Skip to content

Latest commit

 

History

History
33 lines (20 loc) · 2.52 KB

Output.adoc

File metadata and controls

33 lines (20 loc) · 2.52 KB
Raw

Output

By default, Test Bench emits the prose supplied to context and test blocks as well as the structure of the test scripts themselves.

Test Script Structure

Test Bench tests are structured with context and test, which can be nested similarly to describe, context, specify, and it in RSpec.

In general, the goal of a context block is to provide contextual information about a test so that the reader of the test output may better understand the what is being tested and why. For this reason, a context is always supplied prose — a context block in Test Bench will not accept a class the way RSpec allows describe SomeClass. The structure of a test script should describe behavior that is implemented by some part of the system, and never the implementation itself. This is a subtle but significant departure from RSpec test structures that commonly describe the implementation itself, both through class names supplied to describe and the common convention of method names being used to structure the test script underneath the top level describe block.

The ideal output when running a test script with Test Bench communicates the behavior produced by the test subject. It is common and even desirable to write more than one test script to cover a single class. For instance, a test script for a cache might include a test script for demonstrating a cache hit, and another test script for demonstrating a miss. Even though a class is responsible for a single concern, there are inevitably nuances and edge cases that must be each described and tested independently.

Comments

To provide additional output beyond the prose supplied to context and test, comment can be used to write arbitrary text to the test output. This is useful, for instance, to present data that is about to be used in an assertion — if the assertion fails, the user may be able to read the comment to understand why:

test "Fibonacci" do
  result = fib(3)

  comment "fib(3) = #{result.inspect}"

  assert result == [1, 1, 2]
end

Verbosity

When using the bench executable, the verbosity level of the output can be controlled. In quiet mode, the output will be extremely succinct and merely summarize the entire execution and show any errors or test failures. This is suitable for a CI environment. Verbose mode includes additional information beyond what is included in normal output. The next section explains the bench executable in greater detail.