Question

我的数据传播到3个字符串列; col 1, col 2, col 3。数据集大约有500k行，每月添加一行。

Row setting                     acceptance_rate    undergrads    
1   City                        N/A                2773  
2   198                         Town               77%
3   133                         Suburban           56%
4   55%                         254                Suburban  
5   54%                         Rural              46    
6   63%                         City               247   
7   100%                        210                Rural

我想为特定条件创建2个新列到组号。我希望新列acceptance_rate_new包含介于0和1之间的所有数字，而population的数字大于1.我认为以下CASE ... WHEN足以完成此任务它适用于字符串到字符串，但这次不起作用。我想我需要每个月运行一次查询。

SELECT _name, COALESCE(
  CASE WHEN INTEGER(col1) > 1 THEN INTEGER(col1) ELSE NULL END, 
  CASE WHEN INTEGER(col2) > 1 THEN INTEGER(col2) ELSE NULL END, 
  CASE WHEN INTEGER(col3) > 1 THEN INTEGER(col3) ELSE NULL END
  ) AS population_new
FROM

Answer 1

下面是BigQuery Standard SQL和＆＃34;重新分发＆＃34;相应列的正确值（相对于列的当前随机分布）
使用它你可以把你需要的任何逻辑放在首位

#standardSQL
WITH `yourproject.yourdataset.yourtable` AS (
  SELECT 'City' setting, 'N/A' acceptance_rate, '2773' undergrads UNION ALL
  SELECT '198', 'Town', '77%' UNION ALL
  SELECT '133', 'Suburban', '56%' UNION ALL
  SELECT '55%', '254', 'Suburban' UNION ALL
  SELECT '54%', 'Rural', '46' UNION ALL
  SELECT '63%', 'City', '247' UNION ALL
  SELECT '100%', '210', 'Rural' 
)
SELECT setting, acceptance_rate, undergrads,
  COALESCE(
    IF(setting IN (undergrads_correct, acceptance_rate_correct, 'N/A'), NULL, setting),
    IF(acceptance_rate IN (undergrads_correct, acceptance_rate_correct, 'N/A'), NULL, acceptance_rate),
    IF(undergrads IN (undergrads_correct, acceptance_rate_correct, 'N/A'), NULL, undergrads)
  ) setting_correct,
  CAST(undergrads_correct AS INT64) undergrads_correct,
  acceptance_rate_correct
FROM
(
  SELECT *,
    COALESCE(
      IF(NOT IS_NAN(SAFE_CAST(setting AS INT64)), setting, NULL),
      IF(NOT IS_NAN(SAFE_CAST(acceptance_rate AS INT64)), acceptance_rate, NULL),
      IF(NOT IS_NAN(SAFE_CAST(undergrads AS INT64)), undergrads, NULL)
    ) AS undergrads_correct,
    COALESCE(
      IF(REGEXP_CONTAINS(setting, '%$') AND NOT IS_NAN(SAFE_CAST(REPLACE(setting, '%', '') AS INT64)), setting, NULL),
      IF(REGEXP_CONTAINS(acceptance_rate, '%$') AND NOT IS_NAN(SAFE_CAST(REPLACE(acceptance_rate, '%', '') AS INT64)), acceptance_rate, NULL),
      IF(REGEXP_CONTAINS(undergrads, '%$') AND NOT IS_NAN(SAFE_CAST(REPLACE(undergrads, '%', '') AS INT64)), undergrads, NULL)
    ) acceptance_rate_correct
  FROM `yourproject.yourdataset.yourtable`
)

结果是

setting   acceptance_rate   undergrads  setting_correct undergrads_correct  acceptance_rate_correct  
City      N/A               2773        City                          2773  null     
63%       City              247         City                           247  63%  
198       Town              77%         Town                           198  77%  
54%       Rural             46          Rural                           46  54%  
100%      210               Rural       Rural                          210  100%     
133       Suburban          56%         Suburban                       133  56%  
55%       254               Suburban    Suburban                       254  55%

以下是使用SQL UDF的版本（当然具有相同的输出）

#standardSQL
CREATE TEMP FUNCTION check_int(val STRING) AS (
  (IF(NOT IS_NAN(SAFE_CAST(val AS INT64)), val, NULL))
);
CREATE TEMP FUNCTION check_rate(val STRING) AS (
  IF(REGEXP_CONTAINS(val, '%$') AND NOT IS_NAN(SAFE_CAST(REPLACE(val, '%', '') AS INT64)), val, NULL)
);
CREATE TEMP FUNCTION check_city(val STRUCT<setting STRING, acceptance_rate STRING, undergrads STRING, undergrads_correct STRING, acceptance_rate_correct STRING>) AS (
  COALESCE(
    IF(val.setting IN (val.undergrads_correct, val.acceptance_rate_correct, 'N/A'), NULL, val.setting),
    IF(val.acceptance_rate IN (val.undergrads_correct, val.acceptance_rate_correct, 'N/A'), NULL, val.acceptance_rate),
    IF(val.undergrads IN (val.undergrads_correct, val.acceptance_rate_correct, 'N/A'), NULL, val.undergrads)
  ) 
);
WITH `yourproject.yourdataset.yourtable` AS (
  SELECT 'City' setting, 'N/A' acceptance_rate, '2773' undergrads UNION ALL
  SELECT '198', 'Town', '77%' UNION ALL
  SELECT '133', 'Suburban', '56%' UNION ALL
  SELECT '55%', '254', 'Suburban' UNION ALL
  SELECT '54%', 'Rural', '46' UNION ALL
  SELECT '63%', 'City', '247' UNION ALL
  SELECT '100%', '210', 'Rural' 
)
SELECT setting, acceptance_rate, undergrads,
  check_city(STRUCT(setting, acceptance_rate, undergrads, undergrads_correct, acceptance_rate_correct)) setting_correct,
  CAST(undergrads_correct AS INT64) undergrads_correct,
  acceptance_rate_correct
FROM (
  SELECT *,
    COALESCE(check_int(setting),check_int(acceptance_rate),check_int(undergrads)) AS undergrads_correct,
    COALESCE(check_rate(setting),check_rate(acceptance_rate),check_rate(undergrads)) acceptance_rate_correct
  FROM `yourproject.yourdataset.yourtable`
)

更新 - would I manually need to enter values for each row one by one?

您不需要手动输入所有数据 - 而是应删除包含所有虚拟数据的WITH部分，并将您的项目替换为您的项目。您可以使用自己的项目进行替换 - 类似于以下

#standardSQL
SELECT setting, acceptance_rate, undergrads,
  COALESCE(
    IF(setting IN (undergrads_correct, acceptance_rate_correct, 'N/A'), NULL, setting),
    IF(acceptance_rate IN (undergrads_correct, acceptance_rate_correct, 'N/A'), NULL, acceptance_rate),
    IF(undergrads IN (undergrads_correct, acceptance_rate_correct, 'N/A'), NULL, undergrads)
  ) setting_correct,
  CAST(undergrads_correct AS INT64) undergrads_correct,
  acceptance_rate_correct
FROM
(
  SELECT *,
    COALESCE(
      IF(NOT IS_NAN(SAFE_CAST(setting AS INT64)), setting, NULL),
      IF(NOT IS_NAN(SAFE_CAST(acceptance_rate AS INT64)), acceptance_rate, NULL),
      IF(NOT IS_NAN(SAFE_CAST(undergrads AS INT64)), undergrads, NULL)
    ) AS undergrads_correct,
    COALESCE(
      IF(REGEXP_CONTAINS(setting, '%$') AND NOT IS_NAN(SAFE_CAST(REPLACE(setting, '%', '') AS INT64)), setting, NULL),
      IF(REGEXP_CONTAINS(acceptance_rate, '%$') AND NOT IS_NAN(SAFE_CAST(REPLACE(acceptance_rate, '%', '') AS INT64)), acceptance_rate, NULL),
      IF(REGEXP_CONTAINS(undergrads, '%$') AND NOT IS_NAN(SAFE_CAST(REPLACE(undergrads, '%', '') AS INT64)), undergrads, NULL)
    ) acceptance_rate_correct
  FROM `myproject.mydataset.mytable`
)

多个字符串列中的新整数列

1 个答案: