Refactor dropped non-numeric columns with recover

[kylebaron]

Summary

See #1184

• This used send a message when important column was dropped due to non-numeric data; the PR refactors the behavior to generate error; I also considered warning, but I'd rather just stop the simulation in this case
• Throw error rather than message when an important column is dropped; originally aiming for parameters getting dropped, but there will also be error with a dosing column is dropped because it is non-numeric; same for 64 bit integer
• Refactor recover handling in mrgsim() to defer errors for non-numeric columns that are also parameters; this lets the validate data set functions issue the error

library(mrgsolve)
#> 
#> Attaching package: 'mrgsolve'
#> The following object is masked from 'package:stats':
#> 
#>     filter
library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union


mod <- house(end = 0)

data <- expand.ev(amt = 100, WTVC = "a")

mrgsim(mod, idata = data)
#> Error in `valid_idata_set()`:
#> ! Found input data that cannot be used for simulation
#> ✖ idata set column: WTVC (character)
#> Backtrace:
#>     ▆
#>  1. └─mrgsolve::mrgsim(mod, idata = data)
#>  2.   └─mrgsolve::mrgsim_i(x, idata = idata, ...)
#>  3.     ├─base::do.call(...)
#>  4.     └─mrgsolve (local) `<fn>`(x = `<packmod>`, data = `<int[,1]>`, idata = `<df[,6]>`)
#>  5.       └─mrgsolve::valid_idata_set(idata, x, verbose = verbose)
#>  6.         └─mrgsolve:::signal_drop(dm, x, to_signal, context = "idata set")
#>  7.           └─rlang::abort(...)

mrgsim(mod, idata = data, recover = "WTVC")
#> Error in `valid_idata_set()`:
#> ! Found input data that cannot be used for simulation
#> ✖ idata set column: WTVC (character)
#> Backtrace:
#>     ▆
#>  1. └─mrgsolve::mrgsim(mod, idata = data, recover = "WTVC")
#>  2.   └─mrgsolve::mrgsim_i(x, idata = idata, ...)
#>  3.     ├─base::do.call(...)
#>  4.     └─mrgsolve (local) `<fn>`(...)
#>  5.       └─mrgsolve::valid_idata_set(idata, x, verbose = verbose)
#>  6.         └─mrgsolve:::signal_drop(dm, x, to_signal, context = "idata set")
#>  7.           └─rlang::abort(...)

^{Created on 2024-05-21 with reprex v2.0.2}

[kylebaron]

Updated this logic for clarity.

[kylebaron]

We have to explicitly exit now that errors are generated; I think before, there might have been nothing in drop but we kept going and no one knew.

[kylebaron]

This is mostly called under the hood; but I filled out details since we're referencing the function name in the error message and probably good to get more of this written down somewhere.

[kylebaron]

Is referencing this internal GLOBALS data structure a bad idea? I don't want to have to maintain this list off the contents of what is in there but also feels odd to expose this internal data.

[kyleam]

In my opinion it's okay to reference. It's useful to those that want to see the list, but it being internal should make it clear that it shouldn't be relied on in scripts, etc.

[kylebaron]

Honestly can't remember why we're dropping data columns here when it will happen later. For now, will keep it as is.

[kyleam]

LGTM, left one comment about adjacent doc typo

[kyleam]

In my opinion it's okay to reference. It's useful to those that want to see the list, but it being internal should make it clear that it shouldn't be relied on in scripts, etc.