Hibernate 搜索 JsonB 索引

Question

我正在努力将 jsonB 列索引到 Elasicsearch 后端，使用 Hibernate Search 6.0.2

这是我的实体：

@Data
@NoArgsConstructor
@Entity
@Table(name = "examples")
public class Example {

    @Id
    @GeneratedValue(strategy = GenerationType.AUTO)
    private UUID id;

    @NotNull
    @Column(name = "fields")
    @Type(type = "jsonb")
    private Map<String, Object> fields;
}

这是我用于 Hibernate Search 的 elasticsearch 后端的编程映射：

@Configuration
@RequiredArgsConstructor
public class ElasticsearchMappingConfig implements HibernateOrmSearchMappingConfigurer {

    private final JsonPropertyBinder jsonPropertyBinder;

    @Override
    public void configure(HibernateOrmMappingConfigurationContext context) {
        var mapping = context.programmaticMapping();
        var exampleMapping = mapping.type(Example.class);
        exampleMapping.indexed();
        exampleMapping.property("fields").binder(jsonPropertyBinder);
    }
}

我的自定义属性活页夹实现基于 Hibernate Search 6.0.2 documentation。

@Component
public class JsonPropertyBinder implements PropertyBinder {

    @Override
    public void bind(PropertyBindingContext context) {
        context.dependencies().useRootOnly();
        var schemaElement = context.indexSchemaElement();
        var userMetadataField = schemaElement.objectField("metadata");
        context.bridge(Map.class, new Bridge(userMetadataField.toReference()));
    }

    @RequiredArgsConstructor
    private static class Bridge implements PropertyBridge<Map> {

        private final IndexObjectFieldReference fieldReference;

        @Override
        public void write(DocumentElement target, Map bridgedElement, PropertyBridgeWriteContext context) {
            var map = target.addObject(fieldReference);
            ((Map<String, Object>) bridgedElement).forEach(map::addValue);
        }
    }

}

我知道文档为 Map 中的对象可以是什么定义了多个模板（如 MultiTypeUserMetadataBinder 示例），但我真的不知道里面可以是什么。我所知道的是，它是一个有效的 json，我的目标是将它作为“字段”下的有效 json 结构放入 Elasticsearch：{...}

在我的例子中 jsonB 列可能包含如下内容：

    {
        "testString": "298",
        "testNumber": 123,
        "testBoolean": true,
        "testNull": null,
        "testArray": [
            5,
            4,
            3
        ],
        "testObject": {
            "testString": "298",
            "testNumber": 123,
            "testBoolean": true,
            "testNull": null,
            "testArray": [
                5,
                4,
                3
            ]
        }

但抛出异常：

org.hibernate.search.util.common.SearchException: HSEARCH400609: Unknown field 'metadata.testNumber'.

我还在我的 spring 应用程序中将 dynamic_mapping 设置为 true：

...
spring.jpa.properties.hibernate.search.backend.hosts=127.0.0.3:9200
spring.jpa.properties.hibernate.search.backend.dynamic_mapping=true
...

还有其他想法可以解决这个问题吗？还是我哪里出错了？

Answer 1

I am aware that documentation defines multiple templates for what an Object in Map can be (like in MultiTypeUserMetadataBinder example), but I really do not know what can be inside. All I know, it is a valid json and my goal is to put it into Elasticsearch as valid json structure under "fields": {...}

如果您不知道每个字段的类型是什么，Hibernate Search 将无济于事。如果您真的想将其填充到索引中，我建议您声明一个 native field and pushing the JSON as-is. But then you won't be able to apply predicates to the metadata fields easily, except using native JSON.

像这样：

@Component
public class JsonPropertyBinder implements PropertyBinder {

    @Override
    public void bind(PropertyBindingContext context) {
        context.dependencies().useRootOnly();
        var schemaElement = context.indexSchemaElement();

        // CHANGE THIS
        IndexFieldReference<JsonElement> userMetadataField = schemaElement.field( 
                "metadata",
                f -> f.extension(ElasticsearchExtension.get())
                        .asNative().mapping("{\"type\": \"object\", \"dynamic\":true}");
        )
                .toReference();

        context.bridge(Map.class, new Bridge(userMetadataField));
    }

    @RequiredArgsConstructor
    private static class Bridge implements PropertyBridge<Map> {
        private static final Gson GSON = new Gson();

        private final IndexFieldReference<JsonElement> fieldReference;

        @Override
        public void write(DocumentElement target, Map bridgedElement, PropertyBridgeWriteContext context) {
            // CHANGE THIS
            target.addValue(fieldReference, GSON.toJsonTree(bridgedElement));
        }
    }

}

或者，您可以只将所有字段声明为字符串。然后 Hibernate Search 提供的字符串类型的所有功能都将可用。但是当然，范围谓词或排序之类的东西会导致数值出现奇怪的结果（2 在 10 之前，但是 "2" 在之后 "10").

像这样：

@Component
public class JsonPropertyBinder implements PropertyBinder {

    @Override
    public void bind(PropertyBindingContext context) {
        context.dependencies().useRootOnly();
        var schemaElement = context.indexSchemaElement();
        var userMetadataField = schemaElement.objectField("metadata");

        // ADD THIS
        userMetadataField.fieldTemplate( 
                "userMetadataValueTemplate_default",
                f -> f.asString().analyzer( "english" )
        );

        context.bridge(Map.class, new Bridge(userMetadataField.toReference()));
    }

    @RequiredArgsConstructor
    private static class Bridge implements PropertyBridge<Map> {

        private final IndexObjectFieldReference fieldReference;

        @Override
        public void write(DocumentElement target, Map bridgedElement, PropertyBridgeWriteContext context) {
            var map = target.addObject(fieldReference);
            // CHANGE THIS
            ((Map<String, Object>) bridgedElement).forEach(entry -> map.addValue( entry.getKey(), String.valueOf(entry.getValue())));
        }
    }

}

Hibernate 搜索 JsonB 索引

Hibernate Search JsonB indexing

java

hibernate-search

elasticsearch

spring-boot

jsonb