nw-pbt-erlang-elixir

PBT Erlang/Elixir -- PropEr, PropCheck, StreamData

Framework Selection

| Framework | Language | Stateful | Parallel | Choose When | |-----------|----------|----------|----------|-------------| | PropEr | Erlang | Yes | Yes (linearizability) | Erlang projects needing full stateful/parallel testing | | PropCheck | Elixir | Yes (via PropEr) | Yes (via PropEr) | Elixir projects needing stateful/parallel testing | | StreamData | Elixir | No | No | Elixir projects needing only stateless PBT |

PropCheck wraps PropEr with Elixir syntax. StreamData is pure Elixir but lacks stateful testing.

Quick Start

%% PropEr (Erlang)
-include_lib("proper/include/proper.hrl").

prop_sort_preserves_length() ->
    ?FORALL(List, list(integer()),
        length(lists:sort(List)) =:= length(List)).

%% Run: proper:quickcheck(my_module:prop_sort_preserves_length()).

Syntax Differences

| Concept | PropEr (Erlang) | PropCheck (Elixir) | StreamData (Elixir) | |---------|-----------------|--------------------|--------------------| | Import | -include_lib("proper/include/proper.hrl"). | use PropCheck | use ExUnitProperties | | Property | ?FORALL(X, gen(), body) | forall x <- gen() do body end | check all x <- gen() do assert body end | | Integer | integer() | integer() | integer() | | List | list(integer()) | list(integer()) | list_of(integer()) | | Run | proper:quickcheck(prop()) | mix test | mix test |

Generator Cheat Sheet (PropEr)

Primitives

integer()                  % any integer
integer(0, 100)            % bounded
float()
binary()                   % binary data
boolean()
atom()

Collections

list(integer())            % list of integers
non_empty(list(integer())) % non-empty list
vector(5, integer())       % fixed-length list
{integer(), binary()}      % tuple (direct syntax)

Combinators

oneof([integer(), binary()])      % union
elements([a, b, c])               % pick from list
frequency([{80, integer()}, {20, atom()}])  % weighted

%% ?LET (map/transform)
even() ->
    ?LET(N, integer(), N * 2).

%% ?SUCHTHAT (filter -- use sparingly)
non_empty_list() ->
    ?SUCHTHAT(L, list(integer()), L =/= []).

%% Nested ?LET (dependent generation)
list_and_element() ->
    ?LET(List, non_empty(list(integer())),
        ?LET(Elem, elements(List),
            {List, Elem})).

Recursive

tree(Type) ->
    ?SIZED(Size, tree(Size, Type)).
tree(0, Type) -> {leaf, Type};
tree(Size, Type) ->
    frequency([
        {1, {leaf, Type}},
        {5, ?LAZY({node, Type, tree(Size div 2, Type), tree(Size div 2, Type)})}
    ]).

Shrinking

year() ->
    ?SHRINK(integer(0, 9999), [integer(1970, 2000)]).

date() ->
    ?LETSHRINK([Y, M, D],
               [integer(1, 9999), integer(1, 12), integer(1, 31)],
               {Y, M, D}).

Stateful Testing (proper_statem)

-behaviour(proper_statem).

initial_state() -> #{items => #{}}.

command(#{items := Items}) ->
    oneof([
        {call, ?MODULE, put, [key(), value()]},
        {call, ?MODULE, get, [elements(maps:keys(Items))]}
            || maps:size(Items) > 0
    ]).

precondition(#{items := Items}, {call, _, get, [Key]}) ->
    maps:is_key(Key, Items);
precondition(_, _) -> true.

postcondition(#{items := Items}, {call, _, get, [Key]}, Result) ->
    Result =:= maps:get(Key, Items);
postcondition(_, _, _) -> true.

next_state(State = #{items := Items}, _Var, {call, _, put, [Key, Val]}) ->
    State#{items := Items#{Key => Val}};
next_state(State, _, _) -> State.

prop_store() ->
    ?FORALL(Cmds, commands(?MODULE),
        begin
            {History, State, Result} = run_commands(?MODULE, Cmds),
            cleanup(),
            Result =:= ok
        end).

StreamData: No stateful testing. PropCheck: Same proper_statem callbacks with Elixir syntax.

Parallel Testing (Linearizability)

prop_store_parallel() ->
    ?FORALL(Cmds, parallel_commands(?MODULE),
        begin
            {Sequential, Parallel, Result} = run_parallel_commands(?MODULE, Cmds),
            cleanup(),
            Result =:= ok
        end).

Swap commands for parallel_commands, run_commands for run_parallel_commands. Framework generates sequential prefix + parallel branches and checks all linearizations.

Test Runner Integration

%% PropEr: {deps, [{proper, "1.4.0"}]}. Run: rebar3 proper
%% PropCheck: {:propcheck, "~> 1.4", only: :test}. Run: mix test
%% StreamData: {:stream_data, "~> 1.0", only: :test}. Run: mix test

Unique Features

• proper_fsm: Finite state machine testing module (distinct from generic state machines)
• PULSE integration: Controlled scheduling for parallel test race detection
• Symbolic references: First-class support with two-phase (abstract + concrete) execution
• collect/aggregate: Built-in distribution analysis for generator quality verification
• Linearizability checking: One of the most mature implementations in any PBT framework