Decide whether to use DataFrames as data structures or actual structs

[abelsiqueira]

Decide whether to use DataFrames as data structures or actual structs (using MetaGraphs.jl)

Maybe it is not necessary to use a separate dedicated struct to hold the data for the assets and edges since we are already reading these things from CSV files (probably into DataFrames). That means that they don't need a special structure, and therefore using the DataFrames directly should be viable.

Depends on:

• Describe every CSV that is required #102

[clizbe]

We have several issues about dataframes, structs, namedtuples, dicts.
#118 #105 #36 #16

According to this random person on the internet, dataframes aren't meant for data with named rows, which it seems like we have.
https://discourse.julialang.org/t/when-should-i-choose-a-struct-mutable-struct-dict-named-tuple-or-dataframe/60366/4
Is that what you meant about problems retracing violated constraints?

Should we be using a different data storage type for our parameters and variables? Maybe immutable and mutable structs? Although the response above also sounds like tuples could be an option. I don't know what counts as a "smallish amount of objects."

[abelsiqueira]

These issues are a bit related to https://jump.dev/JuMP.jl/stable/tutorials/getting_started/design_patterns_for_larger_models/ and the subsequent chapter on performance.

The question for all these issues is where we'll stand in the balance between performance and readability (code readability and log readability).

My short answer is to do what looks nicer but not obviously slow, and then benchmark/profile later.
Before we can benchmark, we need a working model from beginning to end with at least the Nordic case, and possibly the solver that we will, in fact, use.

[clizbe]

Blocked until Benchmark #111 #146

[datejada]

waiting on #110