PostgreSQL: update table ensuring rows have unique timestamp (no duplicate unix timestamp) - sql

Question

I am having a table of GPS traces with Unix timestamp as shown below:
SELECT * FROM mytable LIMIT 10;
id | lat | lon | seconds | speed
-----------+------------+------------+------------+-------
536889001 | 41.1794675 | -8.6017187 | 1460465697 | 1.25
536889001 | 41.1794709 | -8.601675 | 1460465698 | 2
536889001 | 41.1794636 | -8.6016337 | 1460465700 | 1.25
536889001 | 41.1794468 | -8.6016014 | 1460465700 | 2.5
536889001 | 41.1794114 | -8.6015662 | 1460465701 | 3.5
536889001 | 41.1794376 | -8.6015672 | 1460465703 | 1.5
536889001 | 41.17944 | -8.6015516 | 1460465703 | 1.5
536889001 | 41.1794315 | -8.6015353 | 1460465704 | 1.5
536889001 | 41.1794367 | -8.6015156 | 1460465705 | 1.25
536889001 | 41.1794337 | -8.6014974 | 1460465706 | 1.75
(10 rows)
Column seconds is the Unix timestamp. I would like to update the table by selecting ONLY one row, for rows with timestamps logged more than 1. So for example in above, we see two rows each at timestamp 1460465700 and 1460465703.

Answer 1

Without a unique id on the row, this is tricky. But assuming that the combination of values is unique, you can use:
update gps
set . . .
from (select gps.*, count(*) over (partition by id, seconds) as cnt,
row_number() over (partition by id, seconds order by seconds) as seqnum
from gps
) gps2
where gps2.cnt > 1 and pgs2.seqnum = 1 and
gps2.seconds = pgs.seconds and
gps2.id = gps.id and
gps2.speed = gps.speed and
gps2.lat = gps.lat and
gps2.lon = gps.lon ;
I would advise you to add a unique id to the table, so this is much simpler (and guaranteed to work even if the table has duplicates).