将具有不同记录的列拆分为bigquery中的行

Question

我有一张表，其中多列中包含单个值和逗号分隔值，如下表所示，

`No  Records value1 value2 value3 value4 value5
--- ------- ------ ------ ------ ------ -------
1.  A,B    F,G   D,E    Noval  Noval  Z,U
2.  Z      X     Noval  P,Q    S,T    Noval`

所需的输出必须如下所示， `

No  Records value1 value2 value3 value4 value5 
--- ------- ------ ------ ------ ------ -------
1.  A       F      D      Noval  Noval   Z
1.  B       G      E      Noval  Noval   U
2.  Z       X      Noval  P      S      Noval
2.  Z       X      Noval  Q      T     Noval

` 使用了 unnest 但不知道，因为我是 bigquery 的新手

Answer 1

首先每列需要用逗号分隔。然后不安就开始发挥作用。如果对每列完成取消嵌套，则会创建所有组合。通过使用偏移量，可以消除不需要的组合。

WITH
  tbl AS (
  SELECT
    SPLIT('A,B') AS Records,
    SPLIT('F,G') AS value1)
SELECT
  *
FROM
  tbl,
  UNNEST(Records) AS Records_name WITH OFFSET Records_id,
  UNNEST(value1) AS value1_name WITH OFFSET value1_id
WHERE Records_id=value1_id

但是，如果存在不同长度的数组，这将会失败。因此，我们为数字 0 到 10.000 生成一个数组

tmp

。对于每个数字，我们获取数组的条目。如果数字大于数组大小，我们就取最后一个条目。

IF(tmp>=ARRAY_LENGTH(value1),value1[SAFE_OFFSET(ARRAY_LENGTH(value1)-1)], value1[SAFE_OFFSET(tmp)]) as value1,

这可以更好地写为

value1[SAFE_OFFSET(LEAST(ARRAY_LENGTH(value1)-1,tmp))] AS value1,

WHERE

条件会删除不需要的条目，其中

tmp

大于最大数组大小。

WITH
  tbl AS (
  SELECT
    1 AS row_num,
    'A,B' AS Records,
    'F,G' AS value1,
    'D,E' AS value2,
    'Noval' AS value3,
    'Noval' AS value4,
    'Z,U' AS value5
  UNION ALL
  SELECT
    2,
    'Z',
    'X',
    'Noval',
    'P,Q',
    'S,T',
    'Noval' ),
  tbl1 AS (
  SELECT
    row_num,
    SPLIT(Records) AS Records,
    SPLIT(value1) AS value1,
    SPLIT(value2) AS value2,
    SPLIT(value3) AS value3,
    SPLIT(value4) AS value4,
    SPLIT(value5) AS value5,
  FROM
    tbl ),
  tbl2 AS (
  SELECT
    row_num,
    tmp,
    Records[SAFE_OFFSET(LEAST(ARRAY_LENGTH(Records)-1,tmp))] AS value1,
    value1[SAFE_OFFSET(LEAST(ARRAY_LENGTH(value1)-1,tmp))] AS value1,
    value2[SAFE_OFFSET(LEAST(ARRAY_LENGTH(value2)-1,tmp))] AS value2,
    value3[SAFE_OFFSET(LEAST(ARRAY_LENGTH(value3)-1,tmp))] AS value3,
    value4[SAFE_OFFSET(LEAST(ARRAY_LENGTH(value4)-1,tmp))] AS value4,
    value5[SAFE_OFFSET(LEAST(ARRAY_LENGTH(value5)-1,tmp))] AS value5,
  FROM
    tbl1,
    UNNEST(GENERATE_ARRAY(0,10000)) AS tmp
  WHERE
    tmp<GREATEST(ARRAY_LENGTH(value1),ARRAY_LENGTH(value2),ARRAY_LENGTH(value3),ARRAY_LENGTH(value4),ARRAY_LENGTH(value5)) )
SELECT
  *
FROM
  tbl2

Answer 2

与此类似应该可以工作。

SELECT
  number,
  records,
  value
FROM
  gaurang.so_q,
  UNNEST(SPLIT(records, ",")) records,
  UNNEST(SPLIT(value, ",")) value ;