如何在 Bigquery 中使用数组/结构/json 将五列混合为一列

Question

I have a table where I have a primary key (cust_id) in one column and five other columns with phone numbers, where I might have repeated or null values.

I would need to have all phones in a single column and in a single row for each customer. I am not sure what format is best for this. I was thinking about an ARRAY or JSON. What would you do? I would like to have something like this:

I need to have all phones, with no null nor duplicates in a string in a way I can access the phone easly. I think having them in a JSON string would be nice to use JSON_EXTRACT_SCALAR.

I have thied something like this:

WITH phones as (
SELECT '6446177' as id, NULL as phone1,NULL as phone2,11572946 as phone3,NULL as phone4,11572946 as phone5 
UNION ALL
SELECT '8523122' as id, NULL as phone1,NULL as phone2,29703165 as phone3,NULL as phone4,29703165 as phone5 
UNION ALL
SELECT '5606494' as id, NULL as phone1,NULL as phone2,51156520 as phone3,NULL as phone4,32247153 as phone5
UNION ALL
SELECT '6560607' as id, 85137 as phone1,NULL as phone2,22185137 as phone3,NULL as phone4,93361637 as phone5),

telefonos_pivot as (
SELECT id,phone1 as phone from phones
UNION ALL
SELECT id,phone2 as phone from phones
UNION ALL
SELECT id,phone3 as phone from phones
UNION ALL
SELECT id,phone4 as phone from phones
UNION ALL
SELECT id,phone5 as phone from phones),

telefonos_pivot_clean as (
SELECT distinct * from telefonos_pivot
WHERE id IS NOT NULL AND phone IS NOT NULL)
SELECT id,array_agg(phone) as phones FROM telefonos_pivot_clean
GROUP BY id

The results are not as I expected:

Yes, information is avaliable and easy to access, but this format would not fit well in my project. I need all phones in the same string.

There is any way to make these columns into a JSON string or something like that?

Thank you!

Answer 1

Consider below approach

select id, 
  array_to_string(array(
    select distinct phone
    from unnest(regexp_extract_all(to_json_string(t), r'"phone\d+":(\d+)')) phone
  ), ', ') phones,
from `project.dataset.phones` t             

if applied to sample phones data in your question - output is