Update rows in SQL every 20 iterations

Question

I have a table with ~1 million rows in Google BigQuery from the NYC Yellow TaxiCab public dataset. As you can see from that link, the schema does not have a primary key. Each row represents a trip/transaction, but there is no customer_id field.

I want to add a column customer_id and distribute random numbers to it such that:

For rows 1-20, `customer_id` should be assigned `1`
For rows 21-40, `customer_id` should be assigned `2`
and so on..

In other words, I want exactly (and any) 20 rows in the table to have a specific value of customer_id.

Answer 1

Assign each row a random id, getting groups of ~20 rows for each new_id:

#standardSQL
SELECT CAST(FLOOR(COUNT(*) OVER()/20*RAND()) AS INT64) new_id, *
FROM (
  SELECT login
  FROM `ghtorrent-bq.ght_2017_04_01.users`
  LIMIT 1000000
)

Proof that 50,000 "customers_ids" are generated when going over a million rows:

Answer 2

Below is for BigQuery Standard SQL and produces exactly 20 entries per customer_id

#standardSQL
SELECT DIV(ROW_NUMBER() OVER() - 1, 20) AS customer_id, *
FROM `yourTable`
-- ORDER BY customer_id

You can test play with this using dummy data as below

#standardSQL
WITH `yourTable` AS (
    SELECT login
    FROM `ghtorrent-bq.ght_2017_04_01.users`
    LIMIT 1000000
)
SELECT DIV(ROW_NUMBER() OVER() - 1, 20) AS customer_id, *
FROM `yourTable`
-- ORDER BY customer_id  

Moreover - below query shows the distribution of counts per customer_id

#standardSQL
WITH `yourTable` AS (
    SELECT login
    FROM `ghtorrent-bq.ght_2017_04_01.users`
    LIMIT 1000000
)
SELECT cnt, COUNT(1) AS distribution FROM (
  SELECT customer_id, COUNT(1) AS cnt FROM (
    SELECT *, DIV(ROW_NUMBER() OVER() - 1, 20) AS customer_id
    FROM `yourTable`
    ORDER BY customer_id
  )
  GROUP BY customer_id
)
GROUP BY cnt
ORDER BY cnt   

with output as below

Row cnt distribution     
--- --- ------------
1    20        50000