MongoDB 3.4聚合$project字段提取与多集合投影优化问题咨询

Got it, let's break down these two MongoDB 3.4 aggregation challenges with efficient, scalable solutions that avoid the pitfalls you mentioned:

1. Restructure partners to flatten country without $map #

Since you're working with large datasets, skipping iterative operations like $map is the right call. Instead of post-processing, you can directly rewrite the partners object in an $addFields stage (which preserves all other fields by default, making it perfect for your second requirement too).

This replaces the original partners object with a new one that includes only the fields you want, cutting out the nested address layer entirely:

{
  $addFields: {
    partners: {
      country: "$partners.address.country",
      partnerName: "$partners.partnerName"
    }
  }
}

No iteration needed—this is a direct field projection that runs efficiently even on large collections, and it keeps all other existing fields intact.

2. Keep all main collection fields while fetching only specific fields from joined collections #

MongoDB 3.4 doesn't support the pipeline option in $lookup (that arrived in 3.6), so we can't filter fields during the join itself. But we can use $addFields to replace the full joined documents with only the fields we need, while preserving every field from your main collection automatically.

Here's a complete example flow (assuming your main collection is main, and you're joining 3 other collections):

[
  // First join with coll1
  {
    $lookup: {
      from: "coll1",
      localField: "coll1_ref_id",
      foreignField: "_id",
      as: "coll1_data"
    }
  },
  // Replace coll1_data with only the fields you need (adjust for one-to-many if needed)
  {
    $addFields: {
      coll1_data: {
        $arrayElemAt: [
          {
            $map: {
              input: "$coll1_data",
              as: "item",
              in: {
                coll1_field1: "$$item.coll1_field1",
                coll1_field2: "$$item.coll1_field2"
              }
            }
          },
          0 // Use this only for one-to-one joins; remove if you expect multiple matches
        ]
      }
    }
  },
  // Repeat the same pattern for your other 3 joined collections
  {
    $lookup: {
      from: "coll2",
      localField: "coll2_ref_id",
      foreignField: "_id",
      as: "coll2_data"
    }
  },
  {
    $addFields: {
      coll2_data: {
        $arrayElemAt: [
          {
            $map: {
              input: "$coll2_data",
              as: "item",
              in: {
                coll2_fieldA: "$$item.coll2_fieldA",
                coll2_fieldB: "$$item.coll2_fieldB"
              }
            }
          },
          0
        ]
      }
    }
  },
  // ... Add lookup + addFields for coll3 and coll4 here ...
  // Finally, apply the partners restructuring from problem 1
  {
    $addFields: {
      partners: {
        country: "$partners.address.country",
        partnerName: "$partners.partnerName"
      }
    }
  }
]

Key Notes: #

• $addFields preserves all existing fields, so you never have to list every field from your main collection—huge time saver.
• The $map here is only applied to the small joined datasets (2-3 fields each), not your entire large main collection, so it won't impact performance significantly.
• For one-to-many joins, remove the $arrayElemAt to keep all matched documents (each filtered to your desired fields).

内容的提问来源于stack exchange，提问作者SHIVAM SHARMA