AvroConvert 2.2.1

AvroConvert

Avro format combines readability of JSON format and compression of binary data serialization. <br></br>

The main purpose of the project was to enhance communication between microservices. Replacing JSON with Avro brought two main benefits:

• Reduced amount of data send via HTTP by about 30%
• Increased communication security - the data was not visible in plain JSON text

Documentation

General information: http://avro.apache.org/ <br></br> Wiki: https://cwiki.apache.org/confluence/display/AVRO/Index

Why choose Avro?

Benchmark with comparison to Newtonsoft.Json:

You can find the benchmark under tests/AvroConvertTests/Benchmark directory.

Conclusion: <br> Using Avro for communication between your services reduces up to almost 50 times (!!!) network traffic. This makes huge time and bandwidth savings.

Code samples

Serialization

 byte[] avroObject = AvroConvert.Serialize(object yourObject);

Using encoding

 byte[] avroObject = AvroConvert.Serialize(object yourObject, CodecType.Snappy);

Supported encoding types:

• Null (default)
• Deflate
• Snappy
• GZip

<br>

Deserialization

//Using generic method
CustomClass deserializedObject = AvroConvert.Deserialize<CustomClass>(byte[] avroObject);

//Using dynamic method
CustomClass deserializedObject = AvroConvert.Deserialize(byte[] avroObject, typeof(CustomClass));

Deserialization when a property value is null, but schema contains information about default value

//Model used for serialization
public class DefaultValueClass
{
    [DefaultValue("Let's go")]
    public string justSomeProperty { get; set; }

    [DefaultValue(2137)]
    public long? andLongProperty { get; set; }
}

//Deserializing object with null data
 DefaultValueClass deserializedObject = AvroConvert.Deserialize<DefaultValueClass>(byte[] avroObject);

//Produces following object:
> deserializedObject.justSomeProperty
> "Let's go"

> deserializedObject.andLongProperty
> 2137

<br>

Generating Avro schema for C# classes

Using simple class

//Model
public class SimpleTestClass
{
	public string justSomeProperty { get; set; }

	public long andLongProperty { get; set; }
}


//Action
string schemaInJsonFormat = AvroConvert.GenerateSchema(typeof(SimpleTestClass));


//Produces following schema:
"{"type":"record","name":"AvroConvert.SimpleTestClass","fields":[{"name":"justSomeProperty","type":["null","string"]},{"name":"andLongProperty","type":"long"}]}"

Using class decorated with attributes

//Model
[DataContract(Name = "User", Namespace = "user")]
public class AttributeClass
{
	[DataMember(Name = "name")]
	public string StringProperty { get; set; }

	[DataMember(Name = "favorite_number")]
	[NullableSchema]
	public int? NullableIntProperty { get; set; }

	[DataMember(Name = "favorite_color")]
	public string AndAnotherString { get; set; }
}


//Action
string schemaInJsonFormat = AvroConvert.GenerateSchema(typeof(AttributeClass));


//Produces following schema:
"{"type":"record","name":"user.User","fields":[{"name":"name","type":["null","string"]},{"name":"favorite_number","type":["null","int"]},{"name":"favorite_color","type":["null","string"]}]}"

<br />

Reading Avro schema from Avro encoded object

  string schemaInJsonFormat = AvroConvert.GetSchema(byte[] avroObject)

<br>

Contribution

We want to improve AvroConvert as much as possible. If you have any idea, found next possible feature, optimization opportunity or better way for integration, leave a comment or pull request.