The Mongo Java driver has annotations for POJOs to tell the driver how they should be serialized and deserialized, through things like marking a constructor to use, marking which getters/fields should be used and which should be ignored, etc. However, you can't provide custom codecs for any of the properties. To do that, you have to use things like the ClassModelBuilder.

I would like annotations to provide custom codecs. I propose something like this:

public @interface BsonDecoder {
    Class<Decoder<?>> value();
}

public @interface BsonEncoder {
    Class<Encoder<?>> value();
}

public class IntegerToStringEncoder implements Encoder<String> {
    @Override
    public void encode(BsonWriter writer, Integer value, EncoderContext encoderContext) {
        writer.writeString(value.toString());
    }

    @Override
    public Class<Integer> getEncoderClass() {
        return Integer.class;
    }
}

public class StringToIntegerDecoder implements Decoder<Integer> {
    @Override
    public Integer decode(BsonReader reader, DecoderContext decoderContext) {
        return Integer.parseInt(reader.readString());
    }
}

public class MyPojo {
    private final Integer myNumberStoredAsString;

    @BsonCreator
    public MyPojo(@BsonProperty("myNumberStoredAsString") @BsonDecoder(StringToIntegerDecoder.class) Integer myNumberStoredAsString) {
        this.myNumberStoredAsString = myNumberStoredAsString;
    }

    @BsonProperty("myNumberStoredAsString")
    @BsonEncoder(IntegerToStringEncoder.class)
    public Integer getMyNumberStoredAsString() {
        return myNumberStoredAsString;
    }
}

Another option would be adding new fields to the @BsonProperty annotation.

public @interface BsonProperty {
    String value() default "";

    boolean useDiscriminator() default false;

    Class<Encoder<?>> encoder() default null;

    Class<Decoder<?>> decoder() default null;
}

My use case for this feature is something I'm working on at work. There's a microservice that has an old data format, and I want to change it to use a new data format in MongoDB. However, I don't want to migrate all the data from the old format to the new format, so I decided to instead make the microservice able to read data in both the old and new format, but only save data in the new format. So I'm implementing custom decoders to handle both formats.

I already have a decent idea of how to implement this, as I've implemented my own solution in the work project, using reflection to scan for POJOs with Bson annotations, then finding my own custom annotation and registering the custom encoders and/or decoders. If this ticket is approved, I can create a Pull Request with the change for this feature. The only potential issue I can see is that to allow only providing an encoder or decoder, the PropertyModel<T> class may have to expose a `getEncoder()` and `getDecoder()` instead of a `getCodec()` method.

The main question with this ticket is "is this something that would be acceptable/desired in the Mongo Java Driver?" I think it'd be a pretty good feature, and I'm keen to implement it.