Back to TILs

Run Tests in Vim

Date: 2018-11-19Last modified: 2022-10-07

Test driven development thrives on a tight feedback loop but switching from the editor to the shell to manually run specs is inefficient.

Tools such as autotest and guard run specs whenever a file gets saved. Although an improvement over a manual workflow, those approaches often run the suite when not needed and run too many or too few specs.

Enter vim-rspec, a lightweight Vim plugin that runs specs directly from within Vim with the press of a key.

It exposes methods such as RunNearestSpec(), RunCurrentSpecFile(), and RunLastSpec(), which can be bound to a key mapping of your choice. In thoughtbot/dotfiles, we bind those methods to <Leader>s, <Leader>t, and <Leader>l.

Cursor over any line within an RSpec spec like this:

describe RecipientInterceptor do
  it 'overrides to/cc/bcc fields' do
    Mail.register_interceptor RecipientInterceptor.new(recipient_string)

    response = deliver_mail

    expect(response.to).to eq [recipient_string]
    expect(response.cc).to eq []
    expect(response.bcc).to eq []
  end
end

Type <Leader>s:

rspec spec/recipient_interceptor_spec.rb:4
.

Finished in 0.03059 seconds
1 example, 0 failures

The screen is overtaken by a shell that runs only the focused spec. Developers using tmux with vim-rspec and tslime sometimes send the output to a nearby shell so the code and spec output display on the screen at the same time.

Feeling good that this new spec passes, run the whole file’s specs with <Leader>t to make sure the class’s entire functionality is still intact:

rspec spec/recipient_interceptor_spec.rb
......

Finished in 0.17752 seconds
6 examples, 0 failures

Red, green, refactor. From within the application’s or library’s code:

def delivering_email(message)
  add_custom_headers message
  add_subject_prefix message
  message.to = @recipients
  message.cc = []
  message.bcc = []
end

Run <Leader>l without having to switch back to the spec:

rspec spec/recipient_interceptor_spec.rb
......

Finished in 0.17752 seconds
6 examples, 0 failures

These tight feedback loops make Test-Driven Development easier by eliminating the switching cost between editor to the shell when running specs.