reduce and foldr/foldl are inconsistent for + and empty vectors

[dlfivefifty]

In 0.6, reduce and foldr/foldl are inconsistent for empty vectors:

julia> reduce(+,Float64[])
0.0

julia> foldr(+,Float64[])
ERROR: BoundsError: attempt to access 0-element Array{Float64,1} at index [0]
Stacktrace:
 [1] foldr(::Function, ::Array{Float64,1}) at ./reduce.jl:160

julia> foldl(+,Float64[])
0.0

[dlfivefifty]

Related issue is #5797 (comment)

[vsartor]

I'm new to Julia so take all of this with a grain of salt.

So the foldr with two arguments (here) is calling mapfoldr with three arguments (here).

It's defined as
mapfoldr(f, op, itr) = (i = endof(itr); mapfoldr_impl(f, op, f(itr[i]), itr, i-1))

So not only would it send -1 as the i argument to mapfoldr_impl, it also tries to apply the function to itr[i], which is the 0-index access that happened here.

I think the check for i == 0 in mapfoldr_impl should (also?) happen before the call, since to pass the v0 argument, it needs to access a member of the array (which in this case is impossible).

reduce and foldl seem to use mapfoldl, which does the proposed check (here).

Should be an easy fix if I'm right.

[vsartor]

So, the check on foldl/reduce was added as a fix for #7465, but it wasn't added to foldr, so I think it really is intended for a check to exist in mapfoldr for empty arguments.

I'm gonna go ahead and work on a PR if that's okay.

[vsartor]

This issue should be closed (see #20160).

[Sacha0]

Thanks @Vsart!