如何合并两个满足最小值的字段的表？

Question

我正在尝试完善我的数据，但找不到好的解决方案。

我正在处理动态生成的周期（示例投影）。我不知道重新创建 table 只是为了加入我的 table TREND 是否更好。

-- SQL CODE TO RECREATE MY TREND TABLE
CREATE OR REPLACE TABLE test.TREND(
Asset STRING,
Metric_Code STRING,
Metric_V FLOAT64,
Trend FLOAT64,
YearMonth INT64,
YearMonthRank INT64
);
INSERT INTO test.TREND
VALUES ('RDX4', 'AXCP32', 88.0853212, 89.0853212, 202103,1),
('RDX4', 'AXCP32', 88.0853212,85.0853212, 202106,2),
('RDX4', 'AXCP34', 79.41087028,81.0853212, 202103,3),
('RDX4', 'AXCP34', 70.44587,69.0853212, 202106,4),
('MRDX4', 'AXCP32', 88.0853212, 89.0853212, 202103,1),
('MRDX4', 'AXCP32', 88.0853212,85.0853212, 202106,2),
('MRDX4', 'AXCP34', 79.41087028,81.0853212, 202103,3),
('MRDX4', 'AXCP34', 70.44587, 69.0853212,202106,4);

-- SQL CODE TO RECREATE MY PROJECTION TABLE
CREATE OR REPLACE TABLE test.PROJECTION(
Asset STRING,
Metric_Code STRING,
Metric_V FLOAT64,
Trend FLOAT64,
YearMonth INT64,
YearMonthRank INT64
);
INSERT INTO PROJECTION
VALUES (NULL,NULL,NULL,NULL, 202103,1),
(NULL,NULL,NULL,NULL, 202106,2),
(NULL,NULL,NULL,NULL, 202109,3),
(NULL,NULL,NULL,NULL, 202112,4),
(NULL,NULL,NULL,NULL, 202203,5),
(NULL,NULL,NULL,NULL, 202206,6),
(NULL,NULL,NULL,NULL, 202209,7),
(NULL,NULL,NULL,NULL, 202212,8);

现在我卡住了，我无法正确合并数据来计算我的预测。我试图合并资产，Metric_Code 和趋势，但它并不像你想象的那样有效：

SELECT COALESCE(B.Asset,(SELECT C.asset FROM gopass.TREND C WHERE C.YearMonthRank=1 AND B.asset=C.asset)) AS Asset,
B.Metric_Code,
B.Metric_V,
B.Trend,
A.YearMonth,
A.YearMonthRank
FROM gopass.PROJECTION A
LEFT JOIN gopass.TREND B  ON A.YearMonthRank = B.YearMonthRank
ORDER BY Asset ASC;

我不知道如何生成如下输出。对于每个资产 + Metric_Code 对，我也需要填写 YearMonth 和 YearMonthRank 列。

之后我就可以添加我的预测栏了！

Asset	Metric_Code	Metric_V	Trend	YearMonth	YearMonthRank
RDX4	AXCP32	88.0853212	89.0853212	202103	1
RDX4	AXCP32	88.0853212	85.0853212	202106	2
RDX4	AXCP32			202109	3
RDX4	AXCP32			202112	4
RDX4	AXCP32			202203	5
RDX4	AXCP32			202206	6
RDX4	AXCP32			202209	7
RDX4	AXCP32			202212	8
RDX4	AXCP34	79.41087028	81.0853212	202103	1
RDX4	AXCP34	70.44587	69.0853212	202106	2
RDX4	AXCP34			202109	3
RDX4	AXCP34			202112	4
RDX4	AXCP34			202203	5
RDX4	AXCP34			202206	6
RDX4	AXCP34			202209	7
RDX4	AXCP34			202212	8
MRDX4	AXCP32	88.0853212	89.0853212	202103	1
MRDX4	AXCP32	88.0853212	85.0853212	202106	2
MRDX4	AXCP32			202109	3
MRDX4	AXCP32			202112	4
MRDX4	AXCP32			202203	5
MRDX4	AXCP32			202206	6
MRDX4	AXCP32			202209	7
MRDX4	AXCP32			202212	8
MRDX4	AXCP34	79.41087028	81.0853212	202103	1
MRDX4	AXCP34	70.44587	81.0853212	202106	2
MRDX4	AXCP34			202109	3
MRDX4	AXCP34			202112	4
MRDX4	AXCP34			202203	5
MRDX4	AXCP34			202206	6
MRDX4	AXCP34			202209	7
MRDX4	AXCP34			202212	8

Answer 1

考虑一下：

For each Asset + Metric_Code couple , i need to fill the YearMonth & YearMonthRank columns too

为此，下面的查询首先通过 CROSS JOIN 为每个（Asset、Metric_Code）组合扩展 PROJECTION table。

CREATE TEMP TABLE TREND (
  Asset STRING,
  Metric_Code STRING,
  Metric_V FLOAT64,
  Trend FLOAT64,
  YearMonth INT64,
  YearMonthRank INT64
);
INSERT INTO TREND
VALUES ('RDX4', 'AXCP32', 88.0853212, 89.0853212, 202103,1),
       ('RDX4', 'AXCP32', 88.0853212,85.0853212, 202106,2),
       ('RDX4', 'AXCP34', 79.41087028,81.0853212, 202103,3),
       ('RDX4', 'AXCP34', 70.44587,69.0853212, 202106,4),
       ('MRDX4', 'AXCP32', 88.0853212, 89.0853212, 202103,1),
       ('MRDX4', 'AXCP32', 88.0853212,85.0853212, 202106,2),
       ('MRDX4', 'AXCP34', 79.41087028,81.0853212, 202103,3),
       ('MRDX4', 'AXCP34', 70.44587, 69.0853212,202106,4)
;

-- SQL CODE TO RECREATE MY PROJECTION TABLE
CREATE TEMP TABLE PROJECTION (
  Asset STRING,
  Metric_Code STRING,
  Metric_V FLOAT64,
  Trend FLOAT64,
  YearMonth INT64,
  YearMonthRank INT64
);
INSERT INTO PROJECTION
VALUES (NULL,NULL,NULL,NULL, 202103,1),
       (NULL,NULL,NULL,NULL, 202106,2),
       (NULL,NULL,NULL,NULL, 202109,3),
       (NULL,NULL,NULL,NULL, 202112,4),
       (NULL,NULL,NULL,NULL, 202203,5),
       (NULL,NULL,NULL,NULL, 202206,6),
       (NULL,NULL,NULL,NULL, 202209,7),
       (NULL,NULL,NULL,NULL, 202212,8)
;

SELECT B.Asset_Key AS Asset,
       B.Metric_Code_Key AS Metric_Code,
       C.Metric_V,
       C.Trend,
       A.YearMonth,
       A.YearMonthRank
  FROM PROJECTION A
 CROSS JOIN (SELECT DISTINCT Asset AS Asset_Key, Metric_Code AS Metric_Code_Key FROM TREND) B
  LEFT JOIN TREND C
    ON B.Asset_Key = C.Asset AND B.Metric_Code_Key = C.Metric_Code AND A.YearMonthRank = C.YearMonthRank
 ORDER BY 1, 2, A.YearMonthRank;

如何合并两个满足最小值的字段的表？

How to merge two tables fullfilling a field with his minimum value?

google-bigquery