sql - STRUCT フィールドのネストされた配列に対するフィルタリングとグループ化による BigQuery 最適化クエリ

Question

GQL (Google SQL) クエリを記述して、深くネストされた構造をフィルタリングし、その後再度ネストして、ARRAY と同じレベルの STRUCT プロパティの最初のレコードを残す方法を見つけようとしています。

スキーマサンプルを用意しました

 WITH
      Sale AS (
      SELECT
        "1" AS _id,
        STRUCT("11" AS _id,
          "SERVICE" AS feedbackType,
          DATE(TIMESTAMP("2017-01-20 14:05:51.655")) AS createDate) AS serviceFeedback,
        [STRUCT("host" AS key,
          "localhost" AS value),
        STRUCT("location" AS key,
          "Paris" AS value)] AS tags,
        TRUE AS reviewed,
        [STRUCT("1" as saleId, STRUCT("101" AS _id,
            "PRODUCT" AS feedbackType,
            DATE(TIMESTAMP("2017-01-20 14:05:51.655")) AS createDate) AS productFeedback),
        STRUCT("1" as saleId, STRUCT("102" AS _id,
            "PRODUCT" AS feedbackType,
            DATE(TIMESTAMP("2017-01-20 14:06:51.655")) AS createDate) AS productFeedback) ] AS saleItems,
        DATE(TIMESTAMP("2017-01-20 14:05:51.655")) AS latestFeedbackDate )

また、フィルタリングには、ネストされたすべてのフィールドを平坦化するソース フィルター クエリが必要です。

SELECT
  saleId,
  serviceFeedback,
  saleTags,
  reviewed,
  saleItems,
  latestFeedbackDate
FROM (
  SELECT
    sale._id AS saleId,
    serviceFeedback,
    sale.tags AS saleTags,
    reviewed,
    saleItems,
    latestFeedbackDate
  FROM
    `Sale` AS sale,
    sale.saleItems AS saleItems
  WHERE
    reviewed = TRUE
    AND serviceFeedback.createDate >= DATE(TIMESTAMP("2017-01-18 14:05:51.655"))
    AND serviceFeedback._id IS NOT NULL
    AND saleItems.productFeedback.createDate >= DATE(TIMESTAMP("2017-01-18 14:05:51.655")))
ORDER BY
  latestFeedbackDate DESC
LIMIT
  20

主な問題は、このフィルタリングの後、すべてをグループ化しsaleItemsてsale._id(初期構造を返す)、serviceFeedbackSTRUCT 型のフィールドを取得することです。

JSON 形式で期待される結果は次のとおりです。

{
    "saleId":"1",
    "serviceFeedback":{"_id":"11","feedbackType":"SERVICE","createDate":"2017-01-20"},
    "saleTags":[{"key":"host","value":"localhost"},{"key":"location","value":"Paris"}],
    "reviewed":"true",
    "saleItems":[
        {"saleId":"1","productFeedback":{"_id":"101","feedbackType":"PRODUCT","createDate":"2017-01-20"},
        {"saleId":"1","productFeedback":{"_id":"102","feedbackType":"PRODUCT","createDate":"2017-01-20"},
    ],
    "latestFeedbackDate":"2017-01-20"
}

頭に浮かぶ最も単純なクエリのアイデアを書きました。正しい結果が得られます。しかし、おそらくもっと効率的な方法で書き直すことは可能です。

SELECT
  saleId,
  serviceFeedback,
  latestFeedbackDate,
  subQuery.saleItems as saleItems
FROM
  sale
RIGHT JOIN (
  SELECT
    saleId,
    ARRAY_AGG(saleItems) as saleItems
  FROM (
    SELECT
      saleId,
      saleItems
    FROM (
      SELECT
        sale._id AS saleId,
        latestFeedbackDate,
        saleItems
      FROM
        `Sale` AS sale,
        sale.saleItems AS saleItems
      WHERE
        reviewed = TRUE
        AND serviceFeedback.createDate >= DATE(TIMESTAMP("2017-01-18 14:05:51.655"))
        AND serviceFeedback._id IS NOT NULL
        AND saleItems.productFeedback.createDate >= DATE(TIMESTAMP("2017-01-18 14:05:51.655")))
    ORDER BY
      latestFeedbackDate DESC)
  GROUP BY
    saleId
    ) AS subQuery
ON
  sale._id = subQuery.saleId

同じ結果を達成するためのより良い解決策を提案してもらえますか?