Anthony Blaom, PhD

Yes. Once you have decided the natural choice of type for `X` in your `fit`, you ensure `coerce` maps provided iterable table to your type. The existing type declarations are...

**Important design question** It would be nice to avoid unnecessary data "coercions" (usually table -> matrix) in meta-modelling. The two main cases are ensembling and resampling by cross-validation. The raison...

The external package algorithm being wrapped may *require* a matrix for its fit method (and a matrix may indeed be the most efficient form for it to work with). Remember,...

Hmm. I think I understand your objections, and do not at all disagree. However, I do not understand the proposed solution: > @ablaom, I see - I still personally think...

@nalimilan Is there not still the problem that Tables.jl does not provide fast random access methods, even if the type supports it, because only iterators are returned? Or is this...

@nalimilan Thanks for that. I did not realise that Tables.columnstable essentially returns a view and not a copy where possible. This makes a big difference in those cases. So, if...

@fkiraly On reflection I am in complete agreement with your objections to my "sophisticated' attempts to circumvent repetitive data coercions. All row/column access to data should be applied through the...

I'm encountering strange behaviour when generating a markdown file within the same script. So this works fine: ```julia # Literate.markdown(@__FILE__, DIR_OUTPUT, execute=false) #src Literate.notebook(@__FILE__, DIR_OUTPUT, execute=true) #src Literate.notebook(@__FILE__, DIR_OUTPUT, execute=false,...

Do you mean, am I calling the conversion from within the script I want converted? Yes. It's a convenient workflow. Should I not do this?

Okay, recent changes seem to have broken something. Now I can't get pluto converter to work at all. Here's a MWE ```julia using Pkg Pkg.activate(temp=true) Pkg.add(url="https://github.com/fredrikekre/Literate.jl", rev="fe/pluto") #- x =...