MongoDB嵌套文档中$graphLookup聚合仅返回单文档的问题排查

问题场景 #

使用MongoDB聚合框架的$graphLookup阶段处理嵌套结构文档时，出现结果不符合预期的问题：

• 非嵌套模式下，$graphLookup能正确返回所有符合配置的文档
• 嵌套模式下，仅返回空的nest数组，尝试$unwind和$replaceRoot后仍未解决

非嵌套正常示例 #

文档结构 #

db={
  "fileSystem": [
    {
      "_id": "a",
      "label": "Root",
      "children": ["b"],
    },
    {
      "_id": "b",
      "label": "Nested folder 1",
      "children": ["c","d","e"],
      "parent": "a"
    },
    {
      "_id": "c",
      "label": "Nested File 1.1",
      "parent": "b"
    },
    {
      "_id": "d",
      "label": "Nested File 1.2",
      "parent": "b"
    }
  ]
}

聚合查询 #

db.fileSystem.aggregate([
  {"$match": {"_id": "a"}},
  {
    "$graphLookup": {
      "from": "fileSystem",
      "startWith": "$children",
      "connectFromField": "children",
      "connectToField": "_id",
      "as": "nest",
      "depthField": "level",
      "maxDepth": 1
    }
  }
])

正确结果 #

[
  {
    "_id": "a",
    "children": ["b"],
    "label": "Root",
    "nest": [
      {"_id": "b","children": ["c","d","e"],"label": "Nested folder 1","level": NumberLong(0),"parent": "a"},
      {"_id": "d","label": "Nested File 1.2","level": NumberLong(1),"parent": "b"},
      {"_id": "c","label": "Nested File 1.1","level": NumberLong(1),"parent": "b"}
    ]
  }
]

嵌套异常示例 #

文档结构 #

db={
  "fileSystem": [
    {
      pp: [
        {
          "_id": "a",
          "label": "Root",
          "children": ["b"],
        },
        // 其余内容与非嵌套示例一致
      ]
    }
  ]
}

错误聚合查询 #

db.fileSystem.aggregate([
  {"$unwind": "$pp"},
  {"$replaceRoot": {"newRoot": "$pp"}},
  {"$match": {"_id": "a"}},
  {
    "$graphLookup": {
      "from": "fileSystem",
      "startWith": "$pp.children",
      "connectFromField": "pp.children",
      "connectToField": "pp._id",
      "as": "nest",
      "depthField": "level"
    }
  }
])

错误结果 #

[
  {
    "_id": "a",
    "children": ["b"],
    "label": "Root",
    "nest": []
  }
]

问题原因与解决方案 #

错误原因 #

1. $replaceRoot之后，当前文档已经是pp数组中的节点，不存在$pp.children字段，应该直接使用$children
2. $graphLookup的from指向fileSystem集合，但该集合的文档嵌套在pp数组中，直接查询无法匹配到_id和children字段

正确聚合查询 #

需要在$graphLookup的pipeline中先对from集合的文档进行$unwind和$replaceRoot处理，将嵌套的pp数组展开为根节点：

db.fileSystem.aggregate([
  {"$unwind": "$pp"},
  {"$replaceRoot": {"newRoot": "$pp"}},
  {"$match": {"_id": "a"}},
  {
    "$graphLookup": {
      "from": "fileSystem",
      "startWith": "$children",
      "connectFromField": "children",
      "connectToField": "_id",
      "as": "nest",
      "depthField": "level",
      "maxDepth": 1,
      "pipeline": [
        {"$unwind": "$pp"},
        {"$replaceRoot": {"newRoot": "$pp"}}
      ]
    }
  }
])

执行结果 #

将返回与非嵌套模式一致的正确结果，包含所有关联的子节点。

内容的提问来源于stack exchange，提问作者Nugget