sql bigquery: extract string after “:” from nested field

Question

I am trying to extract all values after ":" from a nested field in Google BigQuery and add them to a separate column. However, the gets returned empty.

I think the problem is with the regex expression used, r'^:(.*?)'

SELECT
  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
FROM `gdelt-bq.gdeltv2.gkg_partitioned`
WHERE _PARTITIONTIME BETWEEN TIMESTAMP('2019-02-02') AND TIMESTAMP('2019-02-02')

The expected result is a column "GCAM_field" with all values before ":" and a column "GCAM_value" with all values after ":". However the latter gets returned empty.

Answer 1

#standardSQL
SELECT
  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
FROM `gdelt-bq.gdeltv2.gkg_partitioned`
WHERE _PARTITIONTIME BETWEEN TIMESTAMP('2019-02-02') AND TIMESTAMP('2019-02-02')  

or simpler:

#standardSQL
SELECT
  ARRAY(
    SELECT SPLIT(x, ':')[SAFE_OFFSET(0)]
    FROM UNNEST(SPLIT(GCAM,',')) AS x
    WHERE SPLIT(x, ':')[SAFE_OFFSET(0)] IS NOT NULL
  ) AS GCAM_field,
   ARRAY(
    SELECT SPLIT(x, ':')[SAFE_OFFSET(1)]
    FROM UNNEST(SPLIT(GCAM,',')) AS x
    WHERE SPLIT(x, ':')[SAFE_OFFSET(1)] IS NOT NULL
  ) AS GCAM_value
FROM `gdelt-bq.gdeltv2.gkg_partitioned`
WHERE _PARTITIONTIME BETWEEN TIMESTAMP('2019-02-02') AND TIMESTAMP('2019-02-02')

Answer 2

one more way to solve this problem

with

sample_data as (
    select
        *
    from
        unnest(
            array[
                struct(1 as id, 'bbb:111,aaa:222' as gcam),
                struct(2 as id, 'qqq:,k:3,:777,xxx:555:&&&' as gcam)
            ]
        )
)

select
    regexp_extract_all(
        concat(',', gcam, ','), r',(.*?)\:') as gcam_field,
    regexp_extract_all(
        concat(',', gcam, ','), r'\:(.*?),') as gcam_value
from
    sample_data