Blackbird fails to deserialize varargs array

[naveg]

The following code snippet fails. However, the same snippet works if either Blackbird is removed, or the constructor argument is changed to use double[] notation instead of double...

public class Scratch {
  public static void main(String[] args) throws JsonProcessingException {
    ObjectMapper mapper = new ObjectMapper();
    mapper.registerModule(new BlackbirdModule());

    Foo foo = new Foo(new double[] { 1d, 2d, 3d });
    String serialized = mapper.writeValueAsString(foo);
    Foo foo2 = mapper.readValue(serialized, Foo.class);
  }

  static class Foo {

    @JsonProperty("bar")
    double[] bar;

    @JsonCreator
    public Foo(@JsonProperty("bar") double... bar) {
      this.bar = bar;
    }
  }
}

Exception in thread "main" com.fasterxml.jackson.databind.exc.ValueInstantiationException: Cannot construct instance of `com.alloymetrics.restapi.Scratch$Foo`, problem: class [D cannot be cast to class java.lang.Number ([D and java.lang.Number are in module java.base of loader 'bootstrap')
 at [Source: (String)"{"bar":[1.0,2.0,3.0]}"; line: 1, column: 21]
	at com.fasterxml.jackson.databind.exc.ValueInstantiationException.from(ValueInstantiationException.java:47)
	at com.fasterxml.jackson.databind.DeserializationContext.instantiationException(DeserializationContext.java:1907)
	at com.fasterxml.jackson.databind.DeserializationContext.handleInstantiationProblem(DeserializationContext.java:1260)
	at com.fasterxml.jackson.module.blackbird.deser.OptimizedValueInstantiator.createFromObjectWith(OptimizedValueInstantiator.java:51)
	at com.fasterxml.jackson.databind.deser.ValueInstantiator.createFromObjectWith(ValueInstantiator.java:288)
	at com.fasterxml.jackson.databind.deser.impl.PropertyBasedCreator.build(PropertyBasedCreator.java:202)
	at com.fasterxml.jackson.databind.deser.BeanDeserializer._deserializeUsingPropertyBased(BeanDeserializer.java:454)
	at com.fasterxml.jackson.databind.deser.BeanDeserializerBase.deserializeFromObjectUsingNonDefault(BeanDeserializerBase.java:1405)
	at com.fasterxml.jackson.module.blackbird.deser.SuperSonicBeanDeserializer.deserializeFromObject(SuperSonicBeanDeserializer.java:247)
	at com.fasterxml.jackson.databind.deser.BeanDeserializer.deserialize(BeanDeserializer.java:195)
	at com.fasterxml.jackson.module.blackbird.deser.SuperSonicBeanDeserializer.deserialize(SuperSonicBeanDeserializer.java:116)
	at com.fasterxml.jackson.databind.deser.DefaultDeserializationContext.readRootValue(DefaultDeserializationContext.java:322)
	at com.fasterxml.jackson.databind.ObjectMapper._readMapAndClose(ObjectMapper.java:4593)
	at com.fasterxml.jackson.databind.ObjectMapper.readValue(ObjectMapper.java:3548)
	at com.fasterxml.jackson.databind.ObjectMapper.readValue(ObjectMapper.java:3516)

[cowtowncoder]

Ok sounds like a bug. I assume it'd likely fail with Afterburner as well.

[cowtowncoder]

Does not look like this fails with Afterburner, for whatever reason.

[cowtowncoder]

@stevenschlansker You might have a better idea of where type check or casting goes wrong here: I added test DoubleArrayDeser141Test in com.fasterxml.jackson.module.blackbird.failing package (branch 2.13 -- although if fix is simple, can backport in 2.12).

[stevenschlansker]

Strange, usually varargs and arrays behave the same, but there are a ton of edge cases.
I am on vacation this week but can look next week.

[cowtowncoder]

Yeah I thought they should be (i.e. just syntactic sugar). But I guess there are specific edge cases.

[carterkozak]

I've just encountered this as well. Would you be willing to backport a fix to 2.12.x if I can put something together? I imagine the blackbird module will become increasingly popular as jdk17 LTS is released in a couple months and folks replace afterburner.

[carterkozak]

I've created a simple PR which avoids attempting to optimize varargs methods: #145

I'm not sure what dragons lurk in the deep if we attempt to optimize them, however I suspect it's not worthwhile to optimize varargs methods because they're generally expensive due to implicit array allocation, and tend not to be used in performance critical code.

[stevenschlansker]

Thank you for the PR! I still want to take at least one attempt to optimize it, in case it turns out to be easy - but I will definitely borrow the test cases and this is a good fallback option if it turns out to be hard to optimize.