Do not explicitly require Controller#call method

[gammons]

I found it slightly odd that I needed to always write a .call method in my controllers when that method was empty. So instead I decided to check for the existence of the super method in the parent class first.

[coveralls]

Coverage remained the same when pulling 358710a on gammons:do_not_require_controller_call into 8d39557 on lotus:master.

[jodosha]

@gammons Thanks for this PR. What's the use case? I mean, if you don't need to implement #call why do you need an action in the first place?

[jodosha]

@gammons RE test:

class ActionWithoutCall
  include Lotus::Action
end

it "doesn't require #call implementation" do
  action     = ActionWithoutCall.new
  code, _, _ = action.call({})

  code.must_equal(200)
end

[sidonath]

@jodosha FWIW I find this useful as well since sometimes all logic can happen in callbacks or there's no logic at all (e.g. rendering a dumb template)

[gammons]

@jodosha What @sidonath said ;) I have a landing page that had no need for any controller logic.

Also, thank you for that recommendation for a good test. The action was not raising an error, but rather it was throwing a 500. So I was able to validate my change with a spec, which I added above.

[coveralls]

Coverage remained the same when pulling 787d77e on gammons:do_not_require_controller_call into 8d39557 on lotus:master.

[jodosha]

@gammons @sidonath I've evaluated this issue, if we allow to omit it there are two downsides:

1. Because #call is part of the public interface, future or third part modules (Lotus::Action::Something) should always check if the action respond to it. That potentially leads to several checks in the same request cycle.
2. I registered a degradation of ~100 reqs/s

# Master

% wrk -t 2 http://localhost:9292/                
Running 10s test @ http://localhost:9292/
  2 threads and 10 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency     4.87ms   12.05ms  51.85ms   92.54%
    Req/Sec     3.06k     1.02k    4.33k    85.34%
  57151 requests in 10.00s, 5.01MB read
Requests/sec:   5717.08
Transfer/sec:    513.68KB

# This branch

% wrk -t 2 http://localhost:9292/
Running 10s test @ http://localhost:9292/
  2 threads and 10 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency     5.19ms   14.04ms  66.17ms   93.50%
    Req/Sec     3.01k     0.98k    4.33k    83.94%
  56055 requests in 10.00s, 4.92MB read
Requests/sec:   5605.63
Transfer/sec:    503.66KB

[fuadsaud]

The empty method definition could be easily extracted out to a mixing in your application, right?

[jodosha]

@fuadsaud yes, so if a developer adds that mixin, it would solves the problem that @gammons raised.