I come up across this issue
https://github.com/JuliaData/DataFrames.jl/issues/2440
and I was amazed by the idea and the capability of Julia language. I learnt a lot from the code, but wonder why it is not implemented yet?
The reason is that it would complicate the minilanguage (it is easy to keep adding things but at some point they could become so complex that newcomers would feel overwhelmed by them).
We first do https://github.com/JuliaData/DataFrames.jl/pull/2869, which implements fast reductions without having to add new semantics with the limitation that they will work only for most common cases at the benefit that they will be faster than they would be with AsVector. It will be released soon.
Also the mechanics is that the user will be able to add their own fast reductions by adding methods to table_transformation function if needed.
Do you have a need for AsVector in cases of non standard reductions? (by standard I mean things like sum, mean, min, max, etc.) that could not be easily handled by adding a custom method for table_transformation?.
3 Likes
cool!
Nothing important, for each row Iwas looking for the most frequent number.