我如何注册一个自定义转换器，以便@GeoShapeField 可以使用它？

Question

I am using spring-data-elasticsearch (5) to automagically write third-party data into an ES (8) index. The data contains geodata in GML format, which is parsed into a nested Map<String, Object> .

In my POJO I have a field

@GeoShapeField
private Map<String, Object> geometry;

This is written perfectly fine in many cases; however, the data I get can also contain eg Envelope , which is not supported by GeoJson but could be imported without problems into ES.

I can write simple custom ReadingConverter / WritingConverter s - but how can I register them in a way that @GeoShapeField automatically chooses them when appropriate?

I see that org.springframework.data.elasticsearch.core.convert.GeoConverters is responsible for choosing the correct converter, esp. .GeoJsonToMapConverter and .MapToGeoJsonConverter . How would I correctly extend the class/replace it, so that @GeoShapeField looks for an additional (or more) type(s)?

Answer 1

As pointed out by PJMeisch in the comments, I had several understanding problems, which led to the formulation of the question.

The answer to my actual question is straightforward: for Envelope, Elasticsearch expects

"myField": {
    "type" : "envelope",
    "coordinates" : [ [100.0, 1.0], [101.0, 0.0] ]
}

To achieve this using spring-data-elasticsearch, it is enough to provide a simple translation into a Map<String, Object> :

Map<String, Object> myField = new HashMap<>();
myField.put("type", "envelope");
myField.put("coordinates", Arrays.asList(Arrays.asList(100.0, 1.0), Arrays.asList(101.0, 0.0)));

The point over which I tripped was the data . I receive bounding boxes specifying lower left and upper right corners. However, ES expects upper left and lower right corners for a bounding box. After switching the respective positions, everything works now.