prepare 7.4.1 release (#327)

## [7.4.1] - 2024-05-13
### Added:
- Adds warning log if excessive start wait time is used.

### Fixed:
- Improved preprocessing allocations to reduce memory footprint in rare
flag configurations.

---------

Co-authored-by: Eli Bishop <[email protected]>
Co-authored-by: LaunchDarklyReleaseBot <[email protected]>
Co-authored-by: Alex Engelberg <[email protected]>
Co-authored-by: Anton Mostovoy <[email protected]>
Co-authored-by: LaunchDarklyCI <[email protected]>
Co-authored-by: LaunchDarklyCI <[email protected]>
Co-authored-by: Gavin Whelan <[email protected]>
Co-authored-by: ssrm <[email protected]>
Co-authored-by: Harpo Roeder <[email protected]>
Co-authored-by: Ben Woskow <[email protected]>
Co-authored-by: Elliot <[email protected]>
Co-authored-by: Robert J. Neal <[email protected]>
Co-authored-by: Robert J. Neal <[email protected]>
Co-authored-by: Sam Stokes <[email protected]>
Co-authored-by: Ember Stevens <[email protected]>
Co-authored-by: ember-stevens <[email protected]>
Co-authored-by: Alex Engelberg <[email protected]>
Co-authored-by: Louis Chan <[email protected]>
Co-authored-by: Louis Chan <[email protected]>
Co-authored-by: Todd Anderson <[email protected]>
Co-authored-by: tanderson-ld <[email protected]>
Co-authored-by: Matthew M. Keeler <[email protected]>
Co-authored-by: ld-repository-standards[bot] <113625520+ld-repository-standards[bot]@users.noreply.github.com>
Co-authored-by: Kane Parkinson <[email protected]>
Co-authored-by: Ryan Lamb <[email protected]>

Author: 20 people

src/main/java/com/launchdarkly/sdk/server/DataModelPreprocessing.java

@@ -1,6 +1,5 @@
 package com.launchdarkly.sdk.server;
 
-import com.google.common.collect.ImmutableList;
 import com.google.common.collect.ImmutableSet;
 import com.launchdarkly.sdk.EvaluationReason;
 import com.launchdarkly.sdk.EvaluationReason.ErrorKind;
@@ -16,6 +15,7 @@
 
 import java.time.Instant;
 import java.util.ArrayList;
+import java.util.Collections;
 import java.util.List;
 import java.util.Set;
 import java.util.function.Function;
@@ -62,10 +62,10 @@ EvalResult getResult(boolean inExperiment) {
   }
 
   static final class EvalResultFactoryMultiVariations {
-    private final ImmutableList<EvalResultsForSingleVariation> variations;
+    private final List<EvalResultsForSingleVariation> variations;
 
     EvalResultFactoryMultiVariations(
-        ImmutableList<EvalResultsForSingleVariation> variations
+        List<EvalResultsForSingleVariation> variations
         ) {
       this.variations = variations;
     }
@@ -74,6 +74,13 @@ EvalResult forVariation(int index, boolean inExperiment) {
       if (index < 0 || index >= variations.size()) {
         return EvalResult.error(ErrorKind.MALFORMED_FLAG);
       }
+
+      if (variations.get(index) == null) {
+        // getting here indicates that the preprocessor incorrectly processed data and another piece of code is
+        // asking for a variation that was not populated ahead of time.  This is an unexpected bug if it happens.
+        return EvalResult.error(ErrorKind.EXCEPTION);
+      }
+
       return variations.get(index).getResult(inExperiment);
     }
   }
@@ -140,7 +147,7 @@ static final class ValueData {
   static void preprocessFlag(FeatureFlag f) {
     f.preprocessed = new FlagPreprocessed(
         EvaluatorHelpers.offResult(f),
-        precomputeMultiVariationResults(f, EvaluationReason.fallthrough(false),
+        precomputeMultiVariationResultsForFlag(f, EvaluationReason.fallthrough(false),
             EvaluationReason.fallthrough(true), f.isTrackEventsFallthrough())
         );
 
@@ -183,7 +190,7 @@ static void preprocessTarget(Target t, FeatureFlag f) {
   static void preprocessFlagRule(Rule r, int ruleIndex, FeatureFlag f) {
     EvaluationReason ruleMatchReason = EvaluationReason.ruleMatch(ruleIndex, r.getId(), false);
     EvaluationReason ruleMatchReasonInExperiment = EvaluationReason.ruleMatch(ruleIndex, r.getId(), true);
-    r.preprocessed = new FlagRulePreprocessed(precomputeMultiVariationResults(f,
+    r.preprocessed = new FlagRulePreprocessed(precomputeMultiVariationResultsForRule(f, r,
         ruleMatchReason, ruleMatchReasonInExperiment, r.isTrackEvents()));
 
     for (Clause c: r.getClauses()) {
@@ -255,18 +262,52 @@ private static ClausePreprocessed preprocessClauseValues(
     return new ClausePreprocessed(null, valuesExtra);
   }
 
-  private static EvalResultFactoryMultiVariations precomputeMultiVariationResults(
+  private static EvalResultFactoryMultiVariations precomputeMultiVariationResultsForFlag(
       FeatureFlag f,
       EvaluationReason regularReason,
       EvaluationReason inExperimentReason,
       boolean alwaysInExperiment
       ) {
-    ImmutableList.Builder<EvalResultsForSingleVariation> builder =
-        ImmutableList.builderWithExpectedSize(f.getVariations().size());
+    ArrayList<EvalResultsForSingleVariation> variations = new ArrayList<>(f.getVariations().size());
     for (int i = 0; i < f.getVariations().size(); i++) {
-      builder.add(new EvalResultsForSingleVariation(f.getVariations().get(i), i,
+      variations.add(new EvalResultsForSingleVariation(f.getVariations().get(i), i,
           regularReason, inExperimentReason, alwaysInExperiment));
     }
-    return new EvalResultFactoryMultiVariations(builder.build());
+    return new EvalResultFactoryMultiVariations(Collections.unmodifiableList(variations));
+  }
+
+  private static EvalResultFactoryMultiVariations precomputeMultiVariationResultsForRule(
+      FeatureFlag f,
+      Rule r,
+      EvaluationReason regularReason,
+      EvaluationReason inExperimentReason,
+      boolean alwaysInExperiment
+  ) {
+    // Here we create a list of nulls and then insert into that list variations from the rule at their associated index.
+    // This allows the evaluator to then index into the array in constant time. Alternative options are to use a map or
+    // a sparse array. The map has high memory footprint for most customer situations, so it was not used. There is no
+    // standard implementation for sparse array, and it is also not always constant time. Most customers don't have
+    // many variations per flag and so these arrays should not be large on average. This approach was part of a bugfix
+    // and this approach cuts the memory footprint enough to meet the need.
+    List<EvalResultsForSingleVariation> variations = new ArrayList<>(Collections.nCopies(f.getVariations().size(), null));
+    if (r.getVariation() != null) {
+      int index = r.getVariation();
+      if (index >= 0 && index < f.getVariations().size()) {
+        variations.set(index, new EvalResultsForSingleVariation(f.getVariations().get(index), index,
+            regularReason, inExperimentReason, alwaysInExperiment));
+      }
+    }
+
+    if (r.getRollout() != null && r.getRollout().getVariations() != null) {
+      for (DataModel.WeightedVariation wv : r.getRollout().getVariations()) {
+        int index = wv.getVariation();
+        if (index >= 0 && index < f.getVariations().size()) {
+          variations.set(index, new EvalResultsForSingleVariation(f.getVariations().get(index), index,
+              regularReason, inExperimentReason, alwaysInExperiment));
+        }
+      }
+    }
+
+    return new EvalResultFactoryMultiVariations(Collections.unmodifiableList(variations));
   }
 }

src/main/java/com/launchdarkly/sdk/server/DataSourceUpdatesImpl.java

@@ -346,10 +346,11 @@ private void onTimeout() {
         onOutageErrorLog.accept(errorsDesc);
       }
       logger.error(
-          "LaunchDarkly data source outage - updates have been unavailable for at least {} with the following errors: {}",
+          "A streaming connection to LaunchDarkly has not been established within {} after the connection was interrupted. " +
+              "The following errors were encountered: {}",
           Util.describeDuration(loggingTimeout),
           errorsDesc
-          );
+      );
     }
   }
 

src/main/java/com/launchdarkly/sdk/server/LDClient.java

@@ -67,6 +67,8 @@ public final class LDClient implements LDClientInterface {
   private final LDLogger baseLogger;
   private final LDLogger evaluationLogger;
 
+  private static final int EXCESSIVE_INIT_WAIT_MILLIS = 60000;
+
   /**
    * Creates a new client instance that connects to LaunchDarkly with the default configuration.
    * <p>
@@ -239,6 +241,9 @@ public LDClient(String sdkKey, LDConfig config) {
       if (!(dataSource instanceof ComponentsImpl.NullDataSource)) {
         baseLogger.info("Waiting up to {} milliseconds for LaunchDarkly client to start...",
             config.startWait.toMillis());
+        if (config.startWait.toMillis() > EXCESSIVE_INIT_WAIT_MILLIS) {
+          baseLogger.warn("LaunchDarkly client created with start wait time of {} milliseconds.  We recommend a timeout of less than {} milliseconds.", config.startWait.toMillis(), EXCESSIVE_INIT_WAIT_MILLIS);
+        }
       }
       try {
         startFuture.get(config.startWait.toMillis(), TimeUnit.MILLISECONDS);

src/test/java/com/launchdarkly/sdk/server/DataModelPreprocessingTest.java

@@ -19,6 +19,7 @@
 
 import java.time.Instant;
 import java.time.ZonedDateTime;
+import java.util.ArrayList;
 import java.util.List;
 
 import static com.launchdarkly.sdk.server.ModelBuilders.clause;
@@ -136,9 +137,44 @@ public void preprocessFlagAddsPrecomputedPrerequisiteFailedResults() {
   }
 
   @Test
-  public void preprocessFlagAddsPrecomputedResultsToFlagRules() {
+  public void preprocessFlagAddsPrecomputedResultsToFlagRulesWithRollout() {
+
+    List<DataModel.WeightedVariation> variations = new ArrayList<>();
+    variations.add(new DataModel.WeightedVariation(0, 50000, false));
+    variations.add(new DataModel.WeightedVariation(1, 50000, false));
+    DataModel.RolloutKind kind = DataModel.RolloutKind.rollout;
+    Integer seed = 123;
+    DataModel.Rollout rollout = new DataModel.Rollout(null, variations, null, kind, seed);
+
     FeatureFlag f = new FeatureFlag("key", 0, false, null, null, null, null,
-        ImmutableList.of(new Rule("ruleid0", ImmutableList.of(), null, null, false)),
+        ImmutableList.of(new Rule("ruleid0", ImmutableList.of(), null, rollout, false)),
+        null, null,
+        ImmutableList.of(aValue, bValue),
+        false, false, false, null, false, null, null, false);
+
+    f.afterDeserialized();
+
+    Rule rule = f.getRules().get(0);
+    assertThat(rule.preprocessed, notNullValue());
+    assertThat(rule.preprocessed.allPossibleResults, notNullValue());
+    EvaluationReason regularReason = EvaluationReason.ruleMatch(0, "ruleid0", false);
+    EvaluationReason inExperimentReason = EvaluationReason.ruleMatch(0, "ruleid0", true);
+
+    assertThat(rule.preprocessed.allPossibleResults.forVariation(0, false),
+        equalTo(EvalResult.of(aValue, 0, regularReason)));
+    assertThat(rule.preprocessed.allPossibleResults.forVariation(0, true),
+        equalTo(EvalResult.of(aValue, 0, inExperimentReason)));
+
+    assertThat(rule.preprocessed.allPossibleResults.forVariation(1, false),
+        equalTo(EvalResult.of(bValue, 1, regularReason)));
+    assertThat(rule.preprocessed.allPossibleResults.forVariation(1, true),
+        equalTo(EvalResult.of(bValue, 1, inExperimentReason)));
+  }
+
+  @Test
+  public void preprocessFlagAddsPrecomputedResultsToFlagRulesWithJustVariation() {
+    FeatureFlag f = new FeatureFlag("key", 0, false, null, null, null, null,
+        ImmutableList.of(new Rule("ruleid0", ImmutableList.of(), 0, null, false)),
         null, null,
         ImmutableList.of(aValue, bValue),
         false, false, false, null, false, null, null, false);
@@ -157,9 +193,9 @@ public void preprocessFlagAddsPrecomputedResultsToFlagRules() {
         equalTo(EvalResult.of(aValue, 0, inExperimentReason)));
 
     assertThat(rule.preprocessed.allPossibleResults.forVariation(1, false),
-        equalTo(EvalResult.of(bValue, 1, regularReason)));
+        equalTo(EvalResult.error(EvaluationReason.ErrorKind.EXCEPTION)));
     assertThat(rule.preprocessed.allPossibleResults.forVariation(1, true),
-        equalTo(EvalResult.of(bValue, 1, inExperimentReason)));
+        equalTo(EvalResult.error(EvaluationReason.ErrorKind.EXCEPTION)));
   }
 
   @Test