Extended Reference Pattern

• Joins are expensive
• so store some must have data into the child to avoid joins.
• This creates data duplicate, yes, but it's an expense we can pay in exchange for performance.
• Duplicated data should not change frequently in nature.
• If data needs to be updated, int most cases real-time update is not necessary, so period data sync is more than enough.

var reshape_review_docs_pipeline = [
  {
    $lookup: {
      from: "books",
      localField: "product_id",
      foreignField: "product_id",
      as: "product_info",
    },
  },
  {
    $unwind: {
      path: "$product_info",
      includeArrayIndex: "string",
      preserveNullAndEmptyArrays: false,
    },
  },
  {
    $project: {
      _id: "$_id",
      "product.product_id": "$product_id",
      "product.product_type":
        "$product_info.product_type",
      "product.title": "$product_info.title",
      "review.user_id": "$user_id",
      "review.reviewTitle": "$reviewTitle",
      "review.reviewBody": "$reviewBody",
      "review.date": "$date",
      "review.stars": "$stars",
    },
  },
  {
    $merge: {
      into: "reviews",
      on: "_id",
      whenMatched: "replace",
      whenNotMatched: "discard",
    },
  },
]

db.reviews.aggregate(reshape_review_docs_pipeline)