jrwdunham / tegere Goto Github PK

Imagine referencing spec-defined data structures in the Gherkin feature and describing, in human-readable language, the properties or constraints that must hold when certain user actions occur.

Within a step, you should be able to call execute-steps and supply a string of step texts that should be executed.

It would be nice to be able to run a testing system built on TeGere from the command-line with faster feedback. Of course, in Clojure we can execute Gherkin features from the REPL. However, consider the case where you want to test that the CLI for your TeGere system works correctly because you are going to run it in a CI workflow. It's not nice to wait 5-10 seconds between running clj -m vbit.core --tags=... and seeing output.

Options:

• Convert the code to CLJC and use the NodeJS runtime to execute features from the command line.
• Get a TeGere-based system working with GraalVM. I have put some effort into this and so far have failed. Using spec.alpha in a Clojure app seems to break Graal native-image compilation.
• Make TeGere runnable using Babashka? Is this even possible given its dependencies?

We want to know how long it took to run all matching features/scenarios (i.e., tests).

The grammar should be expanded to recognize these constructs. See https://cucumber.io/docs/gherkin/reference/.

You should be able to write scenarios like the following, but currently you cannot:

  Given a monkey
  #When I give him a banana
  Then he is happy
  But he doesn't eat it
  And he looks at me quizzically

I'm finding your parser very useful, thank you. There is a thing I'll want to add, and that is the ability to read Background sections. If you have advice -- I've written a modest amount of Clojure, but I've read very little -- let me know and I will try and submit code.

Potentially also allow callers to redirect stdout to a logger. (Though that responsibility might better be left to the external calling system.)

It should be possible to add tabular (or freeform string) data to arbitrary Gherkin steps. This feature reduces the probability of ad hoc hidden configuration in the Clojure step logic and/or repeated values in a Scenario Outline Examples table.

It is tedious to have to manually connect each step text/definition, e.g., "Given app has a resource with attribute value", to its execution semantics, e.g., (confirm-existence-of-or-create context resource :attribute value). It should be possible to devise a "step grammar" (CFG) that parses step texts (instead of using simple regular expressions) and maps them to expected function names and definitions. This could be a real "value add" to Gherkin from the Clojure/Instaparse side of things.

Imagine this step:

When the button is pressed repeatedly with a 2 second delay

and these step functions:

(When "the button is pressed {method}" (fn [ctx method] ...))
(When "the button is pressed {method} with a {delay} delay" (fn [ctx method delay] ...))

Currently, both step functions have regexes that will match the step. However, the one that is chosen is arbitrary (determined by hashmap key ordering, I believe).

Some more principled method should be implemented for determining a single match. I propose, at least for the interim, that we simply choose the step with the longest regex, i.e., the second in the above example.

There might be benefits to parsing the Gherkin DSL to an in-memory data structure that is queryable via Datalog. For example, this may make filtering scenarios based on user-supplied and/or flags easier as well as facilitate the counting of which scenarios passed, failed, were skipped, etc. See https://github.com/jpmonettas/clindex for ideas.