intersection problem

[meh]

intersection problem

[gvaughn]

nevermind me. I figured it out.

[meh]

Just curious, what was it?

[gvaughn]

Well, that code above will generate that error. I think my issue was both that I expected more magic, and it's too magic for my understanding. You can't (or didn't) add your Data.Set protocol to a raw list object as I naively tried. I looked at your implementation and I can't figure out what L.new is referring to, but it plainly works if I pass in parameters of record Data.Set.BalancedTree.

If you're more curious (and I'm happy for any suggestions to improve) my code that brings in this lib is on the datastructures branch of my elixir_kata repo: https://github.com/gvaughn/elixir_kata/tree/datastructures/conway
I kept it separate from master because I'm not honestly sure this is an overall improvement. It was a good learning experience for me. On master I'm calling into the raw erlang :set to do set behaviors in a single place, otherwise working with a list.

I also know things are changing quickly in elixir-land. I see that there is a Set protocol now for elixir 0.10.0, which I don't think was there when I wrote this on the last release. But I don't see a sorted set in the standard lib. Do you picture this library moving into the standard lib?

[meh]

I started this when I started working on a Stack and Queue implementations, and they were refused.

I sincerely doubt this will be ever merged in, but this isn't going to be abandoned, I use it in all my projects including ones that are going in production.

Also Set in the stdlib is not a protocol, but rather a behaviour like Dict.

Nonetheless I think this is actually a bug, so I'll look at it right now.

[meh]

iex(1)> Data.Set.intersection([1,2,3],[1,2])
[1, 2]

🐼

[gvaughn]

Sorry. The dividing line between behaviours and protocols are still fuzzy to me.

Are you riding elixir HEAD? I'm on the 0.10.0 release, with erlang R16B (if that matters) and the error is still there.

[me@GGV-LS conway (datastructures)]$ iex -S mix
Erlang R16B (erts-5.10.1) [source] [64-bit] [smp:8:8] [async-threads:10] [hipe] [kernel-poll:false] [dtrace]

/Users/enduser/code/elixir_kata/conway/lib/conway.ex:2: unused import Data.Set.BalancedTree
Compiled lib/conway.ex
Generated conway.app
Interactive Elixir (0.10.0) - press Ctrl+C to exit (type h() ENTER for help)
iex(1)> Data.Set.intersection([1,2,3],[1,2])
** (UndefinedFunctionError) undefined function: L.new/1
L.new([1, 2])
/Users/enduser/code/elixir_kata/conway/deps/datastructures/lib/data/set.ex:64: Data.Set.List.intersection/2
erl_eval.erl:569: :erl_eval.do_apply/6
src/elixir.erl:147: :elixir.eval_forms/3
/private/tmp/elixir-FC3J/elixir-0.10.0/lib/iex/lib/iex/server.ex:112: IEx.Server.eval/4
/private/tmp/elixir-FC3J/elixir-0.10.0/lib/iex/lib/iex/server.ex:52: IEx.Server.wait_input/1
/private/tmp/elixir-FC3J/elixir-0.10.0/lib/iex/lib/iex/server.ex:31: IEx.Server.start/1

[meh]

Did you run mix deps.update before trying that?

[gvaughn]

No. I didn't realize you had pushed a fix. I thought you couldn't reproduce the problem. After deps.update I get the proper response! Thanks.

Now I need to go rewrite all my code if that was supposed to work ;-)

[meh]

My fault, I wasn't clear about it :)

[gvaughn]

@meh I got around to changing my code and I'm seeing another bug in intersection when I use tuple elements:

iex(4)> Data.Set.intersection([{1,2},{1,1},{1,0}], [{1,1}])
[]

My older implementation of this logic before finding this library is:

iex(1)> :sets.from_list([{1,2},{1,1},{1,0}]) |> :sets.intersection(:sets.from_list([{1,1}])) |> :sets.to_list
[{1, 1}]

[meh]

Sadly there's nothing I can do, unless I reimplement ordsets.

iex(2)> :ordsets.intersection([{1,2},{1,1},{1,0}], [{1,1}])
[]

You can always use Data.Set.BalancedTree or Data.Set.Standard.

[gvaughn]

Wow. That's core erlang stuff. I wonder if this is a tuple specific problem.

Anyway, I've got another implementation that I'll be able to show-n-tell tonight at our local elixir meet up. Thanks for your help

[meh]

You're welcome.

[gvaughn]

I figured out what's up. ordsets is optimized to assume the inputs are ordered already. Notice:

iex(1)> :ordsets.intersection([{1,0},{1,1},{1,2}], [{1,1}])
[{1, 1}]

[meh]

Oh yeah, the first parameter is always assumed to be already an ordset, only the second one is converted, forgot about it, sorry.

[meh]

Data.Set.List.new([{1,2},{1,1},{1,0}]) |> Data.Set.intersection([{1,1}])

This would be the "proper" way to do it.