@pdeffebach: I still think that a version with
leftjoinand DataFramesMeta.jl’s@withis going to be the simplest and most effective.
See my comment above, for ~100k rows in each data frame, my VSCode crashed. Running on terminal gave me segfault.
It’s hard to tell exactly without a full MWE
Is the code I put on the first post not working or do you mean a large case?
And once more, the context is another problem posted here: Speeding up JuMP model creation with sets that depend on other indexes
I currently don’t use anything close to a data frame. Instead, I have a dictionary to compute the indexes that I have to sum for each row of df_cons.
My expectations are that using a linear structure will be better (df_cons). However, computing the incoming column of the linear structure is proving to be slower than the dictionary approach.
I.e., JuMP is slower because I have the dictionaries instead of the linear indexes, but computing the linear indexes are slower than JuMP, so maybe it is what it is.