Hi everyone,

With the upgrade of supported Erlang version and introduction of Elixir into 
our integration test suite we have an opportunity to replace currently used 
eunit (for new tests only) with Elixir based ExUnit. 
The eunit testing framework is very hard to maintain. In particular, it has the 
following problems:
- the process structure is designed in such a way that failure in setup or 
teardown of one test affects the execution environment of subsequent tests. 
Which makes it really hard to locate the place where the problem is coming from.
- inline test in the same module as the functions it tests might be skipped
- incorrect usage of ?assert vs ?_assert is not detectable since it makes tests 
pass 
- there is a weird (and hard to debug) interaction when used in combination 
with meck 
   - https://github.com/eproxus/meck/issues/133#issuecomment-113189678
   - https://github.com/eproxus/meck/issues/61
   - meck:unload() must be used instead of meck:unload(Module)
- teardown is not always run, which affects all subsequent tests
- grouping of tests is tricky
- it is hard to group tests so individual tests have meaningful descriptions

We believe that with ExUnit we wouldn't have these problems:
- on_exit function is reliable in ExUnit
- it is easy to group tests using `describe` directive
- code-generation is trivial, which makes it is possible to generate tests from 
formal spec (if/when we have one)

Here are a few examples:

# Test adapters to test different interfaces using same test suite

CouchDB has four different interfaces which we need to test. These are:
- chttpd
- couch_httpd
- fabric
- couch_db

There is a bunch of operations which are very similar. The only differences 
between them are:
- setup/teardown needs different set of applications
- we need to use different modules to test the operations

This problem is solved by using testing adapter. We would define a common 
protocol, which we would use for testing.
Then we implement this protocol for every interface we want to use.

```
defmodule Couch.Test.CRUD do
  use ExUnit.Case
  alias Couch.Test.Adapter
  alias Couch.Test.Utils, as: Utils

  alias Couch.Test.Setup

  require Record

  test_groups = [
    "using Clustered API": Adapter.Clustered,
    "using Backdoor API": Adapter.Backdoor,
    "using Fabric API": Adapter.Fabric,
  ]

  for {describe, adapter} <- test_groups do
    describe "Database CRUD #{describe}" do
      @describetag setup: %Setup{}
        |> Setup.Start.new([:chttpd])
        |> Setup.Adapter.new(adapter)
        |> Setup.Admin.new(user: "adm", password: "pass")
        |> Setup.Login.new(user: "adm", password: "pass")
      test "Create", %{setup: setup} do
        db_name = Utils.random_name("db")
        setup_ctx = setup |> Setup.run()
        assert {:ok, resp} = Adapter.create_db(Setup.get(setup_ctx, :adapter), 
db_name)
        assert resp.body["ok"]
      end
    end
  end
end
```

# Using same test suite to compare new implementation of the same interface 
with the old one

Imagine that we are doing a major rewrite of a module which would implement the 
same interface.
How do we compare both implementations return the same results for the same 
input?
It is easy in Elixir, here is a sketch:
```
defmodule Couch.Test.Fabric.Rewrite do
  use ExUnit.Case
  alias Couch.Test.Utils, as: Utils

  # we cannot use defrecord here because we need to construct
  # record at compile time
  admin_ctx = {:user_ctx, Utils.erlang_record(
    :user_ctx, "couch/include/couch_db.hrl", roles: ["_admin"])}

  test_cases = [
    {"create database": {create_db, [:db_name, []]}},
    {"create database as admin": {create_db, [:db_name, [admin_ctx]]}}
  ]
  module_a = :fabric
  module_b = :fabric3

  describe "Test compatibility of '#{module_a}' with '#{module_b}'" do
    for {description, {function, args}} <- test_cases do
      test "#{description}" do
        result_a = unquote(module_a).unquote(function)(unquote_splicing(args))
        result_b = unquote(module_b).unquote(function)(unquote_splicing(args))
        assert result_a == result_b
      end
    end
  end

end
```
As a result we would get following tests
```
Couch.Test.Fabric.Rewrite
  * test Test compatibility of 'fabric' with 'fabric3' create database (0.01ms)
  * test Test compatibility of 'fabric' with 'fabric3' create database as admin 
(0.01ms)
```

The prototype of integration is in this draft PR 
https://github.com/apache/couchdb/pull/2036. I am planing to write formal RFC 
after first round of discussions on ML.

Best regards,
iilyak

Reply via email to