Equality of OrderedDict

[lucaferranti]

I am a little confused by the following, is it intentional that == does not take into account the order of keys for OrderedDict?

julia> a1 = OrderedDict(:A => 1, :B => 2, :C => 3)
OrderedDict{Symbol, Int64} with 3 entries:
  :A => 1
  :B => 2
  :C => 3

julia> a2 = OrderedDict(:B => 2, :A => 1, :C => 3)
OrderedDict{Symbol, Int64} with 3 entries:
  :B => 2
  :A => 1
  :C => 3

julia> a1 == a2
true

this is mainly motivated by testing purposes, as I think in tests one wants to check that also the keys order is correct. Well, arguably one can always do

@test all(kv1 == kv2 for (kv1, kv2) in zip(a1, a2))

so maybe it's not a biggie.

[oxinabox]

This is because it is falling back to the generic == on AbstractDicts, which are unordered.

julia> od = OrderedDict(1=>2)
OrderedDict{Int64, Int64} with 1 entry:
  1 => 2

julia> @which ==(od, od)
==(l::AbstractDict, r::AbstractDict) in Base at abstractdict.jl:468

Whether or not this is sensible is not clear to me.

For what it is worth our behavour disagrees with python.

>>> from collections import OrderedDict
>>>  d = OrderedDict.fromkeys('abcde')
>>>  d2 = OrderedDict.fromkeys('edcba')
>>> d==d2
False

>>> d==d
True

Python treats ordered dicts as equal only if they have the same contents in the same order)

[einarpersson]

I definitely expected that == would differentiate between two OrderedDicts if they differ in order. Isn´t that the whole point, i.e. saying that "no these are not same despite having the same elements, because they are in different order".

[jariji]

For any type T, x::T and y::T should not have x == y unless x and y agree on all aspects of their public API. Since order is part of the public API of Ordered containers, they should not be ==(::T, ::T) unless they agree on order.