mrc-2432 Implement optimiser to strategize across regions

[mwoodbri]

This refactors Dirk's code as a function to be called by the optimise endpoint (which will be implemented in mrc-2433)

Any feedback whatsoever on my novice R code very welcome. Biggest challenge was getting dplyr to pass make check - ideas for tidying this up without making it unreadable gratefully received. We inherited dplyr and could probably do without it but would take me a while to translate into pure R...

[codecov]

Codecov Report

Merging #57 (ec516b4) into master (9369876) will not change coverage.
The diff coverage is 100.00%.

@@            Coverage Diff            @@
##            master       #57   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            5         6    +1     
  Lines          375       409   +34     
=========================================
+ Hits           375       409   +34     

Impacted Files	Coverage Δ
R/optimise.R	100.00% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 9369876...ec516b4. Read the comment docs.

[mwoodbri]

This is a dependency of dplyr but seemingly needs to be specified explicitly here for make check

[richfitz]

I usually put those things in Suggests (and typically order these lists alphabetically as they're easier to maintain that way)

[mwoodbri]

Tried to avoid this using .data$ but it got very verbose, and I still had trouble eliminating all the warnings

[richfitz]

the other thing you can do is to scope these variables within the function so do:

function(...) {
  zone <- intervention <- NULL # used by dplyr
  df %>% select(zone) # etc
}

There's no great way of doing this though. The .data$ trick is the canonical way but it's not nice to read

[mwoodbri]

Not sure of the correct way to do this - it's only required for its side-effects (i.e. loading the plugin), and apparently needs to be done in the context of where the code is invoked, not defined (i.e. it isn't sufficient to put this line in optimise.R.

[richfitz]

This should be done by adding

##' @import ROI.plugin.glpk

somewhere near in R/optimise.R and rerun devtools::document() so that NAMESPACE contains an import call. That will mean that mintr can find the plugin

[richfitz]

I usually put those things in Suggests (and typically order these lists alphabetically as they're easier to maintain that way)

[richfitz]

There should be no calls to library() within package code. I presume this is working around something with dplyr's DSL? The import that you have in the NAMESPACE should sort this?

[richfitz]

the other thing you can do is to scope these variables within the function so do:

function(...) {
  zone <- intervention <- NULL # used by dplyr
  df %>% select(zone) # etc
}

There's no great way of doing this though. The .data$ trick is the canonical way but it's not nice to read

[richfitz]

This should be done by adding

##' @import ROI.plugin.glpk

somewhere near in R/optimise.R and rerun devtools::document() so that NAMESPACE contains an import call. That will mean that mintr can find the plugin