Michael Abbott

> abbreviate the display of long parameterized type names What's a good heuristic here? Within the text which is currently light grey, it could (1) just cap the length at...

Maybe this should be a separate issue from the other sense of abbreviation? Anyway the option to truncate it just by length not by nesting is dead-simple, since there is...

Interesting discussion. Some partial thoughts: * How much could be done in Flux not Zygote? We could change it to always accept a 0D array where it now accepts only...

Thanks for this hint, @piechologist This opens a new terminal in the Bottom Dock, if one is not present. The same as View > Terminal Panel, obviously enough since ```cmd-c...

I think you want this: ```julia julia> wrapdims(df, :data, :x, :y, default=missing) 2-dimensional KeyedArray(NamedDimsArray(...)) with keys: ↓ x ∈ 2-element Vector{Int64} → y ∈ 2-element Vector{Int64} And data, 2×2 Matrix{Union{Missing,...

At present there is no website, but docstrings are accessible at the REPL prompt by pressing `?`: ``` help?> wrapdims(df, :data, :x, :y ) wrapdims(table, value, names...; default=undef, sort=false, force=false)...

The original issue was specifically about `cat` & block-diagonal matrices (and keyword arguments, and blockwise linear algebra). The last two posts are about `hvcat` which is what's used for block-matrix...

What are the types in your second example? One quick solution is to simply make the slow path an error, which https://github.com/FluxML/NNlib.jl/pull/612 allows. If you accidentally have Float64 showing up,...

Ideally someone extends the rule. But until then, can you reshape before & after to fit the single dims case? Or, less efficiently, make a comprehension which just calls view.

Besides inference failures, the one failing test is this: ```julia julia> gradient([2 3; 4 5]) do xs sum([x ^ 2 + y for x in xs, y in xs]) end...