)`, make `map(d, eachcol(x), eachcol(y))` the default, but allow specialized methods to optimize for performance. ```julia using Distances, BenchmarkTools d = Euclidean(); a...

I have a `metric` type which is much costlier than the usual `Euclidean` calculations, that's why I considered doing it in a distributed fashion, even more so since generally the...

Do you mean something like this? ```julia julia> A = reshape(collect(1:9), 3, 3) 3×3 Array{Int64,2}: 1 4 7 2 5 8 3 6 9 julia> inds1 = findall(iseven, A) 4-element...

@nalimilan convinced me that supporting `CartesianIndex`es directly (at least in the context of #194) is awkward, because it targets iterable objects, and `CartesianIndex`es are intentionally non-iterable. I believe the correct...

Shall we finish this, release as v0.10.x, and subsequently merge #233 and release as v0.11.0? Potentially removing the deprecation warnings right away? Package maintainers will need to bump their `Project.toml`...

I would vote for keeping the `deprecated.jl` file, but remove the deprecations, and add a comment that this should be deprecated in the v0.11 release and removed in the v0.12...