Different semantics for pars.character() in term and mbr (embr)

[krlmlr]

The Bug

This gave me pause when working on poissonconsulting/mcmcderive#2 .

A Reprex

library(term)
expr <- "
for (i in 1:nObs) {
  log(eCount[i]) <- b0 + bYear * Year[i] + bKelpLine * KelpLine[i]
}"
pars(expr)
#> Error in `term_impl()` at term/R/term.R:50:4:
#> ! All elements of term vector `string_args_term` must be valid.
#> Backtrace:
#>     ▆
#>  1. ├─universals::pars(expr)
#>  2. └─term:::pars.character(expr)
#>  3.   └─term::term(x) at term/R/pars.R:39:2
#>  4.     └─term:::term_impl(args) at term/R/term.R:50:4
#>  5.       └─term::chk_term(string_args_term, "valid") at term/R/internal.R:67:2
#>  6.         └─chk::abort_chk("All elements of term vector ", x_name, " must be valid") at term/R/chk.R:25:4
#>  7.           └─chk::err(..., n = n, tidy = tidy, class = "chk_error", call = call)
#>  8.             └─rlang::abort(msg, class = class, !!!args[named], call = call)

mbr:::pars.character(expr)
#> [1] "b0"        "bKelpLine" "bYear"     "eCount"    "i"         "KelpLine" 
#> [7] "log"       "nObs"      "Year"

^{Created on 2023-05-30 with reprex v2.0.2}

[joethorley]

Yes there is an exported terms::pars.character() that treats the input as a term vctr and an internal mbr:::pars.character() that treats the input as model code.

I was thinking that the mbr function would be hidden. We could rename it as its internal to avoid confusion?

[krlmlr]

Agree to renaming it to a simple non-generic function.