Hi, > You may try to_date(FROM_UNIXTIME(UNIX_TIMESTAMP()))
That would be a very bad idea for query correctness. The current_date UDF was introduced because of that particular anti-pattern. The unix_timestamp() is evaluated when a row is read, so each row gets a slightly different timestamp when doing ETLs & I have seen it give completely incorrect results when used near midnight (i.e query kicked off at 11:59, executed at 00:01 next day). Plus map-reduce falure handling mandates (logically) identical outputs on map task retries or speculation. I¹ve recently pulled out the current_timestamp() impl out into a reusable UDF recently so that I can use that in hive-1.0.x installs as well. https://github.com/t3rmin4t0r/current-timestamp You can use that UDF JAR with hive-1.0+, the constant folding will replace the value in-place during compilation. Cheers, Gopal
