feat: support isolated API instances by marcozabel · Pull Request #1928 · open-feature/java-sdk

marcozabel · 2026-04-16T15:00:39Z

This PR

This PR introduces support for creating isolated OpenFeature API instances, each with their own providers, hooks, context, and event handling - enabling multi-tenant or side-by-side usage without shared global state.

Related Issues

Fixes #1925

Notes

Follow-up Tasks

How to test

gemini-code-assist

Code Review

This pull request introduces support for isolated OpenFeature API instances by adding a createIsolated method to OpenFeatureAPI and a new OpenFeatureAPIFactory in a dedicated package. These changes allow for independent state management (providers, hooks, context) across different instances, satisfying specification requirements. Feedback includes a necessary fix for a compilation error in the tests due to unhandled exceptions, a recommendation to move the shared process-wide lock to an instance level for better performance isolation, and a concern regarding the completeness of global state restoration in the test cleanup logic.

Signed-off-by: marcozabel <marco.zabel@dynatrace.com>

codecov · 2026-04-16T15:45:20Z

Codecov Report

❌ Patch coverage is 91.17647% with 3 lines in your changes missing coverage. Please review.
✅ Project coverage is 93.22%. Comparing base (c69bb0e) to head (3e7374c).
⚠️ Report is 5 commits behind head on main.

Files with missing lines	Patch %	Lines
...c/main/java/dev/openfeature/sdk/EventProvider.java	85.71%	0 Missing and 2 partials ⚠️
.../main/java/dev/openfeature/sdk/OpenFeatureAPI.java	92.85%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@             Coverage Diff              @@
##               main    #1928      +/-   ##
============================================
+ Coverage     92.25%   93.22%   +0.97%     
- Complexity      653      662       +9     
============================================
  Files            59       59              
  Lines          1589     1610      +21     
  Branches        179      180       +1     
============================================
+ Hits           1466     1501      +35     
+ Misses           76       62      -14     
  Partials         47       47

Flag	Coverage Δ
unittests	`93.22% <91.17%> (+0.97%)`	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

chrfwow · 2026-04-17T05:25:56Z

+     * @return a new API instance
+     * @see OpenFeatureAPI#createIsolated()
+     */
+    public static OpenFeatureAPI createAPI() {


I don't fully see the point of this method/class. OpenFeatureAPI.createIsolated() has the exact same visibility level and does the same thing. I think this will only lead to confusion

chrfwow · 2026-04-17T05:28:38Z

        // relies on a ready event to be emitted
        emitterExecutor.submit(() -> {
-            try (var ignored = OpenFeatureAPI.lock.readLockAutoCloseable()) {
+            try (var ignored = localLock != null ? localLock.readLockAutoCloseable() : null) {


Why is the lock optional? IIRC, we needed that lock to avoid race conditions when emitting events

Signed-off-by: marcozabel <marco.zabel@dynatrace.com>

…Isolated() Signed-off-by: marcozabel <marco.zabel@dynatrace.com>

Signed-off-by: marcozabel <marco.zabel@dynatrace.com>

chrfwow · 2026-04-21T06:20:28Z

+        api1.setProviderAndWait(new NoOpProvider());
+
+        assertThat(api1HandlerLatch.await(1, TimeUnit.SECONDS)).isTrue();
+        assertThat(api2HandlerCalled.get()).isFalse();


I am not sure this test really tests anything, the lambda for api2 might still get called some time after this tests finished. Unless we can read the queue of the emitterExecutor in the EventProvider, we cannot be sure that the lambda isn't scheduled for execution.

We could wait before this line for a few hundred ms. While this would slow down the test suite and still would not be a real proof, it should catch most errors

…vider binding Signed-off-by: marcozabel <marco.zabel@dynatrace.com>

toddbaert · 2026-04-27T19:10:45Z

+ * @see <a href="https://openfeature.dev/specification/sections/flag-evaluation#18-isolated-api-instances">
+ *     Spec &sect;1.8 &mdash; Isolated API Instances</a>
+ */
+public final class OpenFeatureAPIFactory {


Section 1.8 is marked experimental in the spec. Might be worth adding an @apiNote to flag that, e.g.:

* @apiNote This API is experimental and subject to change.

Could go on the class-level Javadoc or on createAPI() itself (or both).

toddbaert · 2026-04-27T19:19:11Z

+     */
+    @Test
+    @DisplayName("isolated instance does not interfere with singleton")
+    void isolatedInstanceDoesNotInterfereWithSingleton() {


This tests that mutating an isolated instance doesn't affect the singleton, which is great. Worth adding the reverse direction too; e.g. set a provider/hook/context on the singleton and assert a previously created isolated instance is unaffected.

Signed-off-by: marcozabel <marco.zabel@dynatrace.com>

sonarqubecloud · 2026-05-07T10:06:06Z

Quality Gate passed

Issues
1 New issue
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

chrfwow · 2026-05-07T10:13:02Z

-        } else {
-            this.onEmit = onEmit;
        }
+        this.attachment = new Attachment(onEmit, lock);


this.attachment is not guaranteed to be null here any more. If we want to ensure that, we need an AtomicReference and a cas operation

+1 to this; we flagged the same race in our earlier review pass. volatile gives visibility but not atomicity. An AtomicReference with compareAndSet would be better I think, or just guarding attach() with a synchronized block since it's not on a hot path.

toddbaert · 2026-05-12T19:04:19Z

+     * <p><strong>Note:</strong> Isolated API instances (per spec section 1.8) are experimental and
+     * subject to change.
+     */
+    public OpenFeatureAPI() {


I see the OpenFeatureAPIFactory was removed in favor of a public constructor; can you share the reasoning? The factory-in-a-distinct-package approach felt closer to spec 1.8.3 (factory function in a distinct module/package/namespace for discoverability), and intentionally less discoverable than the singleton.

IIUC, a public constructor on OpenFeatureAPI itself sits right next to getInstance() in IntelliSense, which works against the "intentionally less discoverable" goal. Was there a specific issue with the factory class that prompted the removal? Happy to be convinced otherwise.

marcozabel requested review from a team as code owners April 16, 2026 15:00

gemini-code-assist Bot reviewed Apr 16, 2026

View reviewed changes

Comment thread src/test/java/dev/openfeature/sdk/isolated/IsolatedAPITest.java Outdated

Comment thread src/main/java/dev/openfeature/sdk/OpenFeatureAPI.java Outdated

Comment thread src/test/java/dev/openfeature/sdk/isolated/IsolatedAPITest.java Outdated

marcozabel force-pushed the feat/isolated-api-instances branch 5 times, most recently from 518c67f to dcf7a52 Compare April 16, 2026 15:11

feat: support isolated API instances

e7b91f3

Signed-off-by: marcozabel <marco.zabel@dynatrace.com>

marcozabel force-pushed the feat/isolated-api-instances branch from dcf7a52 to e7b91f3 Compare April 16, 2026 15:15

marcozabel added 2 commits April 16, 2026 17:28

fix(tests): separate singleton tests

b084f9c

Signed-off-by: marcozabel <marco.zabel@dynatrace.com>

fix: spotless formatting and arch test compliance

3815faa

Signed-off-by: marcozabel <marco.zabel@dynatrace.com>

chrfwow reviewed Apr 17, 2026

View reviewed changes

marcozabel force-pushed the feat/isolated-api-instances branch 5 times, most recently from 946eb6d to c24b936 Compare April 20, 2026 12:17

marcozabel added 4 commits April 20, 2026 14:33

refactor(test): inject lock via constructor instead of field mutation

da9d472

Signed-off-by: marcozabel <marco.zabel@dynatrace.com>

refactor: make OpenFeatureAPI lock field final

7b9b39d

Signed-off-by: marcozabel <marco.zabel@dynatrace.com>

refactor: move OpenFeatureAPIFactory to sdk package and remove create…

5f14bb9

…Isolated() Signed-off-by: marcozabel <marco.zabel@dynatrace.com>

refactor: bundle EventProvider onEmit and lock into atomic attachment

c06e500

Signed-off-by: marcozabel <marco.zabel@dynatrace.com>

marcozabel force-pushed the feat/isolated-api-instances branch from c24b936 to c06e500 Compare April 20, 2026 12:33

chrfwow reviewed Apr 21, 2026

View reviewed changes

marcozabel force-pushed the feat/isolated-api-instances branch from 896fe71 to 013eb7d Compare April 22, 2026 10:45

feat: move factory to isolated package and warn on multi-instance pro…

0e53150

…vider binding Signed-off-by: marcozabel <marco.zabel@dynatrace.com>

marcozabel force-pushed the feat/isolated-api-instances branch from 013eb7d to 0e53150 Compare April 22, 2026 12:27

toddbaert self-requested a review April 22, 2026 18:58

toddbaert reviewed Apr 27, 2026

View reviewed changes

marcozabel added 3 commits April 29, 2026 15:29

docs: add experimental @APinote to OpenFeatureAPIFactory

586e636

Signed-off-by: marcozabel <marco.zabel@dynatrace.com>

test: add reverse-direction singleton isolation test

2da34f5

Signed-off-by: marcozabel <marco.zabel@dynatrace.com>

test: add delay before negative assertion in event isolation test

3a25c3c

Signed-off-by: marcozabel <marco.zabel@dynatrace.com>

leakonvalinka approved these changes May 4, 2026

View reviewed changes

marcozabel added 3 commits May 7, 2026 11:56

test: use direct constructor in isolated tests

4041ebc

Signed-off-by: marcozabel <marco.zabel@dynatrace.com>

refactor: remove OpenFeatureAPIFactory class

27ca0a2

Signed-off-by: marcozabel <marco.zabel@dynatrace.com>

fix: replace @APinote with prose to satisfy javadoc plugin

3e7374c

Signed-off-by: marcozabel <marco.zabel@dynatrace.com>

chrfwow reviewed May 7, 2026

View reviewed changes

toddbaert reviewed May 12, 2026

View reviewed changes

Conversation

marcozabel commented Apr 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

This PR

Related Issues

Notes

Follow-up Tasks

How to test

Uh oh!

gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

Uh oh!

Uh oh!

Uh oh!

codecov Bot commented Apr 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sonarqubecloud Bot commented May 7, 2026

Quality Gate passed

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

marcozabel commented Apr 16, 2026 •

edited

Loading

codecov Bot commented Apr 16, 2026 •

edited

Loading