CODEWITHSUNDEEP
hivehiveqlunix-timestampsql-timestamp
Rocking chief
Rocking chief

Reputation: 1069

Hive SELECT records from 1 hour ago

I have a hive table that contains a column called timestamp. The timestamp is a bigint field generated from java System.currenttimemillis(). I suppose it should be in UTC. Right now I am trying to select records from 1 hour ago. I know in MySQL you can do something like:

SELECT * FROM table WHERE datetimefield >= DATE_SUB(NOW(), INTERVAL 1 HOUR)

In hive, it seems like NOW() is missing. I did some searching and find unix_timestamp(). I should be able to get the current UTC time in milliseconds by doing a unix_timestamp()*1000.

So if i want to get records from 1 hour ago I am thinking about doing something like:

SELECT * FROM hivetable WHERE datetimefield >= (unix_timestamp()*1000-3600000);

Can someone suggest if it's the right way to approach this problem? Also what if I want to select like 1 day ago? Seems inconvenient to convert that to milliseconds. Any help or suggested readings will be highly appreciated. Thanks in advance for your help.

Upvotes: 0

Views: 3030

Answers (1)

nobody
nobody

Reputation: 11080

Yes unix_timestamp() gets you the seconds elapsed since Unix epoch. You can subtract 60*60*1000 milliseconds and compare your field to get the desired records.

For Hive 1.2.0 and higher you can use current_timestamp

select * 
from hivetable 
where 
     datetimefield >= ((unix_timestamp()*1000) - 3600000);

For 1 day,convert the milliseconds to date format and use date_sub 

select * 
from hivetable 
where 
     from_unixtime(unix_timestamp(datetimefield,'MM-dd-yyyy HH:mm:ss')) >= 
     date_sub(from_unixtime(unix_timestamp()),1);

Upvotes: 1

Related Questions