Enable recursive updates to NamespacedHierarchicalStore

[mpkorstanje]

Users of the extension context may occasionally want to build a non-trivial object graph, of which the individual components are also stored in the extension context store. This naturally leads to a pattern where the store is updated recursively. For example:

store.getOrComputeIfAbsent(("a", __ -> {
  B b = store.getOrComputeIfAbsent("b", _ -> new B(), B.class);
  C c = store.getOrComputeIfAbsent(("c", _ -> new C(), C.class);
  return new A(b, c);
}, A.class);

While the backing ConcurrentHashMap does not support recursive updates this does work as getOrComputeIfAbsent defers the recursive update until the after the map has been updated. This is not the case for computeIfAbsent which immediately calls the default creator. This poses a problem as computeIfAbsent is supposed to be a drop-in replacement for getOrComputeIfAbsent with better null and exception semantics.

The solution is to defer the execution of the default value creator until after the map has been updated. This is handled by the DeferredSupplier which uses a FutureTask to support a nice separation of compute and get operations.

Because there are some differences between both compute methods in how deferred values are to be handled, this PR introduces a StoredValue.Value, .DeferredValue and .DeferredOptionalValue containers for put, getOrComputeIfAbsent and computeIfAbsent respectively. This allows:

• DeferredOptionalValue containing an exception to be treated as absent in getStoredValue.
• DeferredOptionalValue containing a a value to return its contents on evaluation.
• DeferredValue to throw an exception or returns its contents on evaluation
• Value to simply return its contents on evaluation

This PR does introduce a small change in behavior. Previously any of the concurrent callers of getOrComputeIfAbsent could be the one executing the default creator. This can now only be done by the caller that created the stored value that was actually inserted. I don't expect this will break anything.

The PR is also slightly bigger than it needs to be. StoredValue.Value could be replaced with .DeferredValue. But as DeferredValue only exists to support a deprecated method, this makes for easier clean up in the future.

Thanks to @martinfrancois who contributed the CollidingKey, DeferredSupplier and several other useful concepts in #5209.

Fixes: #5171

---

I hereby agree to the terms of the JUnit Contributor License Agreement.

---

Definition of Done

• There are no TODOs left in the code
• Method preconditions are checked and documented in the method's Javadoc
• Coding conventions (e.g. for logging) have been followed
• Change is covered by automated tests including corner cases, errors, and exception handling
• Public API has Javadoc and @API annotations
• Change is documented in the User Guide and Release Notes

[testlens-app]

🔎 No tests executed 🔎

🏷️ Commit: 5493055
▶️ Tests: 0 executed
🟡 Checks: 2/12 completed

---

Learn more about TestLens at testlens.app.

[mpkorstanje]

@martinfrancois would you have time to review this?

[martinfrancois]

@martinfrancois would you have time to review this?

Sure @mpkorstanje, I'll have a look!

[marcphilipp]

Thanks for codifying the concepts so clearly! 👍