When cucumber::test is called, it scans for the .feature files in the features_dir directory.

It parses the feature files and runs the step implementations one by one, according to their order in feature files.

Step execution order

Given such a feature file:

# tests/testthat/sum.feature
Feature: Sum
  Scenario: Sum should work for 2 numbers
    Given I have 1
    And I have 2
    When I add them
    Then I get 3

  Scenario: Sum should work for 3 numbers
    Given I have 1
    And I have 2
    But I have 3
    When I add them
    Then I get 6

And such step definitions:

# tests/testthat/steps/steps_definitions.R
given("I have {int}", function(int, context) {
  print("given")
  context$numbers <- c(context$numbers, int)
})

when("I add them", function(context) {
  print("when")
  context$result <- sum(context$numbers)
})

then("I get {int}", function(int, context) {
  print("then")
  expect_equal(context$result, int)
})

The order of calling the steps would be:

> [1] "given"
> [1] "given"
> [1] "when"
> [1] "then"
> [1] "given"
> [1] "given"
> [1] "given"
> [1] "when"
> [1] "then"

Translation to testthat code

Cucumber translates .feature files to testthat::test_that calls.

When a Scenario is run, the context environment is reset so that the state doesn’t leak to the next scenario.

The previous feature file is basically translated to the following test code, so that we can leverate testthat reporters and not reinvent the wheel.

test_that("Scenario: Sum should work for 2 numbers", {
  context <- new.env()
  context$numbers <- c(context$numbers, 1)
  context$numbers <- c(context$numbers, 2)
  context$result <- sum(context$numbers)
  expect_equal(context$result, 3)
})

test_that("Scenario: Sum should work for 3 numbers", {
  context <- new.env()
  context$numbers <- c(context$numbers, 1)
  context$numbers <- c(context$numbers, 2)
  context$numbers <- c(context$numbers, 3)
  context$result <- sum(context$numbers)
  expect_equal(context$result, 6)
})

Test successes or failures are reported at the level of Scenarios.

A scenario is a test case.

How step definitions are loaded

You don’t have to load step implementations manually. Cucumber loads them automatically when cucumber::test is called.

If you don’t want them to be loaded automatically from the default location, you can use steps_loader argument to provide your own step loader function. See inst/examples/custom_step_loader

Steps from feature files are matched against step definitions defined with given(), when(), and then() functions using regular expressions.

When you define a step by calling any of the step functions, you register that step, making it available for the tests.