Recursively tracking state of customers (Presto SQL)

问题

I have one table with the current state_id of my customers and another table holding all states and their state_ids, but without the corresponding customer_id. However, the historical state table holds the information of which state_id it replaced. Hence, it should be possible to recursively track the states/journey of the customer.

Consider the following example:

"Customer" table:

customer_id    state_created      current_state_id
1              2017-11-09         33
2              2018-04-01         243
3              2018-07-10         254

"Historical_state" table:

state_name     replace_state_id   state_id           state_created
State1                            22                 2015-10-08
State1                            211                2017-06-28
State3                            254                2018-07-10
State4         211                243                2018-04-01
State5         22                 33                 2017-11-09

I am interested in getting the historical state information on every customer, ie. the following table:

customer_id    state_created      state_name       
1              2015-10-00         State1
1              2017-11-09         State5
2              2017-06-28         State1
2              2018-04-01         State4
3              2018-07-10         State3

The data is replaced in AWS' Athena, so the presto sql should be used as the language.

回答1:

Here is one option using a union. The crux behind this query is that we generate a single logical state_id column to which we join the customer table. This table contains both current and replace state_id values for a given state.

SELECT
    c.customer_id,
    t.state_created,
    t.state_name
FROM Customer c
INNER JOIN
(
    SELECT state_id, state_name, state_created
    FROM Historical_state
    UNION ALL
    SELECT h1.state_id, h2.state_name, h2.state_created
    FROM Historical_state h1
    INNER JOIN Historical_state h2
        ON h1.replace_state_id = h2.state_id
) t
    ON c.current_state_id = t.state_id;

Here is a demo in MySQL, because Rextester does not support SQLite, but it at least shows that the query logic is correct.

Demo

来源：https://stackoverflow.com/questions/51823728/recursively-tracking-state-of-customers-presto-sql

标签

sql

amazon-athena

presto