Adjust to variable result grouping

[jacobvjk]

depends on https://github.com/RMI-PACTA/pacta.multi.loanbook.analysis/pull/34 closes #27 closes #25 by coincidence

• adapts the pacta.multi.loanbook.plot package to changes made in pacta.multi.loanbook.analysis, which now allows aggregating results by any user defined variable
• all prep_*() and plot_*() functions gain flexibility in processing grouping variables
  • Where they previously expected a variable "group_id", the variable can now be specified to any name using the group_var argument
  • Omitting the argument aggregates across all input loan books, allowing for a meta view
• coincidentally, the linked changes in pacta.multi.loanbook.analysis remove the need to use the matched_prioritized data set as an input to the sankey plot. this means that pacta.multi.loanbook.plot now solely uses data objects created with pacta.multi.loanbook.analysis
• aligns on using group_var in function signature, changing category to group_var in prep_scatter_alignment_exposure() and plot_scatter_alignment_exposure()
• references to "bank" in scatter plots are changed to "group" and data_level can now be "company" or "group_var", not "bank"

[jacobvjk]

@jdhoffa @MonikaFu I am still keeping this in draft mode, as we probably want to agree on the linked PR first. But it may make sense to take a brief look at this PR now as well. Explanation is above, but just a quick tl;dr:

the aggregate metric needs to be flexible for analysis, allowing slicing and dicing of the loan books by user defined variables (e.g. by banks, but possibly also by membership in alliances, loan structures, etc). The prep and plot functions should cover this in a basic way but could probably use some prettification. open to ideas and suggestions here

[jdhoffa]

Ah this might be a bit of an anti-pattern.

Specifically for the prep_ and plot_ functions, I would tend to prefer that they always expect a single unit of input to prep or plot (so as to not overcomplicate them). And if iterating over groups is desired, I would imagine doing that either in:

• theworkflow.*
• some wrapper function
• a vignette

is probably desired?

I don't think hard-baking the grouping functionality into the plotting functions is a good idea...

[jacobvjk]

Ah this might be a bit of an anti-pattern.

Specifically for the prep_ and plot_ functions, I would tend to prefer that they always expect a single unit of input to prep or plot (so as to not overcomplicate them). And if iterating over groups is desired, I would imagine doing that either in:

• theworkflow.*
• some wrapper function
• a vignette

is probably desired?

I don't think hard-baking the grouping functionality into the plotting functions is a good idea...

It's definitely not pretty at this point and it may be worth reviewing which of the plots realisitically need to show any group information. But I think in some cases we explicitly want the group info inside the plot. See e.g. https://github.com/RMI-PACTA/workflow.aggregate.loanbooks/pull/95

more generally: the group will - at least in most cases - be used as an aesthetic (in ggplot terms). So iterating over the groups will likely not do it

[jdhoffa]

Ok got it, then totally fair enough!