Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Don't silently ignore arbitrary method expectations when combining them with 'and_call_original' #382

Merged
merged 3 commits into from Oct 5, 2013

Conversation

@JonRowe
Copy link
Member

@JonRowe JonRowe commented Jul 28, 2013

I had a discussion with @myronmarston about a problem I ran into during the conversion of our RSpec suite to the new expect/allow syntax. The gist of it can be found here.

Funnily what I thought was a bug, never quite worked the way I thought it was and I refactored the spec after reading his suggestions. Myron asked me to also add an issue for it, though. So here we go :-)

The spec we talked about had used the following expectation.

CachedUser.should_receive(:where) do |args|
    expect(args[:id]).to have(3).items
    expect(contact_ids).to include(*args[:id])
end.and_call_original

The values passed to 'where' are sort of random so you can't really setup an argument matcher with 'with'. That's why we tried to verify them in the block. The whole code under test there was also part of an AREL call comparable to this one. That's where the 'and_call_original' came into the game

CachedUser.where(id: random_related_user_ids).all

The spec passed. Though, as Myron pointed out, 'and_call_original' completely replaces the block expectation, resulting in the inner expectations never to be executed.

That was a bit surprising to find out, but to be honest also a fault on my side, since I probably never saw the expectation fail in the first place :-/

To make that behavior more obvious I think it would be good to either raise an exception or to output a warning in that case.

@coveralls
Copy link

@coveralls coveralls commented Jul 28, 2013

Coverage Status

Coverage decreased (-0%) when pulling c05a378 on warn_when_overriding_implementation into 9202615 on master.

@BjRo
Copy link
Author

@BjRo BjRo commented Jul 28, 2013

Awesome, thx!


extend(Mocks::Warning) unless respond_to?(:warning) && respond_to?(:warn_with)
end

This comment has been minimized.

@myronmarston

myronmarston Jul 31, 2013
Member

This file is really small and seems to have a lot in common with lib/rspec/mocks/deprecation.rb. Given that ruby (particularly 1.9) has had well-known perf problems with requires, I think it behooves us to not split things into so many small files. What do you think about combining these into one file?

This comment has been minimized.

@myronmarston

myronmarston Jul 31, 2013
Member

Also, do you think this kind of warning should be printed in the deprecation stream or always to STDERR?

This comment has been minimized.

@JonRowe

JonRowe Jul 31, 2013
Author Member

I've taken care of this in the other PR, and I will rebase off that when we've worked it into something we want to use :)

This comment has been minimized.

@xaviershay

xaviershay Oct 1, 2013
Member

@JonRowe which other PR? Has it been merged yet and can we move forward with this fix?

This comment has been minimized.

@JonRowe

JonRowe Oct 1, 2013
Author Member

Really wishing I had noted the number...

This comment has been minimized.

This comment has been minimized.

@JonRowe

JonRowe Oct 2, 2013
Author Member

Ah hah! Yes that makes sense :)

@@ -22,24 +22,30 @@ def self.new_instance
let(:instance) { klass.new }

it 'passes the received message through to the original method' do
instance.should_receive(:meth_1).and_call_original
allow(instance).to receive(:meth_1).and_call_original

This comment has been minimized.

@myronmarston

myronmarston Jul 31, 2013
Member

Why did you change this from a message expectation to a stub? IMO, and_call_original doesn't really make sense with a stub, given that the and_call_original is only possible on a partial mock, and it would call the original normally anyway....

expect(instance.meth_1).to eq(:original)
end

it 'passes args and blocks through to the original method' do
instance.should_receive(:meth_2).and_call_original
allow(instance).to receive(:meth_2).and_call_original

This comment has been minimized.

@myronmarston

myronmarston Jul 31, 2013
Member

Same here: why the change from a mock expectation?

@@ -116,6 +116,9 @@ def and_call_original
if @method_double.object.is_a?(RSpec::Mocks::TestDouble)
@error_generator.raise_only_valid_on_a_partial_mock(:and_call_original)
else
if implementation.inner_action

This comment has been minimized.

@myronmarston

myronmarston Jul 31, 2013
Member

I think implementation.has_inner_action? would make more sense...thoughts?

@@ -464,6 +467,7 @@ def initial_implementation_action=(action)
end

def inner_implementation_action=(action)
RSpec.warning("You're overriding a previous implementation for this stub") if implementation.inner_action

This comment has been minimized.

@myronmarston

myronmarston Jul 31, 2013
Member

I think RSpec.warn reads better (as its a verb) than RSpec.warning (as warning is a noun). Thoughts?

This comment has been minimized.

@JonRowe

JonRowe Oct 2, 2013
Author Member

This matches up with the stuff in rspec/rspec-core#1024, happy to have that discussion there :)

@@ -116,6 +116,9 @@ def and_call_original
if @method_double.object.is_a?(RSpec::Mocks::TestDouble)
@error_generator.raise_only_valid_on_a_partial_mock(:and_call_original)
else
if implementation.inner_action
RSpec.warning("You're overriding a previous implementation for this stub")

This comment has been minimized.

@myronmarston

myronmarston Jul 31, 2013
Member

I think the "for this stub" language could be a bit confusing when it's a mock expectation and not a stub. Not sure if there's a more generic way to phrase it, though...

@myronmarston
Copy link
Member

@myronmarston myronmarston commented Jul 31, 2013

In @BjRo's gist, it also came up that an expression like this:

      expect(CachedUser).to receive(:where) do |args|
        expect(args[:id]).to have(3).items
        expect(contact_ids).to include(*args[:id])
      end.and_call_original

...raises a confusing error (NoMethodError: undefined method 'and_call_original' for []:Array). It would be good to fix that as well.

@coveralls
Copy link

@coveralls coveralls commented Oct 2, 2013

Coverage Status

Changes Unknown when pulling d8bca94 on warn_when_overriding_implementation into * on master*.

@coveralls
Copy link

@coveralls coveralls commented Oct 2, 2013

Coverage Status

Changes Unknown when pulling 650bee8 on warn_when_overriding_implementation into * on master*.

@coveralls
Copy link

@coveralls coveralls commented Oct 2, 2013

Coverage Status

Changes Unknown when pulling 650bee8 on warn_when_overriding_implementation into * on master*.

@JonRowe
Copy link
Member

@JonRowe JonRowe commented Oct 2, 2013

Ready for a review I guess.

@@ -0,0 +1,31 @@
module RSpec

unless respond_to?(:deprecate)

This comment has been minimized.

@xaviershay

xaviershay Oct 5, 2013
Member

Where else would this be defined? Comment required.

This comment has been minimized.

@JonRowe

JonRowe Oct 5, 2013
Author Member

This whole file is for when the gems are used individually.

This comment has been minimized.

@xaviershay

xaviershay Oct 5, 2013
Member

Otherwise it picks up the one from rspec-core. I get it, think it's unusual enough to deserve a comment.

This comment has been minimized.

@JonRowe

JonRowe Oct 5, 2013
Author Member

Yep I was already writing one ;)

@xaviershay
Copy link
Member

@xaviershay xaviershay commented Oct 5, 2013

Could final commit be squashed into another?

@xaviershay
Copy link
Member

@xaviershay xaviershay commented Oct 5, 2013

Looks good otherwise when build is green.

@JonRowe
Copy link
Member

@JonRowe JonRowe commented Oct 5, 2013

It's awaiting rspec/rspec-core#1024 (to sync the warning stuff)

@coveralls
Copy link

@coveralls coveralls commented Oct 5, 2013

Coverage Status

Changes Unknown when pulling 894dd8b on warn_when_overriding_implementation into * on master*.

@JonRowe
Copy link
Member

@JonRowe JonRowe commented Oct 5, 2013

Actually I think we can merge this now and sync up the warnings stuff later if need be

module RSpec

# We don't redefine the deprecation helpers
# when they already exist (defined by rspec-core etc)

This comment has been minimized.

@xaviershay
Copy link
Member

@xaviershay xaviershay commented Oct 5, 2013

Yeah I don't see a reason to gate this on the core PR. Worst case scenario we just need to come back here and change it again if core PR changes, that's not terrible.

@coveralls
Copy link

@coveralls coveralls commented Oct 5, 2013

Coverage Status

Changes Unknown when pulling 659b039 on warn_when_overriding_implementation into * on master*.

JonRowe added a commit that referenced this pull request Oct 5, 2013
Don't silently ignore arbitrary method expectations when combining them with 'and_call_original'
@JonRowe JonRowe merged commit f434af5 into master Oct 5, 2013
1 check passed
1 check passed
default The Travis CI build passed
Details
@JonRowe JonRowe deleted the warn_when_overriding_implementation branch Oct 5, 2013
@myronmarston
Copy link
Member

@myronmarston myronmarston commented Oct 5, 2013

This looks good, but did we address the other odd error (NoMethodError: undefined method 'and_call_original' for []:Array) I mentioned above ?

@JonRowe
Copy link
Member

@JonRowe JonRowe commented Oct 5, 2013

Hmm, no, but thats a block capture issue... e.g.

expect(user).to receive(:where) do |args|
  expect(args[:id].size).to eq 3
end.and_call_original

produces the weird error

expect(user).to( receive(:where) do |args|
  expect(args[:id].size).to eq 3
end.and_call_original )

Works fine.

@myronmarston
Copy link
Member

@myronmarston myronmarston commented Oct 5, 2013

Ah...I know what's going on: that expression passes the block to to due to ruby's precedence rules. So and_call_original is being called from the return value of to. which is the return value of handle_matcher, which returns whatever matcher.matches? returns, which, in this case, is returning @recorded_customizations, which is an array:

@recorded_customizations.each do |customization|
customization.playback_onto(expectation)
end

So, I think if we changed RSpec::Mocks::Matchers::Recive#setup_method_substitute to return expectation, it would allow further chaining. Want to take a stab at writing a failing spec, adding that, and seeing if it passes?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Linked issues

Successfully merging this pull request may close these issues.

None yet

5 participants