Repeatable random object on the same seed #413

seregamorph · 2020-05-04T11:37:15Z

The problem: EasyRandom object initialized with the same seed parameter should generate same POJO instances. This contract came from java.util.Random is respected by most of Randomizer implementations. But there is still at least two places that generate new Random() object without seed. As a result the generated POJOs are different.

The context: in my project I validate equals/hashCode/toString implementations for the POJOs that have back references. I generate random beans by EasyRandom instance initialized with same seed value and then call equals/hashCode for such pairs, separate test calls toString(). Expected result: equals=true, hashCode should also equal, all three methods do not throw StackOverflowError (if POJO developer is not careful with equals/hashCode/toString exclude (lombok), the call will lead to infinite nested traversals).

Possible workarounds: for now it can be solved by setting EasyRandomParameters().objectPoolSize(1). Because if the populatedBeans list collection size is single element, there is no options to randomly choose other than first.

Proposed solution: share Random object in RandomizationContext.

TDD: RepeatableRandomTest has test that should fail with old implementation. It's the minimal model that reproduces the issue.

seregamorph · 2020-05-10T09:05:53Z

@benas any ideas, objections, or maybe additional clarifications required?

fmbenhassine · 2020-05-10T10:50:58Z

Hi @seregamorph , thank you for the PR! I did not look at it yet, I have very little spare time during these days which I'm using for another project. I will take a look asap and let you know. Thank you.

seregamorph · 2020-09-27T18:40:52Z

@benas any chance it will be reviewed?

fmbenhassine · 2020-10-09T21:31:24Z

Good catch! With the same seed, we should end-up with the same generated objects (assuming equals/hashcode are correctly implemented). If this is not the case, I consider it as a bug in Easy Random. I will test your PR in details and get back to you if I need more clarification. Thank you!

seregamorph · 2020-10-10T18:57:08Z

With the same seed, we should end-up with the same generated objects

Exactly, this PR addresses an issue with it.

fmbenhassine · 2020-10-31T23:08:07Z

easy-random-core/src/main/java/org/jeasy/random/util/CollectionUtils.java

- *
- * @author Mahmoud Ben Hassine (mahmoud.benhassine@icloud.com)
- */
-public final class CollectionUtils {


Even if this is documented to be intended only for internal use, I do not remove public APIs without notice. Please put this back and mark it as deprecated (See how DateUtils has been deprecated in the same way).

reverted and marked deprecated

fmbenhassine · 2020-10-31T23:10:16Z

easy-random-core/src/main/java/org/jeasy/random/util/CollectionUtils.java

-        return list.get(nextInt(0, list.size()));
-    }
-
-    private static int nextInt(int startInclusive, int endExclusive) {


This is actually duplicated in RandomizationContext. Both methods seems to be called always with startInclusive = 0, which mean this is no different than Random.nextInt(bound), so we can get rid of these methods and use the built-in method instead (I will show how in other comments).

fmbenhassine · 2020-10-31T23:14:08Z

easy-random-core/src/main/java/org/jeasy/random/api/RandomizerContext.java

+     * @param <T>  the type of elements in the list
+     * @return a random element from the list or null if the list is empty
+     */
+    <T> T randomElementOf(List<T> list);


In addition to being a breaking change, this does not belong to the contract of RandomizerContext. We can manage to have a random element from a list where this is used (ie in FieldPopulator and ObjenesisObjectFactory) without this method, see other comments.

fmbenhassine · 2020-10-31T23:15:52Z

easy-random-core/src/main/java/org/jeasy/random/FieldPopulator.java

@@ -128,7 +127,7 @@ private Object generateRandomValue(final Field field, final RandomizationContext
            value = mapPopulator.getRandomMap(field, context);
        } else {
            if (context.getParameters().isScanClasspathForConcreteTypes() && isAbstract(fieldType) && !isEnumType(fieldType) /*enums can be abstract, but can not inherit*/) {
-                Class<?> randomConcreteSubType = randomElementOf(filterSameParameterizedTypes(getPublicConcreteSubTypesOf(fieldType), fieldGenericType));
+                Class<?> randomConcreteSubType = context.randomElementOf(filterSameParameterizedTypes(getPublicConcreteSubTypesOf(fieldType), fieldGenericType));


FieldPopulator has a reference to an EasyRandom instance (which is initialized with the same seed), so this can be replaced with:

List<Class<?>> parameterizedTypes = filterSameParameterizedTypes(getPublicConcreteSubTypesOf(fieldType), fieldGenericType); if (parameterizedTypes.isEmpty()) { throw new ObjectCreationException("Unable to find a matching concrete subtype of type: " + fieldType); } else { Class<?> randomConcreteSubType = parameterizedTypes.get(easyRandom.nextInt(parameterizedTypes.size())); value = easyRandom.doPopulateBean(randomConcreteSubType, context); }

thanks for exact code sample. done

fmbenhassine · 2020-10-31T23:20:11Z

easy-random-core/src/main/java/org/jeasy/random/ObjenesisObjectFactory.java

@@ -47,7 +46,7 @@
    @Override
    public <T> T createInstance(Class<T> type, RandomizerContext context) {
        if (context.getParameters().isScanClasspathForConcreteTypes() && isAbstract(type)) {
-            Class<?> randomConcreteSubType = randomElementOf(getPublicConcreteSubTypesOf((type)));
+            Class<?> randomConcreteSubType = context.randomElementOf(getPublicConcreteSubTypesOf((type)));


We can use a Random instance here and initialize it (only once) with the seed from the context:

private Random random; @Override public <T> T createInstance(Class<T> type, RandomizerContext context) { if (random == null) { random = new Random(context.getParameters().getSeed()); } if (context.getParameters().isScanClasspathForConcreteTypes() && isAbstract(type)) { List<Class<?>> publicConcreteSubTypes = getPublicConcreteSubTypesOf(type); if (publicConcreteSubTypes.isEmpty()) { throw new InstantiationError("Unable to find a matching concrete subtype of type: " + type + " in the classpath"); } else { Class<?> randomConcreteSubType = publicConcreteSubTypes.get(random.nextInt(publicConcreteSubTypes.size())); return (T) createNewInstance(randomConcreteSubType); } } else { try { return createNewInstance(type); } catch (Error e) { throw new ObjectCreationException("Unable to create an instance of type: " + type, e); } } }

fmbenhassine · 2020-10-31T23:23:55Z

easy-random-core/src/main/java/org/jeasy/random/RandomizationContext.java

@@ -105,7 +108,7 @@ boolean hasExceededRandomizationDepth() {
    }

    private int nextInt(int startInclusive, int endExclusive) {


This method can be removed and the line where it is called can be replaced with:

int randomIndex = actualPoolSize > 1 ? random.nextInt(actualPoolSize) : 0;

fmbenhassine · 2020-10-31T23:24:41Z

easy-random-core/src/main/java/org/jeasy/random/RandomizationContext.java

@@ -148,4 +151,12 @@ public Object getRootObject() {
    public EasyRandomParameters getParameters() {
        return parameters;
    }
+
+    @Override
+    public <T> T randomElementOf(List<T> list) {


This should be removed as explained below.

…random

Issue #413

fmbenhassine · 2020-11-01T18:59:29Z

Perfect! Thank you for these updates. Rebased, squashed and merged as 15e5a82. Thank you for your contribution 👍

Repeatable random on the same seed

bd320f2

fmbenhassine added the bug label Oct 9, 2020

fmbenhassine added this to the 4.3.0 milestone Oct 9, 2020

fmbenhassine requested changes Oct 31, 2020

View reviewed changes

fmbenhassine added the waiting-for-feedback label Oct 31, 2020

seregamorph added 2 commits November 1, 2020 14:49

Merge remote-tracking branch 'origin/master' into feature/repeatable-…

6e55488

…random

Address review comments

10aceba

fmbenhassine removed the waiting-for-feedback label Nov 1, 2020

fmbenhassine pushed a commit that referenced this pull request Nov 1, 2020

Configure all Random instances with the same seed

15e5a82

Issue #413

fmbenhassine closed this Nov 1, 2020

seregamorph deleted the feature/repeatable-random branch November 2, 2020 09:07

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repeatable random object on the same seed #413

Repeatable random object on the same seed #413

seregamorph commented May 4, 2020

seregamorph commented May 10, 2020

fmbenhassine commented May 10, 2020

seregamorph commented Sep 27, 2020

fmbenhassine commented Oct 9, 2020

seregamorph commented Oct 10, 2020

fmbenhassine Oct 31, 2020

seregamorph Nov 1, 2020

fmbenhassine Oct 31, 2020

fmbenhassine Oct 31, 2020

seregamorph Nov 1, 2020

fmbenhassine Oct 31, 2020 •

edited

Loading

seregamorph Nov 1, 2020

fmbenhassine Oct 31, 2020 •

edited

Loading

seregamorph Nov 1, 2020

fmbenhassine Oct 31, 2020 •

edited

Loading

seregamorph Nov 1, 2020

fmbenhassine Oct 31, 2020

seregamorph Nov 1, 2020

fmbenhassine commented Nov 1, 2020

		@@ -105,7 +108,7 @@ boolean hasExceededRandomizationDepth() {
		}

		private int nextInt(int startInclusive, int endExclusive) {

Repeatable random object on the same seed #413

Repeatable random object on the same seed #413

Conversation

seregamorph commented May 4, 2020

seregamorph commented May 10, 2020

fmbenhassine commented May 10, 2020

seregamorph commented Sep 27, 2020

fmbenhassine commented Oct 9, 2020

seregamorph commented Oct 10, 2020

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

fmbenhassine Oct 31, 2020 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

fmbenhassine Oct 31, 2020 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

fmbenhassine Oct 31, 2020 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

fmbenhassine commented Nov 1, 2020

fmbenhassine Oct 31, 2020 •

edited

Loading

fmbenhassine Oct 31, 2020 •

edited

Loading

fmbenhassine Oct 31, 2020 •

edited

Loading