Consider using Gherkin for tests

[pjoe]

Just wondering if it could make sense to use Gherkin for some tests.

https://cucumber.io/docs/gherkin/reference/

Maybe using tables and outlines could be useful:

https://github.com/b-yond-infinite-network/jest-cucumber-fusion/blob/master/docs/GherkinTables.md

https://github.com/b-yond-infinite-network/jest-cucumber-fusion/blob/master/docs/ScenarioOutlines.md

[pjoe]

Found this for rust: https://github.com/bbqsrc/cucumber-rust

[kvark]

Thank you for filing this! It's great to see a native Rust implementation, this would be a non-option without it :)

Is there an elevator pitch on how we'd benefit from it, and what the downsides are? So far, it looks like a meta-language for tests. Is there a benefit of introducing this meta-language instead of just writing it in Rust to begin with?

[pjoe]

I'm sure there is a better pitch to be found somewhere, but for me I've found gherkin to be useful for these reasons:

• Common language across requirements, tasks, tests, bugs

• Gherkin features are programming language agnostic, but you still have to implement the steps

• It provides a standard way to do parameterized tests using tables and scenario outlines. Guessing that could be useful here for running variations of the same test, e.g. for different types: vec2, vec3, vec4, f16, etc. See some example of this:

  • https://github.com/bbqsrc/cucumber-rust/blob/master/features/test.feature#L12
  • https://github.com/b-yond-infinite-network/jest-cucumber-fusion/blob/master/docs/ScenarioOutlines.md
  • https://github.com/b-yond-infinite-network/jest-cucumber-fusion/blob/master/docs/GherkinTables.md

That being said, I'm not sure it is always beneficial for use in unit tests.

Anyway just an idea for consideration :)

See also:

• https://en.wikipedia.org/wiki/Cucumber_(software)
• https://en.wikipedia.org/wiki/Behavior-driven_development