Question

I have a dataset that appears to have time in an “elapsed” format (possibly in SAS format, where time is expressed as number of seconds elapsed since Jan 1st 1960). In a sql query I would like to convert this to a timestamp or date format. Is there a nice, clean way to do this?

Something similar to the answer here, except I need it in psql, not Python:
Time and date on the basis of seconds elapsed since 1970 year

Here is a sample of my data:

time_var      
-------------
1344444527000  
1344448596000  
1344455497000  
1344445125000  

Here is the type of query I’d like to make:

select
   time_var
   cast(time_var as timestamp) as mytimestamp,
   cast(time_var as date)      as mydatetime, 
   date(time_var)              as mydate 
from
   source_dataset
;

Answer 1

Turns out my dataset is using UNIX timestamps, so the values are big integers (the extra zeros need to be stripped) and the time is from Jan 1st 1970. I obtained the following code from my boss to convert these:

select
   timestamp with time zone 'epoch' + interval '1 second' *
      cast((cast(time_var as bigint) / 1000) as integer) as postgres_timestamp,
   date(timestamp with time zone 'epoch' + interval '1 second' *
      cast((cast(time_var as bigint) / 1000) as integer)) as postgres_date