let NLset() coerce a numeric worldArray layer to character

[achubaty]

Relaxes the restriction added in #59, as requested.

Before

Assigning character values to only some patches of a numeric worldArray layer was an error:

NLset(world = w, agents = cbind(pxcor = 0, pycor = 0), var = "num", val = "oops")
#> Error: cannot assign character values to only some patches of the numeric worldArray layer 'num'.

The reasoning was that the numbers left in the unassigned patches would be reread as codes and silently corrupted.

After

It follows base R instead, where assigning a string into a numeric vector coerces the whole vector:

x <- 1:5; x[1] <- "a"; x
#> "a" "2" "3" "4" "5"

So the layer becomes a character one, and the patches that were not assigned keep their values as characters:

w <- NLset(world = w, agents = cbind(pxcor = 0, pycor = 0), var = "num", val = "oops")
of(world = w, agents = cbind(pxcor = 0, pycor = 0), var = "num")   #> "oops"
of(world = w, agents = cbind(pxcor = 1, pycor = 1), var = "num")   #> "17"

Simply deleting the guard would have caused exactly the corruption it was there to prevent. The values already in the layer are turned into categories in their own right before the assignment is applied, so all of them survive and none is reinterpreted. Verified across a whole layer: every one of the 25 original values is present afterwards, alongside the new one.

Unchanged:

• numbers that merely arrive as strings stay numeric, so cbind(num = 1, hab = "ice") still works and does not turn num into categories
• whole-layer assignment, which never hit the restriction
• the other layers of the array
• the underlying array stays numeric, so nothing regresses on Error with plot() and of() when the world contains a layer with character values #49

Verification

• R CMD check --as-cran: 0 errors, 0 warnings, 0 notes
• devtools::test(): 0 failures, 1231 passing
• plot() on a coerced world still emits no warnings

🤖 Generated with Claude Code