Tuple and NamedTuple indexing corner cases

[bkamins]

I want to raise some cases with indexing Tuple and NamedTuple.

Tuple

You are allowed to index Tuple with values that are not allowed for arrays:

julia> (1,2,3)[1.0]
1

julia> (1,2,3)[true]
1

Should we disallow this?

NamedTuple

You are not allowed to index NamedTuple with vectors, e.g.

julia> (a=1,b=2,c=3)[[1,2]]
ERROR: MethodError: no method matching getindex(::NamedTuple{(:a, :b, :c),Tuple{Int64,Int64,Int64}}, ::Array{Int64,1})
Closest candidates are:
  getindex(::NamedTuple, ::Int64) at namedtuple.jl:101
  getindex(::NamedTuple, ::Symbol) at namedtuple.jl:102
Stacktrace:
 [1] top-level scope at none:0

julia> (a=1,b=2,c=3)[[:a,:b]]
ERROR: MethodError: no method matching getindex(::NamedTuple{(:a, :b, :c),Tuple{Int64,Int64,Int64}}, ::Array{Symbol,1})
Closest candidates are:
  getindex(::NamedTuple, ::Int64) at namedtuple.jl:101
  getindex(::NamedTuple, ::Symbol) at namedtuple.jl:102
Stacktrace:
 [1] top-level scope at none:0

julia> (a=1,b=2,c=3)[[true,false,true]]
ERROR: MethodError: no method matching getindex(::NamedTuple{(:a, :b, :c),Tuple{Int64,Int64,Int64}}, ::Array{Bool,1})
Closest candidates are:
  getindex(::NamedTuple, ::Int64) at namedtuple.jl:101
  getindex(::NamedTuple, ::Symbol) at namedtuple.jl:102
Stacktrace:
 [1] top-level scope at none:0

Here the question is if we should allow this.

CC @nalimilan, @pdeffebach

[nalimilan]

I think we should support the same operations for NamedTuple as for Tuple. I don't see why we would be more restrictive for the latter, except maybe for things like nt[true] and nt[1.0] which don't work with arrays and should probably never have been allowed for tuples.

[bkamins]

This is what I mean. The question is if nt[true] and nt[1.0] is a bug or a feature 😄.

[mcabbott]

xref issues #27021, #38834, and PRs #31423, #38878