fix diff for 32bit arrays

[dpinol]

Fixes #20

[Teo-ShaoWei]

I would be hesitant to add in a specific hack for vectors. Effectively :(variable_name[index]) is Expr(:ref, :variable_name, index) and so there's a potential regression point. Most importantly, we will open a can of worms as we now need to cater for vectors of different container types but same inner values...

[Teo-ShaoWei]

Since e::Expr, we have isa(e.args, Vector) to always be true. So the following suffice

Suggested change

	if e.head === :ref && isa(e.args, Vector)
	return true
	end
	return false
	return e.head === :ref

[ReubenJ]

Just a note: this fix is for arrays of any type that is not listed here—not just 32bit arrays. For example:

julia> string([true, true])
"Bool[1, 1]"

When Meta.parseing that string, the result is an expression of type :ref, too:

julia> Meta.parse(string([true, false]))
:(Bool[1, 0])

julia> Meta.parse(string([true, false])).head
:ref

julia> Meta.parse(string([true, false])).args
3-element Vector{Any}:
  :Bool
 1
 0

[ReubenJ]

And this is not quite true. The arrays get here with the same syntax as a Ref, but they're not actually a ref to a vector. Just a bit of a confusing overlap in the display of vectors and the syntax for Refs.