标准 sql：将数组中的字符串转换为 float64

Question

我正在尝试将字段 GCAM_value 和 Themes_value 转换为数字，例如 float64。

我试过 CAST(regexp_extract(x, r'^(.*?):') as FLOAT64) 但当代码遇到 NULL 值时收到错误消息。

#standardSQL

SELECT
  GKGRECORDID,
  DATE, 
  ARRAY(
    SELECT regexp_extract(x, r'^(.*?)\:')
    FROM UNNEST(split(GCAM,',')) AS x
    WHERE regexp_extract(x, r'^(.*?)\:') IS NOT NULL
  ) AS GCAM_field,
   ARRAY(
    SELECT regexp_extract(x, r'[^:]*$')
    FROM UNNEST(split(GCAM,',')) AS x
    WHERE regexp_extract(x, r'[^:]*$') IS NOT NULL
  ) AS GCAM_value,

    ARRAY(
    SELECT regexp_extract(x, r'^(.*?)\,')
    FROM UNNEST(split(V2Themes,';')) AS x
    WHERE regexp_extract(x, r'^(.*?)\,') IS NOT NULL
  ) AS Theme_field,
   ARRAY(
    SELECT regexp_extract(x, r'[^,]*$')
    FROM UNNEST(split(V2Themes,';')) AS x
    WHERE regexp_extract(x, r'[^,]*$') IS NOT NULL
  ) AS Theme_value,

  TranslationInfo,
  V2Tone,
  SourceCollectionIdentifier,
  SourceCommonName,
  DocumentIdentifier
FROM `gdelt-bq.gdeltv2.gkg_partitioned`
WHERE _PARTITIONTIME BETWEEN TIMESTAMP('2019-02-02') AND TIMESTAMP('2019- 
02-02')

我希望 GCAM_value 和 Themes_value 的数组项采用某种数字格式，如 float64。

Answer 1

下面应该是您不错的起点

#standardSQL
SELECT
  GKGRECORDID,
  DATE, 
  ARRAY(
    SELECT REGEXP_EXTRACT(x, r'^(.*?)\:')
    FROM UNNEST(SPLIT(GCAM,',')) AS x
    WHERE REGEXP_EXTRACT(x, r'^(.*?)\:') IS NOT NULL
  ) AS GCAM_field,
  ARRAY(
    SELECT CAST(REGEXP_EXTRACT(x, r'[^:]*$') AS FLOAT64)
    FROM UNNEST(SPLIT(GCAM,',')) AS x
    WHERE IFNULL(REGEXP_EXTRACT(x, r'[^:]*$'), '') != ''
  ) AS GCAM_value,
  ARRAY(
    SELECT REGEXP_EXTRACT(x, r'^(.*?)\,')
    FROM UNNEST(SPLIT(V2Themes,';')) AS x
    WHERE REGEXP_EXTRACT(x, r'^(.*?)\,') IS NOT NULL
  ) AS Theme_field,
  ARRAY(
    SELECT CAST(REGEXP_EXTRACT(x, r'[^,]*$') AS FLOAT64)
    FROM UNNEST(SPLIT(V2Themes,';')) AS x
    WHERE IFNULL(REGEXP_EXTRACT(x, r'[^,]*$'), '') != ''
  ) AS Theme_value,
  TranslationInfo,
  V2Tone,
  SourceCollectionIdentifier,
  SourceCommonName,
  DocumentIdentifier
FROM `gdelt-bq.gdeltv2.gkg_partitioned`
WHERE _PARTITIONTIME BETWEEN TIMESTAMP('2019-02-02') AND TIMESTAMP('2019-02-02')

它 returns REPEATED FLOAT 而不是 REPEATED STRING 各个字段

标准 sql：将数组中的字符串转换为 float64

standard sql: casting strings in array as float64

google-bigquery

google-cloud-datalab