What is the policy for contributions using java based tests instead of groovy?

[rferreira]

Hello folks

I'm a small code contributor for micronaut (mostly views) and I'm planning a largish contribution and I feel pretty confident that I would be able to write more robust testing if I could write them in good old java/junit but all testing appears to be done in groovy/spock which seems like a fine combination but not what I'm most productive with. It's not a big deal if the answer is "no" and all tests must be written in groovy/spock, I'm just looking for confirmation.

Thanks!

[graemerocher]

We have several test with JUnit 5 and Java as well, no problem using it

[ilopmar]

He was explicitly asking about micronaut-views.

Even if we have JUnit5 tests in core, I think we should have more consistent with the tests. What we should have probably done is what Sergio proposes. If we received a PR with JUnit5 we help the user to convert them to Spock (or we do it before we merge the PR).
If we decided to use Spock it was for some reasons: easy to read/write tests, more powerful, full use of Groovy for testing,... so we should stick with it. If not, we should have chosen JUnit (no please!)

[graemerocher]

I standby my point that we are just limiting contributions. If we are going to have a policy of rejecting PRs for being in JUnit 5 I strongly disagree with that and believe we will have to take the burden on of porting the tests to Spock for the user and not push back and ask them to learn a new language just to write tests. People shouldn't be expected to learn Groovy because of silly personal preferences.

[ilopmar]

As I said, yes, we should have ported the tests to Spock in the same PR before merging it.
I don't think using Spock is a silly personal preference but a choice we (in this case you) made when starting the framework (and the right one from my point of view). The same as using Gradle instead of Maven.

[graemerocher]

Absolutely, I love Spock and think it is the best 🙂

At the same time I appreciate that Java is 70% of our user base and Groovy is single digits and the Micronaut codebase is 99% written in Java (excluding tests). I am also empathetic now to folks who have not used Groovy or Kotlin.

Also we are not talking about some monstrosity here, JUnit 5 is a vast improvement to JUnit 4 and actually quite pleasant in many ways.

In my experience as well there are cases where you simply have to use Java to write tests for a Java API because Groovy doesn't allow you to test how a Java API works in terms of generics, lambdas etc.

It will also not be possible to effectively write native image tests with Spock with the upcoming GraalVM support so I anticipate there will be more and more JUnit tests around in modules.

In addition if a subproject has correct documentation then it must have a mix of Spock and JUnit 5 anyway to support multi language documentation.

So I don't agree with a blanket ban on JUnit 5 tests and stand by that.

[jameskleeh]

We should never reject a PR because it contains Junit5 tests. I think it's up to the person who has taken primary responsibility of that module how they want to proceed. To me there are 2 choices, either merge the PR as is, or take the task on to convert them for the contributor.

From my perspective if the PR has to include dependencies to Junit to get the tests to pass I would rather switch the tests to Spock. If the module already has Junit tests then I likely wouldn't bother.

We shouldn't put the responsibility of doing the conversion on the submitter in order to get the PR merged in any case.