Behaviour driven development framework for testing Vim plugins written in Riml.

For people unfamiliar with Riml. Riml is a programming language that compiles to Vimscript. It's constructs are very similar to languages like Ruby and Coffeescript. It also provides a layer of abstraction to write OOP code that is converted to Vimscript's Funcref prototype chains.

Speckle uses Rimls object-oriented constructs to provide a BDD testing framework for Riml. You can use it for both unit testing and functional testing. Using Riml and Speckle you can write Vim plugins using common OOP software idioms and have the same familar tools available in your development workflow.

Speckle is both a test compiler and test execution engine. It does compilation of the _spec.riml specs with Riml and runs them in a Vim instance, capturing assertions, stacktraces and errors and reports them back after closing the launched vim instance.

$ gem install speckle

Here's an example using Speckle.

riml_include 'dsl.riml'

class MyFirstSpec
  defm describe
    return 'My First Spec'
  end

  defm it_can_check_equality_of_strings
    my_string = 'v' . 'i' . 'm'
    expect(my_string).to_equal(vim)
  end
end

A test example is a Riml class with a test method that begins with it. You can optionally provide a describe method to change the displayed name for test in the reporter output.

Note the riml_include dsl.riml. This include provides Speckle's expect dsl for use inside the test examples. The path to this include is autoconfigured by Speckle. For including your own classes specify the path like -I lib. Where lib is the directory containing your Vim plugin .riml files.

Speckle looks for tests in the spec folder by default. The tests should be named in the form {name}_spec.riml. The above example could be saved as spec/my_first_spec.riml. To run this test use,

$ speckle

If the test succeeded you will get a message like,

✓ 1 tests completed (5ms)
Passed: 1, Failures: 0, Errors: 0, Assertions: 1

If you change the input string to only vi and run it again you will see a message like,

EqualityMatcher #it can check equality of strings
    AssertionError: expected “vi” to equal “vim”


✖ 1 tests completed (5ms)
Passed: 0, Failures: 1, Errors: 0, Assertions: 0

expect(actual).to_equal(expected)  # passes if actual == expected
expect(actual).to_not_equal(expected) # passes if actual != expected

Note: A type mismatch will be thrown when comparing values with different types.

expect(actual).to_be_gt(expected) # -ve to_not_be_gt
expect(actual).to_be_gte(expected) # -ve to_not_be_gte
expect(actual).to_be_lte(expected) # -ve to_not_be_lte
expect(actual).to_be_lt(expected) # -ve to_not_be_lt
expect(actual).to_be_within([delta, expected]) # -ve to_not_be_within
expect(actual).to_be_between([min, max]) # -ve to_not_be_between

expect(actual).to_match(pattern) # -ve to_not_match

expect(actual).to_be_true(expected) # -ve to_be_false
expect(actual).to_be_ok(expected) # -ve to_not_be_ok

expect(actual).to_exist()  # passes if exists(actual) is not false
expect(actual).to_not_exist() # -ve

expect(actual).to_have_key(expected) # passes if actual is a dict with key expected
expect(actual).to_not_have_key(expected) # -ve

expect(actual).to_have_length(expected)  # passes if len(actual) == expected
expect(actual).to_not_have_length(expected) # -ve

In addition to the above default matchers you can create custom matchers easily.

Consider a Person class in Riml.

class Person
  def initialize(name)
    self.name = name
  end

  defm get_name()
    return self.name
  end
end

To write matcher to check if a person has the correct name we can write a PersonNameMatcher. The class must implement the methods, match, failure_message_for_match and failure_message_for_mismatch as shown below.

class PersonNameMatcher
  defm match(expected, actual)
    self.result = actual.get_name()
    return self.result == expected
  end

  defm failure_message_for_match(expected, actual)
    return "expected person name to be “#{expected}” but was “#{self.result}”"
  end

  defm failure_message_for_mismatch(expected, actual)
    return "expected person name to not be “#{expected}” but was “#{self.result}”"
  end
end

Then inside your spec you need to register this matcher using define_matcher,

matcher = new PersonNameMatcher()
define_matcher('to_have_name', 'to_not_have_name', matcher)

Now in a test you can use the custom matcher methods, to_have_name, and it's negative to_not_have_name.

defm it_can_check_for_persons_name
  expect(self.person).to_have_name('john')
end

defm it_can_check_for_negation_of_persons_name
  expect(self.person).to_not_have_name('foo')
end

Speckle supports before, before_each, after and after_each hooks that will be run before/after the entire suite or before/after every test.

defm before
end

defm before_each
end

defm after
end

defm after_each
end

Speckle provides a logger that captures output messages without halting execution of tests. This allows development of vim plugins without needing to use echomsg extensively.

The logger api is,

get_logger().log(msg, ...)
get_logger().info(msg, ...)
get_logger().warn(msg, ...)
get_logger().error(msg, ...)
get_logger().debug(msg, ...)

When running the tests the error messages are shown inline like below.

LoggerSpec
  ✓ it logs a message

      log: Hello World
      log: A test warning
      log: An error warning

Another useful feature of Speckle is it displays stacktraces in a meaningful form to help with debugging.

The example below has a call to an undefined function.

defm it_has_unknown_function
  CallFooFunction()
end

When this test is run the following stacktrace will be shown.

VariousErrorsSpec #it has unknown function
    Vim(call):E117: Unknown function: <SNR>144_CallFooFunction
       at <SNR>143_s:Speckle_run
       at <SNR>143_s:Runner_start
       at <SNR>143_s:SpecRunner_start
       at <SNR>144_s:VariousErrorsSpec_it_has_unknown_function, line 2

Speckle integrates with Vim's built-in profiler. Use the -p or --profile switches to turn on profiling. When profiling is turned on, Speckle prints a summary of the Total Time and Self Time of functions executed in your tests as shown below,

------------------------------------------------------------------------------
PROFILE SUMMARY
------------------------------------------------------------------------------

FUNCTIONS SORTED ON TOTAL TIME
count  total (s)   self (s)  function
  111   1.305531   0.003775  <SNR>212_s:CommandRegistry_run_command()
  605   1.268269   0.008920  <SNR>181_expect()
   98   1.221755   0.002080  <SNR>212_s:CommandRegistry_run_action()
   81   1.068915   0.001752  <SNR>222_s:CommandRegistry_run_action()
   81   1.066899   0.002717  <SNR>222_s:CommandRegistry_run_command()
    9   0.893186   0.000369  <SNR>222_s:GetFileCommandSpec_verify_gf()
   42   0.827668   0.000382  <SNR>212_s:Controller_process()
   41   0.764384   0.000386  <SNR>222_s:Controller_process()

The full profile containing further details like per-line execution times can be viewed in the file build/speckle.profile.

This option works with all the isolated testing options below. Profiling can be used alongside --grep and --tag to only profile specific tests.

Speckle can be made to run only specific tests using the --grep option. For example, to only run tests in the spec/models folder use

    $ speckle --grep spec/models

Further to only run specific tests we can tag the test with a keyword, and then use --tag keyword to run only those tests. Tagging is done by adding a _tagname suffix to the test method name. For example to tag a spec with the tag perf, you would use,

defm it_will_work_perf
end

And run the test with,

    $ speckle --tag perf

Both these flags can be combined to limit the tests run.

$ speckle [options] [file(s) OR directory]
-a, --all                        Compile and run tests (default)
-t, --test                       Only run tests
-c, --compile                    Only compile tests

Options:

-I, --libs <libs>                Specify additional riml library path(s)
-g, --grep <pattern>             Only run tests matching the pattern
-i, --invert                     Inverts --grep matches
    --tag <tag>                  Only run tests matching tag
-r, --reporter <reporter>        Specify the reporter to use (spec, min, dot, tap, fivemat)
-b, --bail                       Bail on first test failure
-w, --watch                      Watch tests for changes
-m, --vim <vim>                  Vim program used to test, default(vim)
-s, --slow-threshold <ms>        Threshold in milliseconds to indicate slow tests
-p, --profile                    Profiles the tests run
-k, --skip-vimrc                 Does not load ~/.vimrc file
-C, --no-colors                  Disable color output
-v, --verbose                    Display verbose output
-D, --debug                      Display debug output
-V, --version                    Print Speckle version
-h, --help                       Print Speckle help

1. Speckle uses git flow based branching model.
2. Pull requests should go against the develop branch.
3. Try to include failing tests if possible.

MIT License. Copyright (c) 2013 Darshan Sawardekar.