Делаем перекрестный поворот в Google BigQuery

Question

Я задал предыдущий вопрос о выполнении многоуровневого запроса агрегации по оси X здесь: Получите лучшие страны патентов, коды в общедоступном наборе данных BQ .

Вот как работает запрос (скопированный из принятого ответа):

Лучшие 2 страны по количеству, а внутри этих стран 2 лучших кода по количеству

WITH A AS (
  SELECT country_code
  FROM `patents-public-data.patents.publications`
  GROUP BY country_code
  ORDER BY COUNT(1) DESC
  LIMIT 2
), B AS (
  SELECT
    country_code,
    application_kind,
    COUNT(1) application_kind_count
  FROM `patents-public-data.patents.publications`
  WHERE country_code IN (SELECT country_code FROM A)
  GROUP BY country_code, application_kind
), C AS (
  SELECT
    country_code,
    application_kind,
    application_kind_count,
    DENSE_RANK() OVER(PARTITION BY country_code ORDER BY application_kind_count DESC) AS application_kind_rank
  FROM B
)
SELECT
  country_code,
  application_kind,
  application_kind_count
FROM C
WHERE application_kind_rank <= 2  

И я получаю что-то вроде:

country_code     application_kind   count
JP               A                  125
JP               U                  124
CN               A                  118
CN               U                  101

Теперь я хотел бы добавить следующий ось Y: чтобы получить следующее:

• X : 2 лучших страны по количеству, а внутри этих стран 2 лучших кода по количеству
• Y : лучшие 2 family_id по количеству, лучшие 2 priority_date по количеству

Окончательные результаты будут выглядеть следующим образом:

Я могу построить Y-запрос во втором запросе -

WITH A AS (
  SELECT family_id
  FROM `patents-public-data.patents.publications`
  GROUP BY family_id
  ORDER BY COUNT(1) DESC
  LIMIT 2
), B AS (
  SELECT
    family_id,
    priority_date,
    COUNT(1) priority_date_count
  FROM `patents-public-data.patents.publications`
  WHERE family_id IN (SELECT family_id FROM A)
  GROUP BY family_id, priority_date
), C AS (
  SELECT
    family_id,
    priority_date,
    priority_date_count,
    DENSE_RANK() OVER(PARTITION BY family_id ORDER BY priority_date_count DESC) AS priority_date_rank
  FROM B
)
SELECT
  family_id,
  priority_date,
  priority_date_count
FROM C
WHERE priority_date_rank <= 2 

Однако я не совсем уверен, как объединить их в один запрос или в два.

Answer 1

Ниже для стандартного SQL BigQuery и просто демонстрация подхода и не претендует на 100%, представляющих запрошенную логику

WITH A_X AS (
  SELECT country_code FROM `patents-public-data.patents.publications`
  GROUP BY country_code ORDER BY COUNT(1) DESC LIMIT 2
), B_X AS (
  SELECT country_code, application_kind, COUNT(1) application_kind_count
  FROM `patents-public-data.patents.publications` WHERE country_code IN (SELECT country_code FROM A_X)
  GROUP BY country_code, application_kind
), C_X AS (
  SELECT country_code, application_kind, application_kind_count,
    DENSE_RANK() OVER(PARTITION BY country_code ORDER BY application_kind_count DESC) AS application_kind_rank
  FROM B_X
), X AS (
  SELECT country_code, application_kind, application_kind_count
  FROM C_X WHERE application_kind_rank <= 2  
), A_Y AS (
  SELECT family_id FROM `patents-public-data.patents.publications` 
  JOIN X USING(country_code, application_kind)
  GROUP BY family_id 
  ORDER BY COUNT(1) DESC LIMIT 2
), B_Y AS (
  SELECT family_id, priority_date, COUNT(1) priority_date_count
  FROM `patents-public-data.patents.publications` WHERE family_id IN (SELECT family_id FROM A_Y) 
  GROUP BY family_id, priority_date
), C_Y AS (
  SELECT family_id, priority_date, priority_date_count,
    DENSE_RANK() OVER(PARTITION BY family_id ORDER BY priority_date_count DESC) AS pos_date
  FROM B_Y
), Y AS (
  SELECT family_id, priority_date, pos_date, DENSE_RANK() OVER(ORDER BY family_id) pos_family
  FROM C_Y WHERE pos_date <= 2 
)
SELECT country_code, application_kind,
  COUNTIF(pos_family = 1 AND pos_date = 1) `family1_date1`,
  COUNTIF(pos_family = 1 AND pos_date = 2) `family1_date2`,
  COUNTIF(pos_family = 2 AND pos_date = 1) `family2_date1`,
  COUNTIF(pos_family = 2 AND pos_date = 2) `family2_date2`
FROM `patents-public-data.patents.publications` 
JOIN Y USING(family_id, priority_date)
WHERE country_code IN (SELECT country_code FROM X)
AND application_kind IN (SELECT application_kind FROM x) 
GROUP BY country_code, application_kind 

результат

Очевидно, что число нулей выше из-за логики пересечения