MongoDB聚合：unwind保留空数组后分组避免生成空对象

问题分析 #

当原文档的produceItems为空数组时，开启preserveNullAndEmptyArrays: true的$unwind会生成一个produceItems为null的文档。后续$addFields阶段设置"produceItems.properties"时，MongoDB会自动将null的produceItems转为空对象{}，最终$group的$push操作会把这个空对象存入数组，导致结果不符合预期。

解决方案 #

以下三种方案均可解决该问题，按需选择：

方案一：分组后过滤空对象

在$group阶段之后添加$filter，移除数组中的空对象：

db.produce.aggregate([
  {
    $unwind: {
      path: "$produceItems",
      preserveNullAndEmptyArrays: true
    }
  },
  {
    $lookup: {
      from: "fruits",
      localField: "produceItems.item",
      foreignField: "_id",
      as: "fruitProperties"
    }
  },
  {
    $lookup: {
      from: "vegetables",
      localField: "produceItems.item",
      foreignField: "_id",
      as: "vegetableProperties"
    }
  },
  {
    $addFields: {
      "produceItems.properties": {
        $ifNull: [
          { $first: "$fruitProperties" },
          { $first: "$vegetableProperties" }
        ]
      }
    }
  },
  {
    $group: {
      _id: "$_id",
      produceItems: { $push: "$produceItems" }
    }
  },
  // 过滤数组中的空对象
  {
    $addFields: {
      produceItems: {
        $filter: {
          input: "$produceItems",
          cond: { $ne: ["$$this", {}] }
        }
      }
    }
  }
])

方案二：分组时仅推送非空项

在$group的$push中通过$cond判断，仅推送非空的produceItems：

db.produce.aggregate([
  {
    $unwind: {
      path: "$produceItems",
      preserveNullAndEmptyArrays: true
    }
  },
  {
    $lookup: {
      from: "fruits",
      localField: "produceItems.item",
      foreignField: "_id",
      as: "fruitProperties"
    }
  },
  {
    $lookup: {
      from: "vegetables",
      localField: "produceItems.item",
      foreignField: "_id",
      as: "vegetableProperties"
    }
  },
  {
    $addFields: {
      "produceItems.properties": {
        $ifNull: [
          { $first: "$fruitProperties" },
          { $first: "$vegetableProperties" }
        ]
      }
    }
  },
  {
    $group: {
      _id: "$_id",
      produceItems: {
        $push: {
          $cond: [
            { $ne: ["$produceItems", {}] },
            "$produceItems",
            "$$REMOVE"
          ]
        }
      }
    }
  }
])

方案三：提前处理null的produceItems

在$unwind后直接移除null的produceItems字段，避免后续生成空对象：

db.produce.aggregate([
  {
    $unwind: {
      path: "$produceItems",
      preserveNullAndEmptyArrays: true
    }
  },
  // 移除null的produceItems字段
  {
    $addFields: {
      produceItems: {
        $ifNull: ["$produceItems", "$$REMOVE"]
      }
    }
  },
  {
    $lookup: {
      from: "fruits",
      localField: "produceItems.item",
      foreignField: "_id",
      as: "fruitProperties"
    }
  },
  {
    $lookup: {
      from: "vegetables",
      localField: "produceItems.item",
      foreignField: "_id",
      as: "vegetableProperties"
    }
  },
  {
    $addFields: {
      "produceItems.properties": {
        $ifNull: [
          { $first: "$fruitProperties" },
          { $first: "$vegetableProperties" }
        ]
      }
    }
  },
  {
    $group: {
      _id: "$_id",
      produceItems: { $push: "$produceItems" }
    }
  },
  // 将null转为空数组
  {
    $addFields: {
      produceItems: {
        $ifNull: ["$produceItems", []]
      }
    }
  }
])

效果验证 #

三种方案均能实现：

• 当原文档produceItems为空数组时，返回{"_id": 1, "produceItems": []}
• 当原文档produceItems非空时，正常返回包含properties字段的数组

内容的提问来源于stack exchange，提问作者Grant