如何在MongoDB Atlas Search中跨根与嵌入文档执行Compound/Must查询？

需求 #

实现跨根文档和嵌入文档的Compound/Must查询，要求在所有可搜索字段中同时匹配“Big”和“Piano”，二者必须同时存在，目标返回_id为“101”的文档（该文档中“Big”位于根字段PublisherName，“Piano”位于嵌入字段Books.Title）。

示例文档 #

{
    "_id": "101",
    "PublisherName": "Big book publishers",
    "PublisherAddress": "123 Mian Street",
    "Books": [
        {
            "_id": "6791913210",
            "Title": "Piano made simple",
            "Author": "John Bent"
        },
        {
            "_id": "6638200210",
            "Title": "Guitar made simple",
            "Author": "Kim Larry"
        }
    ]
}

搜索索引配置 #

{
  "mappings": {
    "dynamic": false,
    "fields": {
      "Books": {
        "dynamic": false,
        "fields": {
          "Author": {
            "analyzer": "lucene.standard",
            "type": "string"
          },
          "Title": {
            "analyzer": "lucene.standard",
            "type": "string"
          }
        },
        "type": "embeddedDocuments"
      },
      "PublisherAddress": {
        "analyzer": "lucene.standard",
        "type": "string"
      },
      "PublisherName": {
        "analyzer": "lucene.standard",
        "type": "string"
      }
    }
  }
}

无效查询问题分析 #

当前查询的核心错误在于：

1. 强制要求根字段PublisherName/PublisherAddress同时包含“Big”和“Piano”，但目标文档根字段无“Piano”；
2. 嵌入文档查询部分要求单个Books子文档同时包含“Big”和“Piano”，但目标文档的子文档分别只含其中一个词，无法匹配。

无效查询语句：

[
    {
        "$search": {
            "index": "default",
            "compound": {
                "must": [
                    {
                        "text": {
                            "query": "Big",
                            "path": [
                                "PublisherName",
                                "PublisherAddress"
                            ]
                        }
                    },
                    {
                        "text": {
                            "query": "Piano",
                            "path": [
                                "PublisherName",
                                "PublisherAddress"
                            ]
                        }
                    },
                    {
                        "embeddedDocument": {
                            "operator": {
                                "text": {
                                    "query": "Big",
                                    "path": [
                                        "Books.Title",
                                        "Books.Author"
                                    ]
                                },
                                "text": {
                                    "query": "Piano",
                                    "path": [
                                        "Books.Title",
                                        "Books.Author"
                                    ]
                                }
                            },
                            "path": "Books"
                        }
                    }
                ]
            }
        }
    }
]

正确查询方案 #

只需在compound.must中定义两个独立的text查询，分别匹配“Big”和“Piano”，并将路径覆盖所有可搜索的根字段和嵌入字段即可。该查询会检查整个文档（含根和所有嵌入子文档）是否同时存在两个关键词，完全符合需求。

通用匹配方案（覆盖所有可搜索字段） #

[
  {
    "$search": {
      "index": "default",
      "compound": {
        "must": [
          {
            "text": {
              "query": "Big",
              "path": ["PublisherName", "PublisherAddress", "Books.Title", "Books.Author"]
            }
          },
          {
            "text": {
              "query": "Piano",
              "path": ["PublisherName", "PublisherAddress", "Books.Title", "Books.Author"]
            }
          }
        ]
      }
    }
  }
]

精准匹配方案（限定关键词出现范围） #

如果明确知道“Big”仅出现在根字段、“Piano”仅出现在嵌入字段，可分别指定路径以提升查询效率：

[
  {
    "$search": {
      "index": "default",
      "compound": {
        "must": [
          {
            "text": {
              "query": "Big",
              "path": ["PublisherName", "PublisherAddress"]
            }
          },
          {
            "text": {
              "query": "Piano",
              "path": ["Books.Title", "Books.Author"]
            }
          }
        ]
      }
    }
  }
]

内容的提问来源于stack exchange，提问作者O'Neil Tomlinson