[erlang-questions] Tried this... Thought it interesting... But it didn't work.

zxq9 <>
Tue Sep 1 20:55:37 CEST 2015


On 2015年9月1日 火曜日 20:01:38 Jesper Louis Andersen wrote:
> On Tue, Sep 1, 2015 at 6:43 PM, zxq9 <> wrote:
> 
> > Honestly, I don't really think this is the best use of list comprehension
> > syntax. I can easily imagine people being confused at that, or at least
> > requiring a few moments thought to figure out wtf is going on in that line.
> > The fact this confounded you (and the example was unclear in the blog post,
> > imo) is a good reason why you shouldn't do this in your code.
> >
> 
> I must admit I sometimes do this. Consider:
> 
> f(X) ->
>     E = case X > 5 of
>         true -> [7];
>         false -> []
>     end,
>     [1,2,3] ++ E ++ [4,5].
> 
> This is easier written as
> 
> f(X) ->
>    [1,2,3] ++ [7 || X > 5] ++ [4,5].
> 
> but as a way to get the list comprehension to append like this is the only
> way I tend to use the construction. For real-world use, consider the
> following snippet from my Maps R18 tests:
> 
> https://github.com/jlouis/maps_eqc/blob/96d06da56053e87dd33c830b293dface525be17d/src/maps_eqc.erl#L693-L696
> 
> remove_args(#state { contents = C } = State) ->
>     frequency(
>       [{5, ?LET(Pair, elements(C), [element(1, Pair)])} || C /= [] ] ++
>       [{1,  ?SUCHTHAT([K], [map_key(State)], find(K, 1, C) == false)}]).
> 
> The idea here is that we want to generate arguments for removing an element
> from a map, and C contains the current contents of the map. If the map is
> non-empty, C /= [], and we can pick elements from C. Otherwise, we generate
> a map key such that it is really not an element of the map in question
> (which is trivially true if C = []). The neat part is that the first
> variant with frequency 5 is never generated for the empty map.
> 
> A case analysis in this case would tend to repeat code, so I find this
> somewhat more nimble. But these situations are probably the only situations
> on which I use this way of writing.


I sort of see your point, but I tend to write far less exciting code. I would probably do something more like:


remove_args(#state { contents = C } = State) ->
    frequency(prefix_on_empty(C) ++
              [{1,  ?SUCHTHAT([K], [map_key(State)], find(K, 1, C) == false)}]).

prefix_on_empty([]) -> [];
prefix_on_empty(C)  -> [{5, ?LET(Pair, elements(C), [element(1, Pair)])}].



Note, I have no idea what a fitting name for prefix_on_empty/1 should be, but you get the idea.

Anyway, interesting technique.

-Craig


More information about the erlang-questions mailing list