eachindex should throw DimensionMismatch for differently sized tuples

[cafaxo]

Currently, we have

julia> eachindex((), (1,))
Base.OneTo(1)

From the documentation of eachindex, I expected a DimensionMismatch instead. Like this:

julia> eachindex(1:0, 1:1)
ERROR: DimensionMismatch: all inputs to eachindex must have the same indices, got Base.OneTo(0) and Base.OneTo(1)

(Previous discussion on slack: https://julialang.slack.com/archives/C6A044SQH/p1671032794309049)

[fredrikekre]

Current behavior introduced in #14100, when eachindex returned the max, and #25108 missed to update this to the new behavior for tuples.

[cafaxo]

This also seems to behave incorrectly:

julia> itr0 = (i for i in 1:0);

julia> itr1 = (i for i in 1:1);

julia> eachindex(itr0)
0-element LinearIndices{1, Tuple{Base.OneTo{Int64}}}

julia> eachindex(itr1)
1-element LinearIndices{1, Tuple{Base.OneTo{Int64}}}:
 1

julia> eachindex(itr0, itr1)
ERROR: MethodError: no method matching keys(::Base.Generator{UnitRange{Int64}, typeof(identity)}, ::Base.Generator{UnitRange{Int64}, typeof(identity)})

Not sure if eachindex even makes sense for generators... but then it should give Base.OneTo instead of LinearIndices.
Should keys(::Tuple) give LinearIndices to be consistent with keys(::Generator) and keys(::AbstractArray)?
(keys(::Generator) was introduced in #34678)

[Snimm]

Hi, I am new to coding, and I want to work on this issue.

[piyushjain16]

Is the issue closed ?. if not I would like to work on it.

[oscardssmith]

not closed, but #48125 looks like it will fix this once it merges.

[piyushjain16]

Ok thank you

[uroosa114]

I would like to work on this