Deeply nested contexts can be confusing. I wouldn't use any more than 2 levels.

I agree with this best practice, but I feel there should also be example code of how to set up the context. The context method doesn't do anything by itself, just makes it more readable.

I agree with @andyw8 and disagree with the rule.
If there would be only one it block inside the context block, it reads much easier with using one it instead of using context and it.
It also requires you to setup the conditions in a before block which is another two lines just for before do and end

I disagree with using "should" in describe blocks.

Do or do not - there is no "should".

Edit: Wait - you already have this guideline here.

Why is it not used in this example (or others consistently)?

This particular example confuses me.

it 'has 200 status code if logged in' do
  response.should respond_with 200
end

=>

context 'when logged in' do
  it { should respond_with 200 }
end

My concern is that response has magically dissapeared. Where is it?

What's the best way to go about setting up the context? Should I just be using before? In the example, the context is magically set up.

It should be mentioned somewhere that you need shoulda included as gem for this matchers to actually work

@kikito response is the subject. Since this would be in a controller test, it gets assigned as the subject.

I agree that the comment by @Balauru should be incorporated. Also the clarification by @warmwaffles would be nice too. These things may be obvious for experience Rails hackers but they aren't clear for a newbie.

@russellsilva essentially in the background this happens with controller tests:

subject { response }

If you wish to be more explicit, then by all means, use response

I really fail to see how one is any better than the other. You're adding extra syntax and work for the program just to save yourself a few characters? Also forgot to mention that shoulda isn't part of RSpec and therefore it {should.. doesn't work in straight RSpec. So now I have to load a whole 'nother gem just to save a few extra characters?

Most of these tips for writing more clear specs are appreciated, and make a lot of sense. But this one in particular struck me as a little snake oily.

I am new to rspec. First I type a good solution and in result I have obvious for experienced users 'undefined method context'. Where is application of 'devoted to how to create a great RSpec test suite'? I wish to find reliable resource.

How are nested contexts supposed to be done?

context 'when something is set' do
  context 'and when another thing is set' do
  end
end

is it proper to use and in the description?

I usually don't write "and" in my context descriptions, but if I could alias context "when #{message}" to when, I would :)- T

On Thu, Jan 9, 2014 at 1:15 PM, Matthew Johnston notifications@github.com
wrote:

How are nested contexts supposed to be done?

context 'when something is set' do
  context 'and when another thing is set' do
  end
end

is it proper to use and in the description?

Reply to this email directly or view it on GitHub:
#3 (comment)

You can't use when with a lower case, however you can alias context to be When, however I don't think this is a good idea.

Yeah, I realized when was a reserved word almost immediately after I typed that. My bad.

The switch from inline matching to custom matchers from the shoulda gem in the before/after examples are extremely misleading for those of us new to RSpec.

Personally, I stay away from shoulda matchers. They are confusing and need to be avoided.

This is inconsistent with http://betterspecs.org/#expect

It should be updated to follow the other guideline.

As a one liner syntax, I would rather recommend is_expected.to instead of should. While a little more verbose, it would keep the logic more consistent.
Cf. https://www.relishapp.com/rspec/rspec-core/v/3-0/docs/subject/one-liner-syntax

For this particular topic (and perhaps others), I would like to see what rspec prints out. That would make me better able to evaluate the idea myself. Often I write specs and then rewrite them because the output when the test is run is useless.

Is it really okay to have nested context blocks?

@xirukitepe I think it depends on your workflow and preference. At work, we use Capybara integration tests. Capybara DSL uses feature in place of context. Both work, but it just adds readibility to the spec. For us, a feature is a page (pretty much the controller used) in the app. I then create sub-contexts, like: context "SAD PATHS", because we do not use the it {should not} style. This way we can read the output and know if a test is for happy or sad paths

Do know, if you run random specs, they will be randomized as follows:

1. By spec file
2. By context (or feature or whatever DSL you use)
3. By examples within the context

This means, if you have 2 contexts (one saying "Users page", and another saying "sad paths"), what will be randomized is:

1. Spec file
2. top level context
3. second level context
4. (and so on)
5. examples

How about amending:

When describing a context, start its description with "when" or "with"

... to ...

When describing a context, start its description with "when", "with", or "without"

Not so hot (in my opinion)

context 'with sprinkles' ...
context 'when there are no sprinkles on top' ...

Better (again, in my opinion)

context 'with sprinkles' ...
context 'without sprinkles' ...

🍦 🍨 🍧

@nagi I like your idea. How about submitting a pull request with the change?

@onebree Great, thanks! #157

I'm confused what this particular guideline is about: The good/bad example both adds the context and uses an anonymous it block.

+1 @warmwaffles about writing nested context with context 'and when...'.

How are nested contexts supposed to be done?

context 'when something is set' do
  context 'and when another thing is set' do
  end
end

is it proper to use and in the description?

I miss "but" even more. Consider this example:

context "when we're looking at a banana" do
  subject { Banana.new }

  context "when it is green" do
    subject { Banana.new(color: 'green') }

    it { is_expected.to be_good_for_storage }
  end

  context "when it is yellow" do
    subject { Banana.new(color: 'yellow') }

    it { is_expected.to be_good_to_eat }

    context 'but it has black dots' do
      subject { Banana.new(color: 'yellow', dots: true) }

      it { is_expected.to be_good_for_milk_shake }
    end
  end
end

Even if betterspecs doesn't suggest to use a lot of nested contexts: people are doing it. So let's assume that there are many nested contexts in reality:

Contexts are really helpful for branching, but they could be much more helpful if we could mark edge cases explicitly. I'd propose to suggest "but" in nested contexts for edge cases.

How are nested contexts supposed to be done?

context 'when something is set' do
  context 'and when another thing is set' do
  end
end

is it proper to use and in the description?

I miss "but" even more. Consider this example:

context "when we're looking at a banana" do
  subject { Banana.new }

  context "when it is green" do
    subject { Banana.new(color: 'green') }

    it { is_expected.to be_good_for_storage }
  end

  context "when it is yellow" do
    subject { Banana.new(color: 'yellow') }

    it { is_expected.to be_good_to_eat }

    context 'but it has black dots' do
      subject { Banana.new(color: 'yellow', dots: true) }

      it { is_expected.to be_good_for_milk_shake }
    end
  end
end

Even if betterspecs doesn't suggest to use a lot of nested contexts: people are doing it. So let's assume that there are many nested contexts in reality:

Contexts are really helpful for branching, but they could be much more helpful if we could mark edge cases explicitly. I'd propose to suggest "but" in nested contexts for edge cases.

Interesting. I really dislike "and" in nested context but I've seen a similar use case to "but ..." with "except ..."

How about guiding describe vs context and change of test state, e.g.

context "when logged in" do
  before(:each) { log_in_as_new_user }

  it "lets you through" 
   ... 
end 

context "when logged in" do
  it "will be annoying if the test case doesn't log you in"
  it "may be confusing to write more examples in this context, is the state is set or not"
  ... 
end 

describe "when logged in" do
  it "seems more likely that the state isn't set up already by the block" 
   ... 
end 

A simple rule might be "use describe if you aren't adding any let! or before code to implement the context", but I'm not sure it's a good idea. One problem I've noticed is it encourages let spaghetti hell by preventing self-contained test cases. But I'm wondering if it's better to group those with describe instead of context, since no context is set up, even if the description strings end up the same.