Add InstanceOfPatternMatch recipe to java-version-17.yml

[knutwannheden]

Related: openrewrite/rewrite#2690

[knutwannheden]

@timtebeek Is this the right place to add the recipe?

[knutwannheden]

@timtebeek Also, I don't know if this can be integrated before a new release has been made. Is the CI using the released version or SNAPSHOT builds?

[timtebeek]

I'm not sure of your intentions with the System.out.println in 507f67d;
but I've approved the run just this once; if that's what hacks us well done. ;)

[knutwannheden]

I'm not sure of your intentions with the System.out.println in 507f67d;
but I've approved the run just this once; if that's what hacks us well done. ;)

I just added a test to make sure that the new recipe is also being used. If the System.out.println() call in the test (which of course won't get executed) bothers you, we can of course change it 😄

[knutwannheden]

Looks like the recipe wasn't applied properly. Makes me wonder why the test didn't fail before I modified it. Shouldn't it complain if it can't find the recipe? I could debug that.

[timtebeek]

It should; surprised it didn't. Thanks for troubleshooting either way; it's good to uncover these cross-repository issues this way. Likely wouldn't have registered if it was indeed in the same repository.

[knutwannheden]

It should; surprised it didn't. Thanks for troubleshooting either way; it's good to uncover these cross-repository issues this way. Likely wouldn't have registered if it was indeed in the same repository.

Actually, my test is wrong. I will fix it.

[knutwannheden]

Actually, my test is wrong. I will fix it.

The problem with the test was that the instanceof and type cast expressions were method invocations, which we can't know if they have side effects. So theoretically it could be important that the method gets called twice. That is why the recipe doesn't get applied and I had to extract the expression into a variable first. I had myself forgotten about this safety guard.

I will add an issue for support of IntelliJ's @Contract annotation, which could be used for more precise side-effects calculations. Why Map#get(Object) isn't annotated as @Contract(pure=true) (see https://github.com/JetBrains/intellij-community/blob/master/java/jdkAnnotations/java/util/annotations.xml#L3249-L3254) I am unsure about, so that wouldn't have made a difference in my example.