I ran into the following (Julia v1.5.3):
julia> A = rand(2,2);
julia> Meta.@lower A[end,1]
:($(Expr(:thunk, CodeInfo(
@ none within `top-level scope'
1 ─ %1 = Base.lastindex(A, 1)
│ %2 = Base.getindex(A, %1, 1)
└── return %2
))))
julia> Meta.@lower A[begin,1]
:($(Expr(:thunk, CodeInfo(
@ none within `top-level scope'
1 ─ %1 = Base.axes(A, 1)
│ %2 = Base.first(%1)
│ %3 = Base.getindex(A, %2, 1)
└── return %3
))))
It seems that when using two indices for a matrix begin and end lowers to different implementations. Is there any reason for this? (I would have expected A[begin,1] to call Base.firstindex(A, i) the same way end calls Base.lastindex(A, i).
Note that if we simply use A[begin] then it calls Base.firstindex(A) (and similarly A[end] calls Base.lastindex(A)).
