SQL 比较同一张表多行记录

Question

I have table with: Port_nbr, Pair, length, add_date. example:

port_nbr        | 4
pair            | pairC
length          | 33.12
add_date        | 2020-06-16 00:01:13.237164

There is 4 pairs intotal ( A,B,C,D) and each port_nbr has a pair(AD) atleast once. It means, each port_nbr can be multiple times. I need to find out if that port any pair length is increased by 30m. To do it, i must order each port_nbr by date, because i need to compare 1 record with next record and find out when it increased by 30m

example: RECORD 1

port_nbr | 1                          
pair     | pairA
length   | 30.00
add_date | 2020-06-16 00:01:13.237164

RECRORD 2

port_nbr | 1                              
pair     | pairA
length   | 65.00
add_date | 2020-06-16 00:02:13.237164

expected output- Length increased on: port 1, pairA, 2020-06-16 00:02:13.237164 by 35m

There is hundreds of records and i have to compare one to another record, but it has to be same port and pair.

Everything i have right now is:

Select *
from diags a join
     (SELECT port_nbr, count(*) from diags group by port_nbr having count(*) > 1 ) b
     on a.port_nbr = b.port_nbr
order by a.port_nbr

It would help a lot if someone could explain what i should to try add to get expected result.

Answer 1

I think you just want lead() :

select d.*
from (select d.*,
             lead(length) over (partition by port_nbr, pair order by add_date) as next_length
      from diags d
     ) d
where next_length > length + 30;

This gets the next length for the same part/pair combo and returns rows where the next value exceeds your threshold.

Answer 2

With LAG() window function:

select port_nbr, pair, length, add_date, diff
from (
  select *, length - lag(length) over (partition by port_nbr, pair order by add_date) diff
 from tablename
) t
where diff > 30