unique() changes input order

[diegozea]

julia> unique(collect("AABBCC"))
3-element Char Array:
 'A'
 'C'
 'B'

julia> unique(collect("ABC"))
3-element Char Array:
 'A'
 'C'
 'B'

julia> @which  unique(collect("AABBCC"))
unique(C) at set.jl:97

The actual definition of unique using Set and union! change the input order of the array.

[StefanKarpinski]

I'm unclear what this issue means. What's the problem here?

[diegozea]

I'm not sure if this is an issue or not.
I expect that the order doesn't change with respect to the order of the first apparition of every element... Like in R:

> unique(c("A","B","C"))
[1] "A" "B" "C"
> unique(c("A","A","B","B","C","C"))
[1] "A" "B" "C"

I notice it because I was using unique in a context where the order is important.

[diegozea]

In my particular case, I was doing something like:

enumerate(unique(collect("AABBCC")))

And waiting for

(1,'A')
(2,'B')
(3,'C')

Instead of

(1,'A')
(2,'C')
(3,'B')

[diegozea]

In this case sort! ( like was mentioned in #1645 ) can not be used. Here the important order is the input (user defined) order an not the numeric or alphabetic order.

But #1845 can be very useful in this case.

For now I'm using ( #550 can be useful here for avoid the last line ):

function unique_order{T}(ab::Array{T,1})
  intset = IntSet(ab...)
  out = [ delete!(intset,elem,0) for elem in ab ]
  out[ out .!= 0 ]
end

[JeffBezanson]

I guess it might as well preserve order.

[diegozea]

Thanks Jeff!!!

[pao]

Note (for future archaeologists) that this now does not behave like MATLAB's unique. The MATLAB version uses a string argument 'stable' to get this behavior. http://www.mathworks.com/help/matlab/ref/unique.html