Bookish spork

Version: 0.5.1

Authors: Alexey Nikitin ([email protected]) (web site: https://twitter.com/tank_bohr).

An erlang library to test http requests. Inspired by Ruby's WebMock.

Suitable for Elixir.

Rationale

There are several ways to test your http interaction

Real http request to real servers: not very reliable, requires internet
You can use external http server like https://httpbin.org/ (hackney approach)
You can mock your http client library
Also you can run an http-server within your application on your localhost on a particular port

The last approach is the best IMHO. It is absolutely http-client agnostic. It doesn't require internet connection or any external utilities.

bookish_spork provides you facilities to test your requests with real http server.

Usage

Bookish spork supports Erlang/OTP 20.3 or later.

First step: add to your rebar config

{profiles, [
    {test, [
        {deps, [
            {bookish_spork, "0.5.1"}
        ]}
    ]}
]}.

Second: start server in your tests.

bookish_spork:start_server().

It starts process without link. Thus you can use it in init_per_group and in init_per_suite callbacks. Default port is 32002 but you can specify any port you like with bookish_spork:start_server/1

Stub request

The simplest stub you can do is

bookish_spork:stub_request().

It will stub your requests with 204 No Content response with empty body.

If you need specify response you easily can do this:

bookish_spork:stub_request([Status, Headers, Content]).

Capture request

As usual the main goal is to test that you send the correct request

{ok, Request} = bookish_spork:capture_request().

It returns you an opaque structure of the request. You can inspect it with

Bypass comparison

An elixir library bypass does pretty much the same. And illustrates the same approach. It starts a cowboy web-server to replace a real service for test. It's a beautiful library with great API, documentation, and very concise source code. If you are an elixir developer, most likely, it will be a good fit for you.

But nevertheless bookish_spork has some advantages:

Bypass depends on cowboy and plug. Bookish spork has zero dependencies.
Bookish spork works seamlessly with both erlang and elixir. Bypass is supposed to be an elixir only library.
~~Bookish spork much simpler (I believe)~~ (not any more).
Bookish spork allows you to inspect the request very deeply and accurate. For example take a look at bookish_spork_request:raw_headers/1 and bookish_spork_request:ssl_info/1 and bookish_spork_request:tls_ext/1. It can be useful for HTTP clients testing.

Elli comparison

Very often people use elli for this purpose. But elli is a full-featured web-server while bookish_spork is a testing library. It allows you to stub requests as close to your tests as possible. Without callback module and supervisor.

Examples

Setup and teardown

init_per_group(_GroupName, Config) ->
    {ok, _} = bookish_spork:start_server(),
    Config.

end_per_group(_GroupName, _Config) ->
    ok = bookish_spork:stop_server().

Set expectation

init_per_testcase(random_test, Config) ->
    bookish_spork:stub_request([200, #{}
        <<"{\"value\": \"Chuck Norris' favourite word: chunk.\"}">>]),
    Config.

Make assertions

random_test(_Config) ->
    ?assertEqual(<<"Chuck Norris' favourite word: chunk.">>, testee:make_request()),
    {ok, Request} = bookish_spork:capture_request(),
    ?assertEqual("/jokes/random", bookish_spork_request:uri(Request)).

As you can see there are two types of assertions:

we check a testee function result
we check a side effect: verifying outgoing request has correct attributes (uri in this case)

More complex expectations

There are cases when the testee function initiates more than one request. But if you know the order of your requests, you can set several expectations

bookish_spork:stub_request([200, #{}, <<"{\"value\": \"The first response\"}">>]),
bookish_spork:stub_request([200, #{}, <<"{\"value\": \"The second response\"}">>]).

The library will response in the order the stubs were defined.

Sometimes you can't guarantee the order of requests. Then you may stub request with the fun

bookish_spork:stub_request(fun(Request) ->
    case bookish_spork_request:uri(Request) of
        "/bookish/spork" ->
            [200, #{}, <<"Hello">>];
        "/admin/sporks" ->
            [403, #{}, <<"It is not possible here">>]
    end
end)

Module to work with request

Module to work with response

Stub multiple requests with one response

It can be useful to stub several requests with one command

bookish_spork:stub_request([200, #{<<"Content-Type" => "text/plan">>}, <<"Pants">>], _Times = 20)

The same with the fun

bookish_spork:stub_request(fun(Req) ->
    Body = bookish_spork_request:body(Req),
    [200, #{<<"X-Respond-With">> => <<"echo">>}, Body]
end, _Times = 150)

As you can see that it's not necessary to build response structure yourself. You can use handy three-element tuple or list syntax to define the response. But the bookish_spork_response:new/1 still works.

Elixir example

defmodule ChuckNorrisApiTest do
  use ExUnit.Case
  doctest ChuckNorrisApi

  setup do
    {:ok, _} = :bookish_spork.start_server()
    on_exit(fn -> :bookish_spork.stop_server() end)
  end

  test "retrieves a random joke" do
    :bookish_spork.stub_request([200, %{}, "{
      \"value\": \"Chuck norris tried to crank that soulja boy but it wouldn't crank up\"
    }"])
    assert ChuckNorrisApi.random == "Chuck norris tried to crank that soulja boy but it wouldn't crank up"

    {:ok, request} = :bookish_spork.capture_request()
    assert request.uri === "/jokes/random"
  end
end