Add #unstub method, replace deprecated watir-webdriver with watir

[bwilczek]

In some test cases, it is required to dynamically decide which requests should be stubbed, and which not. Using proxy.reset is not the best way to it since it provides no control over which stubs are to be removed.

This PR adds a new method: proxy.unstub, which acts as a reverse to proxy.stub, and removes only a single stub.

See spec/lib/billy/handlers/stub_handler_spec.rb:66 for examples.

Additionally, this PR removes dependency watir-webdriver. That gem was outdated, deprecated and in fact already has already been removed from GitHub. It is now a part of watir gem, and this one has been added instead.

watir-webdriver was throwing a ton of warnings (Fixnum) and an exception on just require 'watir-webdriver' making work on this change impossible.

[ronwsmith]

@bwilczek thanks for the contribution! Looks like the watir upgrade requires us to finally drop support for ruby 1.93. Mind updating the travis config to remove that and add in the more recent ruby versions?

[ronwsmith]

@bwilczek can you fix up the conflicts here? I just merged a couple smaller PRs in preparation for this breaking change.

[bwilczek]

@ronwsmith
done: fork freshly rebased onto master, updated list of supported ruby versions.

[AlanFoster]

Should unstub be given a previously created stub, rather than providing the values to search for one? 🤔

stub = proxy.stub(...)
proxy.unstub(stub)

[AlanFoster]

@bwilczek I wonder if it's easier to conditionally add stubs to the proxy, rather than conditionally remove stubs? I'd be interested to see an example scenario of this new api being applied :)

[bwilczek]

@AlanFoster Here's an example :)

Let's consider a suite of WebDriver tests running in the same browser.
The application under test consists of JS-heavy frontend, and few backend services.
One of them is search, which is slow and heavy, and for most of the scenarios
requests to it are stubbed out with Puffing Billy.

But, there are some tests (tagged search) which verify integration with the real service:

def stub_seach
  Billy.proxy.stub("#{SEARCH_HOST}/query/...").and_return(
    {some: 'fake objects'}
  )
end

def unstub_search
  Billy.proxy.unstub("#{SEARCH_HOST}/query/...")
end

RSpec.configure do |c|
  c.before(:suite) do
    stub_seach
  end

  c.around(:each, search: true) do |example|
    unstub_search
    example.run
    stub_search
  end
end

Not sure if the API suggested in one of the previous comments would me more elegant though:

def stub_seach
  @search_stub = Billy.proxy.stub("#{SEARCH_HOST}/query/...").and_return(
    {some: 'fake objects'}
  )
end

def unstub_search
  Billy.proxy.unstub(@search_stub)
end

RSpec.configure do |c|
  c.before(:suite) do
    stub_seach
  end

  c.around(:each, search: true) do |example|
    unstub_search
    example.run
    stub_search
  end
end

Hope the example is clear.

[AlanFoster]

@bwilczek Interesting! Won't Puffing Billy's default rspec configuration cause you issues?

puffing-billy/lib/billy/init/rspec.rb

Lines 9 to 15 in 4468e23

	RSpec.configure do |config|
	config.include(Billy::RspecHelper)
	config.after(:each) do
	proxy.reset
	end
	end

Although - I guess you're not making use of that, if you're referencing Billy.proxy directly? 🤔

[bwilczek]

@AlanFoster
Yes, that's my case: we're explicitly setting Billy as the HTTP(S) proxy for the browser in our test suite.

And regarding the API:

stub = proxy.stub(...)
proxy.unstub(stub)

the more I think about it the more I like this approach, I'll test it and implement it if practical.

[AlanFoster]

@bwilczek For that scenario what are your thoughts on something like:

require 'billy/capybara/rspec'

def stub_seach
  Billy.proxy.stub("#{SEARCH_HOST}/query/...").and_return(...)
end

RSpec.configure do |config|
  config.before(:each, type: :feature) do |example|
    stub_search unless example.metadata.key? :with_real_search_functionality
  end
end

That would give you the functionality you're after at the minute I think 🤔

[bwilczek]

@AlanFoster Your example would work as long as all stubs are reset after each scenario. Given that we use Billy as a regular browser proxy once the stubs are set they will remain there and affect the consecutive tests. In fact, we use cucumber for this, I used rspec here as an example because it's much nicer :)

BTW: I've updated the unstub method to accept the stub (not URL/method anymore) as a parameter.

[ronwsmith]

@bwilczek @AlanFoster I haven't been following your back and forth closely. Is this PR ready for review, or are you still working through some ideas?

[bwilczek]

@ronwsmith It is done from my perspective, so as long as @AlanFoster is OK with the changes the PR is ready for review/merge.

[AlanFoster]

[ronwsmith]

Released with 1.0.0

[AlanFoster]

@ronwsmith / @bwilczek Was this version change intentional? 🤔

[ronwsmith]

Maybe watir required it?

[bwilczek]

Yes, I believe once I changed watir-webdriver to watir I had to change versions of certain other gems - it wouldn't work otherwise. But I don't remember the exact details now.