Adding Array.uniq_by(proc) to core_ext, allowing selection of unique elements based on their response to a proc

I can't download the original attachment, so trying another upload to see if that fixes it.

Thanks! I can't think of any off the top of my head, but I'll definitely take a look.

+1 for the idea. I wrote a version of the method a while back (except using inject + a hash instead of a select within a select as you've got). It does come in handy.

Question: Why does your uniq_by accept a Proc object? Shouldn't it just accept a block? I.e.:

arr.uniq_by {|x| ... }

as opposed to

arr.uniq_by(Proc.new {|x| ... })

Thanks for the suggestion. You're right - blocks result in much nicer-looking method calls than procs. I updated the patch to use blocks instead.

Why not just override Array#uniq and uniq! instead of adding a new method ?

It seemed like overriding a Ruby method as common as [].uniq might cause confusion so I opted to create a new, explicitly-named method instead.

I think doing something like this might be ok :

  def uniq
    if block_give?
      select {|item| select {|i| yield(i) == yield(item)}.index(item) == 0 }
    else
      super
    end
  end

Just a thought. I'm fine with a new method too. I just think it's more intuitive to make uniq accept a block.

Joey's implementation O(n^3 / m) where n is the number of elements in the array and m is the number of unique values. I came up with a different implementation that is more performant: http://gist.github.com/159799

I like Pratik's idea of overriding uniq. There've been more than a few times that I've used uniq expecting that it takes a block.

Otherwise I like the general idea of this patch :)

+1 since was always using the same method for a long time available in Facets http://facets.rubyforge.org/git?p=facets.git;a=blob_plain;f=lib/cor...

Also if you guys came with more efficient implementation I think it's worth sending it to Facets as well.

Ok, the Facets link got crippled, here is the shortened one http://bit.ly/QMjcp

+1 I could use something like this right now. Dan's patch is nice as well--those speed improvements are excellent.

Just discovered this ticket while Googling for my version of Enumerable#uniq_by. ;)

Here's the version I came up with if it helps: http://gist.github.com/168544