如何将一列中的字符串解析为 SQL 中的分隔值

Question

This is my column in Redshift

SHIPMENT_ID
-----------------------------------------
FBA15KS66741, FBA15KS6673D
FBA15NHV7PXX (Oct 20th)
FBA15XNW0SWY 27 balance 2 of 2
FBA15M575MDL &  FBA15M59W1Y5
FBA15NHV7PXX (Oct 20th)
FBA15D7WPZVR /FBA15D7WWTPK/FBA15D7WW1GL

I would like to make it

SHIPMENT_ID
-----------------------------------------
FBA15KS66741, FBA15KS6673D
FBA15NHV7PXX
FBA15XNW0SWY
FBA15M575MDL, FBA15M59W1Y5
FBA15NHV7PXX
FBA15D7WPZVR, FBA15D7WWTPK, FBA15D7WW1GL

In SQL only, what is the best way to handle this?

Answer 1

This works in PostgreSQL, so may work in Redshift depending on feature availability in PG8.

WITH items AS
(
  SELECT shipment_id,
         ARRAY_TO_STRING(REGEXP_MATCHES(shipment_id,'FBA15[0-9a-zA-z]{7}','g'),'') AS unique_shipment_ids
  FROM dat
)
SELECT shipment_id,
       STRING_AGG(unique_shipment_ids,',') AS shipment_id_csv
FROM items
GROUP BY shipment_id;

I've assumed:

• Each item begins with the characters 'FBA15'
• There are exactly 7 characters after the first 5

You can edit the regexp pattern if my assumptions are incorrect.

The approach is:

1. Use REGEXP_MATCHES to capture each item within each row. This creates multiple rows per unique value in shipment_id
2. Use ARRAY_TO_STRING to convert those values to text , rather than text[]
3. Use STRING_AGG to join them back together with a comma separator

I found that I could not use STRING_AGG directly around REGEXP_MATCHES as I get the error aggregate function calls cannot contain set-returning function calls , so opted for a CTE. I assume a subquery would work as well.