Combine multiple rows with different dates with overlapping variables (to capture first and last change dates)

问题

I have the following data represented in a table like this:

User	Type	Date
A	Mobile	2019-01-10
A	Mobile	2019-01-20
A	Desktop	2019-03-01
A	Desktop	2019-03-20
A	Email	2021-01-01
A	Email	2020-01-02
A	Desktop	2021-01-03
A	Desktop	2021-01-04
A	Desktop	2021-01-05

Using PostgreSQL - I want to achieve the following:

User	First_Type	First Type Initial Date	Last_Type	Last_Type_Initial_Date
A	Mobile	2019-01-10	Desktop	2021-01-03

So for each user, I want to capture the initial date and type but then also, on the same row (but diff columns), have their last type they "switched" to but with the first date the switch occurred and not the last record of activity on that type.

回答1:

Consider using a LAG window function and conditional aggregation join via multiple CTEs and self-joins:

WITH sub AS (
  SELECT "user"
       , "type"
       , "date"
       , CASE 
            WHEN LAG("type") OVER(PARTITION BY "user" ORDER BY "date") = "type"
            THEN 0
            ELSE 1
         END "shift"
  FROM myTable 
), agg AS (
   SELECT "user"
         , MIN(CASE WHEN shift = 1 THEN "date" END) AS min_shift_dt
         , MAX(CASE WHEN shift = 1 THEN "date" END) AS max_shift_dt
   FROM sub
   GROUP BY "user"
)


SELECT agg."user"
     , s1."type" AS first_type
     , s1."date" AS first_type_initial_date
     , s2."type" AS last_type
     , s2."date" AS last_type_initial_date
FROM agg
INNER JOIN sub AS s1
  ON agg."user" = s1."user"
  AND agg.min_shift_dt = s1."date"
  
INNER JOIN sub AS s2
  ON agg."user" = s2."user"
  AND agg.max_shift_dt = s2."date"

Online Demo

user	first_type	first_type_initial_date	last_type	last_type_initial_date
A	Mobile	2019-01-10 00:00:00	Desktop	2021-01-03 00:00:00

回答2:

Here is my solution with only windows functions and no joins:

with 
  prep as (
select *, 
  lag("Type") over(partition by "User" order by "Date") as "Lasttype"
from your_table_name
)
select distinct "User", 
  first_value("Type") over(partition by "User") as "First_Type", 
  first_value("Date") over(partition by "User") as "First_Type_Initial_Date", 
  last_value("Type") over(partition by "User") as "Last_Type", 
  last_value("Date") over(partition by "User") as "Last_Type_Initial_Date"
from prep 
where "Type" <> "Lasttype" or "Lasttype" is null
;

回答3:

I think this will work, but it sure feels ugly. There might be a better way to do this.

SELECT a.User, a.Type AS First_Type, a.Date AS FirstTypeInitialDate, b.Type AS Last_Type, b.LastTypeInitialDate
FROM table a
INNER JOIN table b ON a.User = b.User
WHERE a.Date = (SELECT MIN(c.Date) FROM table c WHERE c.User = a.User)
    AND b.Date = (SELECT MIN(d.Date) FROM table d WHERE d.User = b.User
       AND d.Type = (SELECT e.Type FROM table e WHERE e.User = d.User 
          AND e.Date = (SELECT MAX(f.Date) FROM table f WHERE f.User = e.User)))

来源：https://stackoverflow.com/questions/65585137/combine-multiple-rows-with-different-dates-with-overlapping-variables-to-captur

标签

sql

postgresql