Limit nothing interpolation error to AbstractCmd subtypes

[omus]

The introduction of disallowing print/string with nothing (#27829) has introduced some headaches. From a recent #gripes session:

Mike Innes [2018-11-07 07:50]

julia> xs = [1, 2, 3]

julia> xs[nothing]
ERROR: ArgumentError: `nothing` should not be printed; use `show`, `repr`, or custom output > instead.

Simon Danisch [08:32]
I'm getting so sick of nothing should not be printed errors 😄 The worst is, they usually happen in the display stack, which has the bad habbit of removing stack traces

Eric Davies [10:43]
It feels like an evil wizard is like "haha! you've fallen into my 'printing nothing' trap!"
And it causes so many special cases to be written in Base and in user code, all because special > casing Cmd interpolation was too hard to do in the very short period of time before release.

Stefan Karpinski [10:46]
I think it would be fine to limit the nothing error to splicing nothing into command objects

Stefan Karpinski [10:46]
if someone wants to make a PR for that

[JeffBezanson]

I think this is actually a bug in the error throwing code, which should be using show/repr instead of interpolation. For example:

julia> xs["my horse is"]
ERROR: ArgumentError: invalid index: my horse is of type String

[omus]

I've also found this to be an annoyance with the change:

julia> x = nothing

julia> @error "Found unexpected value: $x"
┌ Error: Exception while generating log record in module Main at REPL[2]:1
│   exception =
│    ArgumentError: `nothing` should not be printed; use `show`, `repr`, or custom output instead.
│    Stacktrace:
│     [1] print(::Base.GenericIOBuffer{Array{UInt8,1}}, ::Nothing) at ./show.jl:566
│     [2] print_to_string(::String, ::Vararg{Any,N} where N) at ./strings/io.jl:122
│     [3] string(::String, ::Nothing) at ./strings/io.jl:155
│     [4] top-level scope at logging.jl:321
│     [5] eval(::Module, ::Any) at ./boot.jl:319
│     [6] eval_user_input(::Any, ::REPL.REPLBackend) at /Users/omus/Development/Julia/1.0/usr/share/julia/stdlib/v1.0/REPL/src/REPL.jl:85
│     [7] macro expansion at /Users/omus/Development/Julia/1.0/usr/share/julia/stdlib/v1.0/REPL/src/REPL.jl:117 [inlined]
│     [8] (::getfield(REPL, Symbol("##28#29")){REPL.REPLBackend})() at ./task.jl:259
└ @ Main REPL[2]:1

[StefanKarpinski]

I agree, let's just restrict the error to splicing nothing into command objects.

[KristofferC]

It seems like most of the time this issue comes up it is actually a bug that is being discovered.

[nickrobinson251]

#32148

[omus]

Thanks for the reminder. This issue has been resolved. At least all of the original gripes.