集計QUERY - 毎月のための一人あたりの3ヶ月平均

Question

は今、私は毎月

すべての人のための最後の3ヶ月間、この平均を取得する必要があり、私はBigQueryの

Person | Amount | yearMonth 
--------------------------- 
AA  | 100 | 201701 
AA  | 200 | 201702 
AA  | 300 | 201703 
AA  | 70 | 201704 
AB  | 10 | 201701 
AB  | 50 | 201702 
AB  | 60 | 201703 
AB  | 70 | 201704 
AC  | 70 | 201701 
AC  | 80 | 201702 
AC  | 30 | 201703 
AC  | 10 | 201704 

にデータのビューを持っていると仮定します

期待される結果：これはどのように計算され

Person | Amount | yearMonth 
--------------------------- 
AA  | 200 | 201703(avg of 201701-201703) 
AA  | 190 | 201704(avg of 201702-201704) 
AB  | 40 | 201703(avg of 201701-201703) 
AB  | 60 | 201704(avg of 201702-201704) 
AC  | 60 | 201703(avg of 201701-201703) 
AC  | 40 | 201704(avg of 201702-201704) 

？

最初の行

• AA = 200は、+ 100（201701）+200（201702）+300（201703）/ 3 = 200
• AA = 100から200（201702）から来る来ます300（201703）+70（201704）/ 3 = 190
• AB = 40、10（201701）+50（201702）60（201703）/ 3 = 40
• とすぐ

から来ます

私はこれでどのようにグループ化するのか分かりません。あなたの答えがこの問題へのリンクを持っているのであれば、私は気にしません。

ありがとうヒープ

レガシーSQLでも可能ですか？私はまだstandardSQLに移行していません。 私の見解は、従来のSQLである以下

Answer 1

は

#standardSQL 
SELECT 
    person, yearMonth, CAST(amount AS INT64) amount 
FROM (
    SELECT 
    person, yearMonth, dt, 
    AVG(amount) OVER(PARTITION BY person ORDER BY dt RANGE BETWEEN 63 PRECEDING AND CURRENT row) amount, 
    COUNT(1) OVER(PARTITION BY person ORDER BY dt RANGE BETWEEN 63 PRECEDING AND CURRENT row) months 
    FROM (
    SELECT 
     person, amount, yearMonth, 
     UNIX_DATE(DATE(DIV(yearMonth, 100), MOD(yearMonth, 100), 1)) AS dt 
    FROM `project.dataset.table` 
) 
) 
WHERE months = 3 
-- ORDER BY person, yearMonth 

あなたが/テストダミーでそれで遊ぶことができます（少なくとも適切なグループ化のロジックにあなたのアイデアを与える必要があります）BigQueryの標準SQLのためであります

#standardSQL 
WITH `project.dataset.table` AS (
    SELECT 'AA' person, 100 amount, 201701 yearMonth UNION ALL 
    SELECT 'AA', 200, 201702 UNION ALL 
    SELECT 'AA', 300, 201703 UNION ALL 
    SELECT 'AA', 70, 201704 UNION ALL 
    SELECT 'AB', 10, 201701 UNION ALL 
    SELECT 'AB', 50, 201702 UNION ALL 
    SELECT 'AB', 60, 201703 UNION ALL 
    SELECT 'AB', 70, 201704 UNION ALL 
    SELECT 'AC', 70, 201701 UNION ALL 
    SELECT 'AC', 80, 201702 UNION ALL 
    SELECT 'AC', 30, 201703 UNION ALL 
    SELECT 'AC', 10, 201704 
) 
SELECT 
    person, yearMonth, CAST(amount AS INT64) amount 
FROM (
    SELECT 
    person, yearMonth, dt, 
    AVG(amount) OVER(PARTITION BY person ORDER BY dt RANGE BETWEEN 63 PRECEDING AND CURRENT row) amount, 
    COUNT(1) OVER(PARTITION BY person ORDER BY dt RANGE BETWEEN 63 PRECEDING AND CURRENT row) months 
    FROM (
    SELECT 
     person, amount, yearMonth, 
     UNIX_DATE(DATE(DIV(yearMonth, 100), MOD(yearMonth, 100), 1)) AS dt 
    FROM `project.dataset.table` 
) 
) 
WHERE months = 3 
ORDER BY person, yearMonth 

以下の期待通りの出力が

person yearMonth amount 
AA  201703  200 
AA  201704  190 
AB  201703  40 
AB  201704  60 
AC  201703  60 
AC  201704  40  

0であるとしてデータ

BigQueryのレガシーSQL

#legacySQL 
SELECT 
    person, yearMonth, INTEGER(amount) amount 
FROM (
    SELECT 
    person, yearMonth, dt, 
    AVG(amount) OVER(PARTITION BY person ORDER BY dt range BETWEEN 63*60*60*24 preceding AND current row) amount, 
    COUNT(1) OVER(PARTITION BY person ORDER BY dt range BETWEEN 63*60*60*24 preceding AND current row) months 
    FROM (
    SELECT 
     person, amount, yearMonth, 
     TIMESTAMP_TO_SEC(TIMESTAMP(CONCAT(STRING(INTEGER(yearMonth/100)), '-', SUBSTR(STRING(100 + yearMonth % 100), 2, 2), '-01'))) AS dt 
    FROM [project:dataset.table] 
) 
) 
WHERE months = 3 
-- ORDER BY person, yearMonth 

ため

追加バージョンあなたがテストすることができます/ダミーデータ

#legacySQL 
SELECT 
    person, yearMonth, INTEGER(amount) amount 
FROM (
    SELECT 
    person, yearMonth, dt, 
    AVG(amount) OVER(PARTITION BY person ORDER BY dt range BETWEEN 63*60*60*24 preceding AND current row) amount, 
    COUNT(1) OVER(PARTITION BY person ORDER BY dt range BETWEEN 63*60*60*24 preceding AND current row) months 
    FROM (
    SELECT 
     person, amount, yearMonth, 
     TIMESTAMP_TO_SEC(TIMESTAMP(CONCAT(STRING(INTEGER(yearMonth/100)), '-', SUBSTR(STRING(100 + yearMonth % 100), 2, 2), '-01'))) AS dt 
    FROM -- [project:dataset.table] 
     (SELECT 'AA' person, 100 amount, 201701 yearMonth), 
     (SELECT 'AA' person, 200 amount, 201702 yearMonth), 
     (SELECT 'AA' person, 300 amount, 201703 yearMonth), 
     (SELECT 'AA' person, 70 amount, 201704 yearMonth), 
     (SELECT 'AB' person, 10 amount, 201701 yearMonth), 
     (SELECT 'AB' person, 50 amount, 201702 yearMonth), 
     (SELECT 'AB' person, 60 amount, 201703 yearMonth), 
     (SELECT 'AB' person, 70 amount, 201704 yearMonth), 
     (SELECT 'AC' person, 70 amount, 201701 yearMonth), 
     (SELECT 'AC' person, 80 amount, 201702 yearMonth), 
     (SELECT 'AC' person, 30 amount, 201703 yearMonth), 
     (SELECT 'AC' person, 10 amount, 201704 yearMonth) 
) 
) 
WHERE months = 3 
ORDER BY person, yearMonth