Refactor rdiffnet

[gvegayon]

Example of flowchart using mermaid (You don't need to use this, draw.io also has this)

flowchart LR
  is_seedpnodes_list{Is seed p nodes list} -->|Yes| B
  is_seedpnodes_list -->|No| C

Loading

The ideal function should have:

• Allow passing a list for some of the arguments, in which case it assumes a multi-behavior model.
• Arguments are are not passed as list, should be recycled, e.g., if seed.p.nodes is a list but seed.nodes as a vector (indicating seed vertices), then the program should do something like seed.nodes <- list(seed.nodes, seed.nodes)

Need to think about the sim process.

For the preprocessing, you could have something like this:

rdiffnet_validate_args <- function(seed.p.nodes, seed.nodes, behavior) {
   ... do all the checks to validate length and stuff ...
   list(
     seed.p.nodes = ...,
     seed.nodes = ...,
     behavior = ...,
     num_of_behaviors = ...
   )
}

rdiffnet <- function( ... ) {
  
    validated_args <- rdiffnet_validate_args(seed.p.nodes, seed.nodes, behavior)

    ....

    # if you want to use seed.p.nodes
    validated_args$seed.p.nodes
}

[gvegayon]

Here is an idea as a generalization.

n <- 20
thr <- .3
thr <- rep(.3, n)

thr <- function(x) runif(n)

thr_final <- if (length(thr) == 1) function() rep(thr, n)
else if (length(thr) == n) function() thr
else if (is.function(thr)) function() thr()

# IN the simulation, replace
# thr < exposure
# with
# thr() < exposure
# So, this handles all the cases.

[aoliveram]

I wondered if the function MUST return another function or if it's enough to always to return a 'matrix' 'array' object. This is more straightforward as we can keep many of the tests already written.

I have implemented the function that always returns 'matrix' 'array' objects. I could change it, but I wonder if it's strongly recommended.

[aoliveram]

Hi @gvegayon! LMK what do you think.

[gvegayon]

Adding a reference to the equations: #46 (comment)