获取上一个和下一个值计数

Question

I have a table which contains set of events. The events can be search, click, open etc. There is a date column associated with each row.

I'm trying to obtain count of rows where the first event was search and next event was click. Is there any way to achieve this using lead/lag in redshift ?

It throws me an error that aggregate functions calls may not have nested function

count(CASE WHEN lead(trim(lower(event)), 1)
                  OVER (
                    ORDER BY timevalue :: TIMESTAMP ASC) = 'click' AND trim(lower(event)) = 'search'
    THEN 1
        ELSE NULL END)              AS my_column,

Answer 1

I would expect your code to work, using a subquery. I would write it as:

select count(*)
from (select t.*,
             lead(event) over (order by timevalue) as next_event
      from t
     ) t
where event = 'search' and next_event = 'click';

I would not expect the lower() and trim() to be necessary, but if your data is really dirty, they might be.

Note: You cannot put a window function as an argument to an aggregation function, so you need to use a subquery.