带avro序列化的C#汇合kafka问题

Question

我正在使用docker从https://github.com/confluentinc/cp-All-in-One运行kafka和其他服务，并在我的测试项目中使用用于kafka、avro和schemaRegistry的汇合nuget包。

如果要发送json消息，我现在还没有问题，但是我正在努力发送avro序列化消息。

我看到https://github.com/confluentinc/confluent-kafka-dotnet/tree/master/examples/avrospecific示例，我尝试用同样的方法执行，但最终我得到了如下所示的异常：

Local：值序列化错误
在confluent.kafka.producer2. D__52.MoveNext()在system.runtime.compilerServices.taskawaiter.throwforNonSuccess（任务任务）在system.runtime.compilerServices.taskawaiter 1.GetResult()在c:\users\lu95eb\source\repos\kafka_playground\kafkaProducer\kafkaService.cs:行126

内部例外

对象引用未设置为对象的实例。
在Confluent.SchemaRegistry.Serdes.SpecificSerializerImpl在Confluent.schemaRegistryClient schemaRegistryClient,Boolean autoRegisterSchema,Int32 initialBufferSize在System.Runtime.CompilerServices.Taskawaiter.ThrowForNonSuccess（Task任务）在

这是我的SpecificRecord类

public class UserInfo : ISpecificRecord
{
    public string Name { get; set; }
    public int[] Numbers { get; set; }

    public Schema Schema => Schema.Parse(@"
        {
          ""name"": ""UserInfo"",
          ""type"": ""record"",
          ""namespace"": ""kafka"",
          ""fields"": [
            {
              ""name"": ""Name"",
              ""type"": ""string""
            },
            {
              ""name"": ""Numbers"",
              ""type"": {
                ""type"": ""array"",
                ""items"": ""int""
              }
            }
          ]
        }
        ");

    public object Get(int fieldPos)
    {
        switch (fieldPos)
        {
            case 0: return Name;
            case 1: return Numbers;
            default: throw new AvroRuntimeException($"Bad index {fieldPos} in Get()");
        }
    }

    public void Put(int fieldPos, object fieldValue)
    {
        switch (fieldPos)
        {
            case 0: Name = (string)fieldValue; break;
            case 1: Numbers = (int[])fieldValue; break;
            default: throw new AvroRuntimeException($"Bad index {fieldPos} in Put()");
        }
    }
}

和用于发送消息的方法

private async Task SendSpecificRecord(UserInfo userInfo)
    {
        using (var schemaRegistry = new CachedSchemaRegistryClient(new SchemaRegistryConfig { Url = _schemaRegistryUrl }))
        using (var producer =
            new ProducerBuilder<string, UserInfo>(new ProducerConfig { BootstrapServers = _brokerUrl })
                .SetKeySerializer(new AvroSerializer<string>(schemaRegistry))
                .SetValueSerializer(new AvroSerializer<UserInfo>(schemaRegistry))
                .Build())
        {

            var message = new Message<string, UserInfo>
            {
                Key = userInfo.Name,
                Value = userInfo
            };


            await producer.ProduceAsync(SpecificTopic, message);
        }
    }

kafkaservice.cs：第126行是Await Producer.ProduceAsync（SpecificTopic,message）；

如果有人能指出我做错了什么，我会很感激的。提前谢谢你。

Answer 1

如果有人对解决方案感到好奇（我无法想象某人是怎样的；））然后我编写了“自定义”avro序列化器和反序列化器，并像一个魅力一样工作。

public class CustomAvroSerializer<T> : IAsyncSerializer<T>
    where T : class, ISpecificRecord
{
    public Task<byte[]> SerializeAsync(T data, SerializationContext context)
    {
        return Task.Run(() =>
        {
            using (var ms = new MemoryStream())
            {
                var enc = new BinaryEncoder(ms);
                var writer = new SpecificDefaultWriter(data.Schema);
                writer.Write(data, enc);
                return ms.ToArray();
            }
        });
    }
}

public class CustomAvroDeserializer<T> : IDeserializer<T>
    where T : class, ISpecificRecord
{
    public T Deserialize(ReadOnlySpan<byte> data, bool isNull, SerializationContext context)
    {
        using (var ms = new MemoryStream(data.ToArray()))
        {
            var dec = new BinaryDecoder(ms);
            var regenObj = (T)Activator.CreateInstance(typeof(T));

            var reader = new SpecificDefaultReader(regenObj.Schema, regenObj.Schema);
            reader.Read(regenObj, dec);
            return regenObj;
        }
    }
}