views:

97

answers:

3

I'm trying to come up with a dictionary-like data structure that I can use in Erlang. The goal is to guarantee that all the values, as well as the keys, are unique. I can do it with explicit consistency checks after every modification, but I'm hoping there's an obscure type that does this for me. Is there one? If not, is there a better way than wrapping a check into every function that modifies the data (or returns a slightly different copy)?

I'm expecting to have at least 120 elements and no more than a few thousand, in case that matters.

+1  A: 

Is there one?

Not in the standard library, I believe. I would use a pair consisting of a dict() and a set() of values.

Alexey Romanov
+3  A: 

What about something like that:

-module(unidict).

-export([
    new/0,
    find/2,
    store/3
    ]).


new() ->
    dict:new().


find(Key, Dict) ->
    dict:find({key, Key}, Dict).


store(K, V, Dict) ->
    Key = {key, K},
    case dict:is_key(Key, Dict) of
        true ->
            erlang:error(badarg);
        false ->
            Value = {value, V},
            case dict:is_key(Value, Dict) of
                true ->
                    erlang:error(badarg);
                false ->
                    dict:store(Value, K, dict:store(Key, V, Dict))
            end
    end.

Example shell session:

1> c(unidict).
{ok,unidict}
2> D = unidict:new().
{dict,0,16,16,8,80,48,
      {[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[]},
      {{[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[]}}}
3> D1 = unidict:store(key, value, D).
{dict,2,16,16,8,80,48,
      {[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[]},
      {{[],[],[],[],[],[],[],[],[],[],[],[],[],[],
        [[{key,key}|value],[{value,...}|{...}]],
        []}}}
4> D2 = unidict:store(key, value, D1).
** exception error: bad argument
     in function  unidict:store/3
5> D2 = unidict:store(key2, value, D1).
** exception error: bad argument
     in function  unidict:store/3
6> D2 = unidict:store(key, value2, D1).
** exception error: bad argument
     in function  unidict:store/3
7> D2 = unidict:store(key2, value2, D1).
{dict,4,16,16,8,80,48,
      {[],[],[],[],[],[],[],[],[],[],[],[],[],[],[],[]},
      {{[],
        [[{key,key2}|value2]],
        [],[],[],[],[],[],[],[],[],[],[],
        [[{value,value2}|{key,key2}]],
        [[{key,key}|value],[{value,...}|{...}]],
        []}}}
8> unidict:find(key, D2).
{ok,value}
9> unidict:find(key2, D2).
{ok,value2}
hdima
I like this, but is there a reason you used `dict:find` instead of `dict:is_key`?
Nathon
Using `dict:find` will work just as well. (Not my code)
rvirding
You're right dict:is_key can be slightly better here. I thought about dict:update first and then just try dict:find
hdima
+1  A: 

You can use a simple {key, value} list for a couple of hundred elements:

put(L, K, V) ->
  case lists:keyfind(K, 1, L) of
    {K, _} -> erlang:error(badarg);
    false ->
      case lists:keyfind(V, 2, L) of
        {_, V} -> erlang:error(badarg);
        false ->  [{K,V} | L]
      end
  end.

get(L, K) ->
  case lists:keyfind(K, 1, L) of
    {K, V} -> {'value', V};
    false ->  'undefined'
  end.
Zed