make recent log agg with unix timestamp in BigQuery - sql

Question

I'm making a webpage user access log table in BigQuery.
But I don't know how agg log with exit_time.
I want recent 5 view page log array.
If view page log is less than 5, the array is shorter.
plz teach me.
USER_EXIT_LOG AS (
SELECT
['A','A','B'] AS user_id,
[1000,2000,1000] AS exit_time
),
VIEW_PAGE_LOG AS (
SELECT
['A','A','A','A','A','B','B'] AS user_id,
['a','b','c','d','e','a','b'] AS view_page_id,
[800,900,1800,1900,1950,800,900] AS time
)
USER_EXIT_LOG
| user_id | exit_time |
| -------- | --------- |
| A | 1000 |
| A | 2000 |
| B | 1000 |
VIEW_PAGE_LOG
| user_id | view_page_id | view_time |
| -------- | ----------- | --------- |
| A | a | 800 |
| A | b | 900 |
| A | c | 1800 |
| A | d | 1900 |
| B | a | 800 |
| B | b | 900 |
WHAT I WANT
| user_id | exit_time | view_page_array |
| ------- | --------- | --------------- |
| A | 1000 | [b,a] |
| A | 2000 | [e,d,c] |
| B | 1000 | [b,a] |

Answer 1

Consider below
select u.user_id, exit_time,
array_agg(view_page_id order by view_time desc limit 5) as view_page_array
from (
select user_id, exit_time,
ifnull(lag(exit_time + 1) over(partition by user_id order by exit_time), 0) start_time
from `project.dataset.USER_EXIT_LOG`
) u
join `project.dataset.VIEW_PAGE_LOG` v
on u.user_id = v.user_id
and view_time between start_time and exit_time
group by user_id, exit_time
If applied t sample data in your question
with `project.dataset.USER_EXIT_LOG` as (
select 'A' user_id, 1000 exit_time union all
select 'A', 2000 union all
select 'B', 1000
), `project.dataset.VIEW_PAGE_LOG` as (
select 'A' user_id, 'a' view_page_id, 800 view_time union all
select 'A', 'b', 900 union all
select 'A', 'c', 1800 union all
select 'A', 'd', 1900 union all
select 'A', 'e', 1950 union all
select 'B', 'a', 800 union all
select 'B', 'b', 900
)
the output is